Tech2Pocket

SpaceX’s historic Falcon Heavy successfully launches | TechCrunch

Tue, 06 Feb 2018 12:53:31 +0800

SpaceX has had a very good first test of its Falcon Heavy rocket, the heavy-lift orbital vehicle that can carry twice the weight of its closest competition in active operation. The massive, three-booster rocket took off from Cape Canaveral in Florida as planned on Tuesday at 1:30 PM EST, lifting off from Kennedy Space Center’s LC-39A.

This is a historic moment for SpaceX, since it has been aiming to build and launch the large capacity rocket since 2011. Initially, the planned schedule had targeted 2013 as a launch window, but various delays and setbacks pushed down its inaugural flight – until today. Musk has previously described SpaceX’s early views on how hard it would be to build this rocket as “naive,” to account in part for the considerable departure from that early timetable.

The rocket, which has 27 Merlin engines (as many as three Falcon 9s) has a low-Earth orbit payload capacity of up to 140,700 lbs, or it can carry 58,900 lbs to geostationary transfer orbit (GTO), or 37,000 lbs to Mars. Those numbers are all important because SpaceX is planning Falcon Heavy as a key linchpin in its Mars plan, setting the stage for a system that can ferry goods and people between Earth and the red planet.

SpaceX’s payload for this mission includes a red Roadster destined for an Earth-Mars orbit, with a dummy on board wearing SpaceX’s new crew suit for its first astronaut missions, and a data storage device with Isaac Asimov’s classic space trilogy loaded up. There’s also a plaque inscribed with all SpaceX employee names, so they can tag along on the payload’s billion+ year journey.

The Kennedy launch complex used for this mission, which SpaceX has leased from NASA, is as historic as its latest resident – it was previously used for both the Apollo program and the Space Shuttle before being taken over and reconfigured for Falcon Heavy by SpaceX.

SpaceX also attempted a recovery of all three of the first stage boosters it used during the launch. It has recovered two of those thus far, and we’re waiting to hear back from SpaceX on the official status of the final, third booster, which was landing at sea.

So what now? From here, the rocket’s payload (read: the aforementioned Tesla, pictured below as the nosecone seperates) will pass through the Van Allen belt — a zone of charged particles and extreme radiation surrounding Earth. Assuming it makes it through that radioactive beatdown, it’ll head for a heliocentric orbit between Earth and Mars, from which point it’ll spend the next few hundred million years whizzing through space.

Update as of 1:50 PM Pacific: Here’s the current view of Earth as seen by the in-car astronaut dummy (“Starman”, as he’s known around SpaceX) and beamed back to us via this livestream. Note the “Don’t Panic” on the dash display, a nod to Douglas Adams’ Hitchhiker’s Guide to the Galaxy; there’s also, according to Elon, a towel in the glove box.

Mozilla announces an open gateway for the internet of things | TechCrunch

Tue, 06 Feb 2018 11:45:05 +0800

Apple, Google, Amazon and Samsung have all been working hard to create their own standard to control all the connected devices around your home. Mozilla just announced that anybody can now create an open gateway to control the internet of things. The organization also confirmed that it is still working on a set of frameworks and open standards so that we don’t end up with an internet of things controlled by big tech companies.

Connected devices are great, until you realize that your connected thermostat only works with the Amazon Echo and your connected lightbulbs only work with Siri and the Home app.

Accessory makers also don’t necessarily want a handful of big tech companies to control the internet of things. Tech giants could end up charging expensive licensing fees to work with their ecosystem. And customers end up having to make tough decisions.

Mozilla is a big proponent of the open web. So it seems natural that the not-for-profit organization has plans for connected devices. Project Things encompasses multiple projects, so let’s look at Mozilla’s work.

First, Mozilla wants to create an open standard with the W3C around the Web of Things. The idea is that accessory makers and service providers should use the same standard to make devices talk to each other. The specifications rely on JSON, and a REST and WebSockets API. Those are standard data and API models on the web, and they should work perfectly fine for connected devices.

Second, Mozilla is working on a Web of Things Gateway so that you can replace your Amazon Echo, Philips Hue hub, Apple TV and Google Home with an open device. You can already create a gateway using a Raspberry Pi 3, and ZigBee and Z-Wave USB dongles.

Eventually, manufacturers could leverage this work to create their own gateways. Maybe Netgear could embed a Web of Things gateway into their next router — your router is connected to the internet and runs 24/7 after all. Developers could also create bridges between the HomeKit API or Amazon’s Smart Home Skill API so that all devices work with your Amazon Echo, Google Home or iPhone without too much effort. Web of Things could become the common language between those proprietary APIs.

Finally, Mozilla is creating the interface to control your connected devices. You can add Mozilla’s progressive web app to your smartphone home screen and control your home. For instance, you can use your voice to turn on the lights, create IFTTT-style rules to automate your house, add a floor-plan to lay out your devices and more.

Mozilla has designed an add-on system so that you can add support for new devices and protocols by installing plugins. It’s important to note that all of this runs on your own gateway in your house. Google or Amazon can’t see when you turn on the light using your voice.

Eventually, I could also see app developers leveraging the Web of Things protocol to create native apps to control your house. But it’s clear that Mozilla wants to attack this issue from all angles. And developers can already start playing around with Project Things and contribute to development.

Quizlet – The World’s Largest User Generated Learning Platform

Wed, 07 Feb 2018 18:03:29 +0800

On the occasion of our portfolio company Quizlet‘s announcement of a new round of funding today, I thought I’d talk about what led us to Quizlet and why it is poised to be a game changer in education.

Quizlet is the world’s largest user generated learning platform. Founded in 2005 in founder Andrew Sutherland’s bedroom (he was 15 at the time), Quizlet has become the wikipedia of the education sector. Over the last twelve and a half years, users have posted over 200 million study sets to Quizlet. A study set is a list of things you want to learn. Think of what happens when flashcards meet the Internet and mobile devices. A study set can be French Adverbs or Heart Muscles or a Pre-Takeoff Flight Safety Checklist. You can study these things anytime and any place that you have a mobile phone on you.

That was all already in place when we invested in Quizlet in the fall of 2015. I wrote this post talking about Quizlet at the time.

What wasn’t there, and is now, is the machine learning team to make sense of all of these study sets. Take 200+ million study sets and 30+ million monthly users and you have a ton of data about what people are learning, how they are learning, and how their learning evolves over time on Quizlet. From that data will come new modes of learning on Quizlet and a lot of help figuring out what study sets are best to learn something.

The other place Quizlet plans to invest in is expanding Quizlet internationally. Over the last year, the Quizlet team has localized into 18 languages. Now, over 90% of the world’s population can use Quizlet. And because Quizlet is free to use for everyone, that means this user powered learning model can be used by students all over the world. Quizlet intends to aggressively expand its content base and user base internationally in the next few years.

Just because Quizlet is free for anyone to use doesn’t mean it is not a good business. Quizlet has two primary revenue streams right now, advertising and subscriptions, both of which are performing very well and there will be more revenue opportunities for Quizlet as they build out their content base and user base around the world. Quizlet was a profitable business when we invested in it back in 2015 and has remained at or near profitability even with the significant investment we have made in the business since then.

Quizlet is a great example of how you can build a very good business while expanding access to knowledge dramatically around the world. One does not have to come at the cost of the other if you architect your product and business model appropriately.

What I Learned about Management and Culture from Growing Uber’s Asia Business from Zero to Billions | First Round Review

Tue, 06 Feb 2018 00:00:00 +0800

When Mike Brown moved to Singapore in 2013 to start his new job overseeing Uber's business in Asia, he knew the next few years would be wild and unprecedented in his career. But he couldn’t have anticipated what came next. In the space of just four years, he went from managing 7 people to overseeing 1,500 full-time employees plus hundreds of contractors — and working with hundreds of thousands of drivers each week. Annual regional revenue jumped from near zero to billions. And his purview expanded from 2 to 15 countries with their own languages and regulatory regimes. It was an unpredictable, demanding ride — very analogous to that experienced by consumer-facing startups during growth.

Reflecting on this experience, he’s taken the step to write down the most impactful lessons he learned during his time at Uber — something that leaders like Brown seldom do. We’re excited to publish these insights here for the benefit of all founders who want to understand the actions and tactics it takes to build a high-growth business in tough markets.

The following knowledge comes from Brown’s first-person perspective. Because there’s so much specific advice packed into this one story, we’ve decided to divide it into multiple installments. Today, we’re publishing the first, covering Management, Organizational Effectiveness and Culture — three areas that made all the difference for Uber’s success in the region.

MANAGEMENT

Great managers stay involved in key decisions.

Good management is about more than setting clear goals and holding people accountable. Truly great managers get deeply and personally involved in key decisions and help shape the outcomes. I don't like it when an employee says, “My manager is great. She leaves me alone to do whatever I want and doesn't micro-manage.” I understand the sentiment here, but to succeed, managers must identify the key decisions or problems that are material to the business and they must to be personally drive results.

This means getting into the weeds on the things that really matter. Senior managers need to review the data themselves, speak directly to affected customers, and be active at meetings with stakeholders. Great management is not “set it and forget it.” The best managers know how to ‘porpoise’ — know what’s going on with everything at a surface level, but go deep on select initiatives.

So which decisions do you go deep on and which do you delegate? On a quarterly or semi-annual basis, I would go deep on the couple of projects that I thought were going to have the biggest impact on our region achieving its KPIs. I would go deep on one project tied to a people/culture KPI, one project tied to a top-line growth KPI, one project tied to cost reduction/efficiency KPI and one project most likely to improve customer experience.

Occasionally a problem or opportunity will arise that either poses an existential threat to your business or that could have a transformative positive impact. I was always hands-on in those circumstances. For example, when the Philippines government shut down our business for a month last year, I camped out in Manila and worked side by side with the team on the ground to figure out how to get cars back on the road and repair our then-strained relationship with the transport regulator.

Know when the stakes are too high to delegate, and don't hesitate to involve yourself in the full stack of that work.

Keep close watch on yourself.

Pay close attention to your own behavior, words and body language at all times — but especially during tough times. Whether it feels like it or not, people are looking to you for strength, comfort and confidence. Whatever fear or anxiety or frustration you feel during trying times, as the leader you need to squirrel it away and project positivity and confidence because that's what your people need most.

They will notice and be deflated by the smallest things. As the leader, you are the company and your words and actions define the culture and the mood of the company at all times. If you’re not feeling like you can project the right energy and will to win, take a break from the office and focus on rallying.

Your energy and attitude are self-fulfilling in terms of the fate of the company and the vibe and spirit of the team.

Leverage stories wherever possible.

Storytelling is bar none the most powerful way to share company history and to reinforce shared values. It's particularly helpful to share stories from the past with all employees often so they feel connected to the company's core and its arc of development. Everyone loves a compelling company origin story, or stories about the early original team — how bought in they were, how hard they were working, bonding in the trenches, the crazy or unorthodox ways they culled together resources, etc. I often talked about putting down my own credit card to rent cars in Singapore at the start of uberX. The first rental companies wouldn’t accept cash from local Singaporean drivers and many drivers didn't have their own credit cards — our scrappy approach at that moment inspired scrappy behavior long afterward.

All-hands meetings are a great time to tell these stories and to empower your early team to do the same. Veteran employees will feel recognized and appreciated, and new people will get a sense of what’s valued and expected within your culture.

Leaderboards are crazy effective.

Ranked lists are a surprisingly effective way to get people on your team focused on moving a metric. If a leaderboard is public, e.g. shared monthly over email or at an all hands or on a staff call, it gets everyone's competitive juices flowing and motivates them to take action (especially if they're currently at the bottom of the board).

I remember many times when a global leaderboard would be circulated over email and our region was at the lower third relative to others. It was an incredibly motivating force for both me and my team to rally and focus on a particular initiative to get the APAC region moving up and ahead — and to show everyone else at Uber that Asia was a well-managed force to be reckoned with. It taps into people's pride in their own work and their work as a team. Harness that energy.

One-on-ones need to be a two-way street.

Managers need to come armed with questions about the business to their 1:1s based on data they’ve reviewed in advance — both qualitative and quantitative. If you don’t do this, you’re only getting the part of the story that your direct report thinks is important. That may or may not be where the real issues and opportunities lie. You’re optimizing within a single frame of reference (the perspective of your direct report) and that’s not good.

If you’re a manager, you need to look at your report’s sphere of influence, and come prepared to push on various aspects of their work and what they touch — particularly any shortcomings you see in the data or learn via other people.

Having this kind of visibility requires accessing the right metrics reporting so you can examine the data yourself and develop your own questions and perspective. If you don’t have this kind of trackable data, you won’t add as much value during these meetings.

If you rely exclusively on your report’s version of how things are going, you’re sure to miss other important factors. You should constantly be collecting data about their experience and performance from skip-level meetings, customer interviews, and the business metrics your reports impact to get a greater sense of the story and ask them the right questions during your precious and rare time 1:1.

Skip-level 1-on-1 meetings are a critical tool.

If you manage managers, it’s very powerful to set up a recurring catch up with their direct reports. These are commonly referred to as “skip-level” meetings, and they're the only way to get an accurate sense of how your direct reports are leading — and how people on their teams are feeling about the business and the strategy you've set. It's also an efficient way to uncover the obstacles that are preventing teams from executing so that you, as the skip-level manager, can work to unblock them in ways their direct manager can't. When teams are running hard, these meetings often get deprioritized or forgotten. They're too important for that. Make them a priority.

Make most meetings about decisions.

If you want people to show up to a meeting, make sure an important decision will be made. Communicate that this is what the meeting is about in advance and be specific about what you want to decide before everyone leaves the room. People are less likely to forego their decision rights. If no decision is being made, why have the meeting?

One of the first things I like to do in a meeting is get clarity on what decision is being made and who has the decision rights. If the answer to either is vague or unclear, you should cancel the meeting and reconvene when you have clarity on these two points. If there’s ambiguity as to who has decision making rights for a particular topic, it may speak to a broader ack of clarity on decision making and ownership that needs to be addressed.

Make managers accountable (for the good and the bad).

When bad things happen in the office, the team looks to you as the senior manager to take action that’s consistent with the company’s values. Your actions to handle an employee relations matter (let’s say a case of bullying or sexual harassment) speak volumes about what kind of culture you want to set and what kind of company you aim to build. My advice: communicate and train people in what good behavior means and then hold a zero tolerance policy toward behaviors that violate the cultural values you’ve set as a leader. All of your managers should follow suit.

How do you set the right tone for the behavior you expect from people? Define it clearly and communicate it often.

When someone crosses the line, they should be held accountable — and so should their manager.

If a manager’s employee is exited from the company for acting inappropriately, the manager must suffer some negative consequences for his or her employee’s actions. Yes, that’s right. Even if the manager was not the instigator or even physically present for the bad act by the employee, the manager must be held accountable with a financial penalty, demotion, or possibly termination if the situation is severe enough.

I used to think that it was unfair to penalize a manager for an employee’s action. After all, how can the manager “control” what the employee does? But over time I realized, it's a core management responsibility to be clear and forceful enough about what's acceptable and unacceptable so that all employees know and understand. If you're not doing this well, you're not doing your job well.

Holding managers accountable for the actions of their employees is the only way to create the right incentives for them to set and enforce the appropriate tone. That doesn’t mean work shouldn’t be fun or people shouldn’t have a great time together. But, it does mean managers are on the hook for how their employees behave and making sure that they're good ambassadors for the company, its brand and broader culture at all times.

Let long-term perception shape present decisions.

Decisions have a long reputation shelf life, and the more successful your company is, the longer their shelf life will be.

Decisions that seem okay or like not a big deal early on may not look so good when you're a bigger company. You have to think, act and plan for success and longer-term reputation. If you're successful, the decisions you make today will reflect on you and your reputation for years to come. I’ve found this to be true in terms of defining and living company culture as well as in the business and strategy decisions and tactics I pursued.

At Uber in Asia, we developed a habit when making important decisions to ask ourselves: “How is this going to look on the cover of the newspaper in two years’ time?” We found this to be a helpful framing to make sure we'd be proud of whatever decisions we were making at the time.

Evaluate managers with upward feedback and culture surveys.

Managers must be evaluated both on what they get done and how they get it done. The “how” is often overlooked and undervalued. In terms of the “how,” determine whether the person works well cross-functionally, represents the company‘s values in both word and deed, receives strong upward feedback, and whether their team of reports has a high level of satisfaction working at the company.

To measure these attributes, it’s very helpful to run a twice-yearly upward feedback survey and a twice-yearly culture survey (in alternating quarters). Both of these allow you to take the pulse of how people are feeling about the company and their place in it.

Upward feedback surveys should ask people:

Does your manager inspire you? How? (Sometimes he answer is no.)
Does your manager represent company values?
Does your manager set clearly-defined goals?
Does your manager provide useful, action-oriented feedback?
Does your manager provide professional development or coaching?

At Uber, I would review the upward feedback that employees gave to each manager on my team, and my manager would review the feedback provided by my direct reports about how I was doing. We would collect and review the upward feedback twice per year. The results of a single reporting period were super interesting and the trajectory of a manager’s upward feedback over time was very valuable to track improvement or regression.

Ultimately, each manager is responsible for creating a great (fun, inclusive, safe) working environment and enabling each employee to grow and succeed. Each manager is responsible for the happiness and success of his/her direct reports and the happiness and success of all of their reports down the chain. If a manager on my team has people that feel they aren't creating an inclusive or safe culture, I’m accountable because I’m the leader of the whole team and I need to do a better job training and coaching them to create a great environment.

Culture feedback surveys should ask people:

To what extent is your experience at work consistent with what’s stated in company-wide cultural values?
Do you perceive that your peers and managers conduct themselves in accordance with these values?
Are there any cultural values that you think are missing or should be added?
Do you see any behavior that's inconsistent with stated values? If so, can you share examples?

We started running culture surveys in the quarters that we didn’t run upward feedback surveys — also twice per year. These surveys were not specific to individual manager performance but rather sought to evaluate whether our company was living the values we espoused. In other words, did we “walk the talk.” In cases where employees felt we were missing the mark, we could dig in and address negative situations very specifically.

Be there in the tough, dig-deep times.

As a manager, you better be there in person for the fire drills, the emergencies, the weekend envelope stuffing sessions, and all the times the team has to go above and beyond.

When we got shut down in the Philippines, I cancelled my planned vacation and got on a plane to Manila to be there with the team. It was important to show the people on the ground that they weren’t facing the biggest crisis of their existence at Uber alone, and it was important to demonstrate to local government officials that the leader of the region cared enough about the Philippines to show up and deal with them personally to make things right.

I was the leader, so people looked to me as the voice and presence of the company. Even if I didn’t have unique expertise to bring to bear on a problem, or I could delegate to someone well positioned to solve the problem themselves, I made a habit of being there in person for the tough times. Not to necessarily take control, but to show support. So show up and don't shirk the busy or grunt work. Make your presence felt and don't act too senior to pitch in and help with anything.

Make a “management view” or “operational view” of the P&L accessible to all managers.

Often, your profit and loss sheet won't show you what you actually need to know. A standard P&L doesn't provide a ton of insight into how the work of individual people or teams impacts profits and losses. Also, a P&L doesn’t always make it clear what specific activities are the most or least efficient — or where there’s opportunity to cut costs and do more with less. You need to create a different type of P&L view that gives you more leverage.

For example, expenses to send SMS messages to riders and drivers got lumped into the marketing line item on our P&L. When we broke out telecomm fees seperately, we realized we were paying especially high per-message fees in certain countries. Only by seeing this, were we able to bring those fees under control. Creating a new manager view of our P&L (sometimes referred to as an activity-based P&L) gave the entire mangement team a powerful tool to openly discuss and change what was being spent and why. This also changed my managers' mindset — they felt more responsibility to be efficient within the context of the entire regional business.

Align around communication between teams.

If your company has a matrix structure — i.e. people report to both project leaders and department leaders — communication can get thorny fast, especially when you have multiple offices, like at Uber. You want to make sure that stakeholders are informed of important decisions at all times, that your entire team is aligned on what to share with whom, and that you’re not selectively sharing with some stakeholders but not others.

Communication channels and who gets brought into what decisions should be part of your decision rights framework. Sit down with your senior team or peer group and be explicit about who needs to make what decisions and who needs to be notified when decisions are made.

I used to make decisions about Operations related matters, but I didn’t proactively share certain key decisions with my peers in other functions. That created big problems for them and for me. I wasn’t trying to be political; I just didn’t understand how the decisions I was making were going to affect other orgs.

We ultimately adopted the “RACI” framework to define who is responsible for the decision, who is accountable, who is consulted and who needs to be informed. By applying a clear decision rights framework, we created a governance structure that people could buy into. It ensured that the right people received the right information to do their jobs or to support the work of others that impacted their jobs.

Don’t lose touch with HQ — and if you’re in HQ, get out to the field.

If you’re a manager working in a remote office, get back to HQ in person at least once per quarter. It’s extremely important to learn what’s going on and invest in-person in the relationships that you’re dependent upon to get stuff done. This is especially important if you need central product or engineering support for your regional priorities.

In person, I was able to make a much stronger case for our region’s priorities and gain a better understanding of the pressures that others in HQ were under, including the competing priorities they were juggling. I was able to take what I learned at HQ and share that information with my teammates in Asia. When you work in a remote office, HQ can feel far away, and employees struggle for connection to the company and its management. They need to feel it, and you can help with that as a conduit.

The flip side of this advice is that senior managers from HQ must visit employees in satellite offices often. It’s a really powerful boost for remote office team morale and team culture, and an important way to show appreciation for the work people are doing for the company outside of HQ.

Fortify your cross-functional communication to anticipate major problems.

I recall one particular crisis that popped up in our region — a vehicle recall. We kicked off a discussion thread about how to solve the problem, but a smaller subset of people (including me) ended up developing an action plan by phone. We didn’t do a good job of then circling back with the other cross-functional leaders to clearly articulate the final decision and action plan. Later, it turned out that some of them felt marginalized or excluded, and others thought we had really mishandled the situation because they didn’t have any context. That was my fault for not bringing people along for the ride together.

If cross-functional leaders don’t have full-visibility into every big decision or plan, you’ll create big problems that are hard to reverse.

In the case of the vehicle recall, I remember receiving a phone call from my boss at HQ because he'd heard about the issue through his peer who had heard about it from someone else entirely. This type of phone tag is both too common and extremely costly. It risks creating thrash when messages get distorted or change. I had to spend considerable time unwinding false narratives and setting the record straight because I didn’t include the right group of leaders in my region from the outset. It’s much more efficient to assemble all the leaders, walk them through the situation, and plan for remediation.

ORGANIZATIONAL EFFECTIVENESS

Every meeting must have a clear goal. Every decision must have a clear owner.

Enough said.

Big changes and launches require cross-functional awareness.

Every time there’s a strategic change or announcement — a feature launch, a pricing shift, entrance into a new market — you need to assemble your cross-functional leaders far ahead of time to not only give them the heads up, but to let them stress test the thinking behind the plan itself. This will help you avoid sloppy mistakes and last-minute fire drills around issues like taxes, PR and other snafus. You want to create a checklist to ensure that product launches and operational changes get in front of the right cross-functional audience at the appropriate time. You want all of them working as one unit to support an overall goal.

There will always be many unknown unknowns when a change or announcement of this magnitude comes up. For example, the introduction of a new product might have tax consequences, require explanation to third-party partners or regulators, the media must be briefed in a particular way, and customer support needs to be lined up to handle new inquiries. The list goes on.

Give the experts on your team who know more about your unknowns the chance to do their jobs and cover your bases.

Cross-functional awareness should never require unanimity in decision making.

Big bold bets require courage. Ideas that make impact can sound outlandish or far-fetched to skeptics at first. If you strive for unanimity, you’ll take fewer risks. This is why good decision making frameworks require agreeing on who owns different classes of decisions — giving one person the authority to make the final call. You also want your team to unanimously establish and prioritize a “disagree and commit” culture. This empowers decision makers to make a call and know everyone will back them up and help implement once it’s done. If you don’t institute this very clearly and have everyone agree to it, mis-alignment on decisions will create confusion, thrash and politics.

Create a decision-making framework and stick to it.

In matrix organizations and at really early-stage companies with nebulous reporting structures, it’s really difficult to maintain alignment, set goals, and create a common set of priorities. When bad stuff happens, people run to their boss to cover their own ass, and that can really damage teamwork and collaboration. At a minimum, you need a clear decision making framework so everyone understands how decisions — especially controversial or contentious ones — are made. Who gets the final say? How are disputes handled? What decisions require group input across functions versus decisions that can be made within a single team or function? Create ground rules around all of this.

At Uber in Asia, we would divide decisions up into three categories:

Those that need to be made with the involvement of the whole leadership team (or rather, with helpful input from the whole team)
Those that only need a sub-group involved (ideal to go faster)
Those that should be made by an individual leader

Early in any decision-making process, we would make sure to communicate with everyone who might be touched by the result so we could suss out who had an opinion they wanted to contribute. Decisions were never held up by this, but it did give us a chance to make everyone feel heard. Informing people early that a decision is coming makes it more likely they’ll happily live with the outcome regardless. They just don’t want to be surprised.

Reporting relationships work much better when they're local.

Reporting and decision making across miles and oceans is crappy because so much context is lost and everything slows down — no matter how hard you try. If you can't avoid long-distance management relationships, make sure the local or on-site manager is sufficiently senior and experienced so that they can handle almost everything locally, and have the trust of people’s actual managers to make the right decisions.

Never assume healthy information flow.

Information shared by leadership doesn’t always make it all the way to junior employees.

It’s safe to assume most information doesn’t flow down through the org in the way or with the clarity that you want unless you’re really focused on it.

You have to be extremely proactive about sharing company goals, strategy, successes and failures repeatedly and in a way that reaches everyone, without relying on cascading information through tiers of management. Good leaders hold Q&A sessions with different groups within the company — like junior employees specifically, support personnel, etc. Maximize your regular face time with as many people internally as possible.

Choose a single point of contact by function.

It’s incredibly helpful if each function or department in the company has a single point of contact for a business owner to interface with — this vastly simplifies coordination and accountability. Also, a single point of contact that's a functional expert can see the big picture and maintain context of the overall situation the business faces.

Alternatively, if the business owner needs to juggle multiple relationships and expertise sets within a functional team, that adds up to a lot more complexity for the business owner, and each “specialist” on the functional team only sees his or her part of the puzzle without full context.

Put every employee on the front line.

In the old days at Uber, the team that ran a given city would process all the customer support tickets themselves. Over time we outsourced a significant amount of this work — but still had it managed by Uber employees who set policies and maintained quality and reliability of support. The benefit was that we could handle support inquiries faster, 24 hours a day in every language that we needed to speak comfortably with our customers. The downside was that our city teams became less connected to the issues raised by our customers — and the great suggestions our customers would provide about product or service improvements.

As a manager, if I didn’t make a concerted effort to get in front of customers, I could easily lose touch.

I learned how important it is to expose every employee, regardless of their function or title, to front-line service delivery, interfacing directly with users or customers. This meant processing support tickets, shipping products, handling calls, and gathering feedback via 1:1 customer conversations or focus groups. This generated customer empathy in everyone, not just those who were on the front-lines of product and service delivery as their full-time jobs. I used to think about the famous example of senior executives at McDonald’s working in a restaurant making french fries or working the cashier, and I gained new appreciation for how smart that was.

It’s critically important to be able to represent the voice of the customer in all decision-making, and there’s no better way to so that than to personally work with customers and to listen their experiences. Nobody is too senior or too busy to get in front of customers regularly, and at Uber in Asia we designed new employee onboarding processes to ensure that as many employees as possible experienced the front lines of customer service delivery. I also made a point to personally visit our customer support centers and physical retail centers (we called them Greenlight Hubs) regularly to interact with customers myself and channel that back to my employees.

Get your management team together quarterly.

This is extremely important and can’t be casually skipped because people are short on time. The goal is to create alignment, but also to have fun and renew relationships.

In the heat of battle, if there isn't excellent trust and genuine respect among cross-functional leaders, you've got a real problem.

This is as true at every level of your org. You need to put gas in the relationship tank for the team (especially cross-functional teams where people are dependent on each other but not necessarily accountable to each other) on a regular basis — at least quarterly. Offsites are a good opportunity to have facilitator-led discussions and foster fierce feedback discussions where people can feel safe being candid.

If offsites happen at all, they typically focus on goal-setting and strategy, but they often miss the interpersonal dynamics element that’s so important to forming and maintaining an awesome team. Investing in these dynamics requires collecting data on how people are truly feeling about working with others on the team and setting the right environment and ground rules for sharing that data in a non-threatening and productive manner.

All feedback sharing needs to be coupled with trust building, or it’s just going to create divisiveness.

At Uber in Asia, we hired a consultant who interviewed each member of my leadership team to truly understand what each person’s views were on how the team was working together, where there was friction, and where there were opportunities for improvement. This consultant also held focus groups with an assortment of employees to discuss how well the leadership team was functioning and where there was room for improvement. And finally, they moderated a two-day offsite for us leaders so we could review the data collected on employee culture and upward feedback. We got a sense of their overall work satisfaction, happiness with management, clarity on goals for our region, and more.

As part of the offsite, we each roleplayed another member on our leadership team and explained to the group what challenges we were facing and how the other members of the group could best support us. Walking in another leader’s shoes fostered empathy and gave us greater insight into the unique pressures we were each under and why we made certain decisions or had certain reactions.

CULTURE

Know and respect your team’s sacred spaces.

Ask each member of your team to share what their sacred space is in life — what’s something that gives their life meaning outside of work that they feel comfortable telling people about? Make sure they have the time to pursue that outside activity consistently. Actually tell them to block out this time and stick to it.

For me, it was leaving the office every day at 6 p.m. to spend 6:30 to 7:30 at home with my kids. I would help with dinner and bedtime. I wouldn’t take calls during this hour, but I would start work back up again around 7:30 or 8. This was my sacred space.

I found that by defining and talking openly and often about my sacred space, others scheduled around it —and felt better about having their own.

Each team member should define and communicate his or her sacred space at the beginning of a new project or at the beginning of a new manager-report relationship. As a manager, a core part of your job is to know and support the sacred space of those around you.

Make vacation a cultural norm.

Force everyone to take vacation. Leaders need to set a good example. It's important for longevity, to avoid burnout and for people to feel comfortable going away and coming back with more creativity and joy for their work. So much of culture is about parroting the behaviors and words of senior management. As one of these senior managers, you want your people to find a degree of work-life balance so that everyone can work hard in a sustainable way.

We used to say at Uber that we’re sprinters but we’re running a marathon. To carry on at high intensity for long periods of time requires regular breaks and vacations. Make it not only okay, but encouraged to take vacations in your organization by setting the right example and telling other leaders to do the same.

Proactively address your blind spots.

I knew my experience working in the organization as a senior white male was very different from someone else’s experience in the org who might have been more junior than me or from a different background. Never assume that your experience of company culture and point of view in the org is the same as someone else’s.

Figure out what your blind spots are by conducting skip-level (or skip-multiple-level) meetings regularly and listening sessions with people who are likely to have a very different perspective on what it’s like to work there. Make sure you have a prior relationship with this individual in order to elicit candid and critical feedback about their experience working in the organization. Ask about:

Morale in the office from their point of view.
How they are feeling about his or her team.
What their manager is doing well and not well.
What obstacles they’re facing in their job. (If you are their skip-level manager, take steps to unblock these obstacles — it will mean a lot to the person that you took action.)
Do you understand the company’s goals? Your team’s goals?
Do you feel like you can do things you believe are right for the business?
Do you think leadership acts consistently with your values?
What would make work better for you?
When was the last time you took a vacation?
What is your sacred space? Do you feel like you have time for it?

This is how you’ll overcome your blind spots. You have to work to not lose touch with how others are perceiving and experiencing the organization. I made a regular habit of holding monthly sessions with all of the women in my region to listen to their viewpoints, as well as monthly sessions with the junior members of the Operations organization to answer questions and take feedback. These sessions were invaluable to me in order to see my cultural blind spots. Regular “walk and talk” sessions with people around the office you don’t work with closely are another valuable way to unlock blind spots and learn new perspectives.

Transparency and consistency build trust.

Consistency starts with clear goals and a clear strategy. If you don’t know your organization’s short-term or longer-term goals and strategy, the organization will likely be paralyzed because it will lack a common framework from which to make decisions. But having clear goals and strategy is just the beginning.

You need to communicate your goals and strategy frequently and in a broad variety of forums. And importantly, they shouldn’t change much.

As one of Uber’s respected former execs used to say, “Repetition never spoils the prayer.”

Transparency and consistency mean the most to people when you're addressing difficult or sensitive topics. For example, explaining how and why someone left the organization when they did (the rumor mill becomes a big distraction when departures are not addressed). The same goes for responses to big competitor announcements. People want to know how leadership is going to react to external shocks or changes in the landscape. Proactively address social issues that arise in the industry around topics like fair pay, gender discrimination, and diversity and inclusion. Be candid about how the company is performing in these areas. If you don’t share diversity stats internally, people will assume you’re hiding something.

Appreciate functions that are often overlooked.

Certain functions rarely get the appreciation they deserve. When things go bad, they get mud thrown at them. When things go well, nobody says anything. This is especially true for Legal, Customer Support and Finance. To be a great business leader, proactively and regularly thank and recognize the accomplishments of the traditionally unsung heroes of the organization.

I learned this lesson when one of the functional leaders on my team let me know I was always recognizing great efforts in our organization, but I often overlooked his function and that members of his team had felt slighted. I immediately began to include often overlooked teammates into my regular recognition efforts. Don't wait to be given this feedback.

Make it an operational discipline to do more with less.

This needs to be an ongoing effort by all team members.

It should be a continuous and top-of mind priority for everyone to pull cost out of the business while maintaining — and hopefully improving — service delivery.

Managers must be relentless about maintaining focus on operational efficiency. It's either part of your culture or it’s not. Make it part of your culture by rewarding it whenever it happens, even on a small scale.

One way to promote operational discipline is to force managers to segment and prioritize spend requests during headcount planning and budgeting cycles. Segmentation of spending priorities brings clarity that’s really helpful for getting managers to do more with less by cutting out all (or many) of the expenditures of lower priority initiatives or hires.

Evaluating our expenditures as a percentage of bookings or revenue, and comparing the percentages to industry benchmarks or competitive benchmarks was another helpful analysis to determine where we were spending more efficiently than industry norms, where we were choosing to over-invest, and where we were simply being inefficient.

This is the first of multiple installments of advice from Mike Brown gleaned during his time running Uber's business in Asia. Stay tuned for more on Thursday covering recruiting, performance management and more.

Images courtesy of S3/Getty Images and Sam Yeh/AFP/Getty Images.

AI Just Learned How to Boost the Brain's Memory | WIRED

Tue, 06 Feb 2018 16:00:00 +0800

When it comes to black boxes, there is none more black than the human brain. Our gray matter is so complex, scientists lament, that it can’t quite understand itself.

But if we can’t grok our own brains, maybe the machines can do it for us. In the latest issue of Nature Communications, researchers led by University of Pennsylvania psychologist Michael Kahana show that machine learning algorithms—notoriously inscrutable systems themselves—can be used to decode and then enhance human memory. How? By triggering the delivery of precisely timed pulses of electricity to the brain.

Researchers, in other words, can use one black box to unlock the potential of another. Which on one hand sounds like a rather elegant solution to an absurdly difficult problem, and on the other sounds like the beginning of a techno-pocalypse horror flick.

When it comes to brain measurements, the best recordings come from inside the cranium. But people—and institutional review boards—aren’t usually amenable to cracking open skulls in the name of science. So Kahana and his colleagues collaborated with 25 epilepsy patients, each of whom had between 100 and 200 electrodes implanted in their brain (to monitor seizure-related electrical activity). Kahana and his team piggybacked on those implants, using the electrodes to record high-resolution brain activity during memory tasks.

Machine learning algorithms learned to associate patterns of electrode measurements with a patients' likelihood of memorizing a word.

Kahana et al.

First, the researchers got a sense of what it looks like when a brain memorizes stuff. As the patients read and attempted to internalize lists of words, Kahana and his team gathered thousands of voltage measurements per second from each of the implanted electrodes. Later, they tested the patients’ recall—building up data about which brain activity patterns were associated with remembering a word vs. forgetting it.

Then they did it again. And again. After two or three visits with each test subject, they'd collected enough training data to produce patient-specific algorithms that could predict which words each patient would likely remember—based on their electrode activity alone.

Here’s the kicker. These electrodes don’t just read neural activity; they can stimulate it, too. So the researchers tried prodding the brain to improve—or, as they put it, "rescue"—the formation of memories in real time. Every few seconds, the subject would see a new word, and the newly trained algorithm would decide whether the brain was ready to remember it. “A closed loop system lets us record the state of the subject's brain, analyze it, and decide whether to trigger a stimulation, all in a few hundred milliseconds," Kahana says.

And it worked. The researchers' system improved patients' ability to recall words by an average of 15 percent.

This isn’t the first time Kahana’s lab has explored the impacts of brain stimulation on memory. Last year, the group showed that electrode pulses seemed to improve or worsen recall, depending on when the researchers delivered them. In that study, test subjects scored higher when the researchers stimulated memory-specific regions of the brain during periods of low functionality (stimulation during high-functioning times had the opposite effect). It was a major finding, but therapeutically useless; the researchers could only identify the link between memory and brain states after the memory tests were performed. What you really want, from a brain-enhancement standpoint, is to deliver pulses in the middle of memorization.

Now, Kahana and his colleagues appear to have closed the loop with the help of their machine learning algorithm. "Only instead of using it to identify images of cats, we're using it to build a decoder—something that can look at electrical activity and say whether the brain is in a state that's conducive to learning," Kahana says. If the brain looks like it’s encoding memories effectively, the researchers leave it alone. If it isn't, their system quickly delivers electrical pulses to jostle it into a higher-functioning state—like a pacemaker for the brain.

"It's not a whomping effect, but it's definitely promising," says UC San Diego neuroscientist Bradley Voytek, who was unaffiliated with the study. The question now is whether future work in this area will yield better results. If patients' brains were implanted with more—and more precise—electrodes, algorithms could decode more neural signatures, with more specificity, on smaller time scales. More training data could help, too; most patients with epilepsy are only able to participate in studies like this one for a few weeks at most, which limits the time researchers can spend with them. A machine learning algorithm trained on more than three sessions might perform better than the ones in Kahana's latest study.

But even with higher resolution and more training data, scientists will need to grapple with the implications of using opaque algorithms to study—and manipulate—brains. The fact remains that while Kahana’s system can improve word recall in specific circumstances, he doesn’t know exactly how it’s improving function. That’s the nature of machine learning.

Luckily, Kahana's team has thought this through, and some algorithms are easier to scrutinize than others. For this particular study, the researchers used a simple linear classifier, which allowed them to draw some inferences about how activity at individual electrodes might contribute to their model's ability to discriminate between patterns of brain activity. "We can't really say at this point if there are interactions between the features that we’re using to record brain activity," says UPenn psychologist Youssef Ezzyat, who oversaw the study's machine learning analyses.

More complicated deep-learning techniques won't necessarily translate to bigger cognitive enhancements. But if they do, researchers could wind up struggling to make sense of the machines’ decision to deliver brain-boosting electrical impulses. Or—if they become truly diabolical—withhold them.

My Accessibility Journey: What I’ve Learned So Far · An A List Apart Article

Mon, 05 Feb 2018 16:18:50 +0800

Last year I gave a talk about CSS and accessibility at the stahlstadt.js meetup in Linz, Austria. Afterward, an attendee asked why I was interested in accessibility: Did I or someone in my life have a disability?

Article Continues Below

New Drupal or WordPress project? Get the fastest hosting with HTTPS and global CDN included. Develop for free, pay when you go live. Learn more

I’m used to answering this question—to which the answer is no—because I get it all the time. A lot of people seem to assume that a personal connection is the only reason someone would care about accessibility.

This is a problem. For the web to be truly accessible, everyone who makes websites needs to care about accessibility. We tend to use our own abilities as a baseline when we’re designing and building websites. Instead, we need to keep in mind our diverse users and their diverse abilities to make sure we’re creating inclusive products that aren’t just designed for a specific range of people.

Another reason we all should think about accessibility is that it makes us better at our jobs. In 2016 I took part in 10k Apart, a competition held by Microsoft and An Event Apart. The objective was to build a compelling web experience that worked without JavaScript and could be delivered in 10 kB. On top of that, the site had to be accessible. At the time, I knew about some accessibility basics like using semantic HTML, providing descriptions for images, and hiding content visually. But there was still a lot to learn.

As I dug deeper, I realized that there was far more to accessibility than I had ever imagined, and that making accessible sites basically means doing a great job as a developer (or as a designer, project manager, or writer).

Accessibility is exciting

Web accessibility is not about a certain technology. It’s not about writing the most sophisticated code or finding the most clever solution to a problem; it’s about users and whether they’re able to use our products.

The focus on users is the main reason why I’m specializing in accessibility rather than solely in animation, performance, JavaScript frameworks, or WebVR. Focusing on users means I have to keep up with pretty much every web discipline, because users will load a page, deal with markup in some way, use a design, read text, control a JavaScript component, see animation, walk through a process, and navigate. What all those things have in common is that they’re performed by someone in front of a device. What makes them exciting is that we don’t know which device it will be, or which operating system or browser. We also don’t know how our app or site will be used, who will use it, how fast their internet connection will be, or how powerful their device will be.

Making accessible sites forces you to engage with all of these variables—and pushes you, in the process, to do a great job as a developer. For me, making accessible sites means making fast, resilient sites with great UX that are fun and easy to use even in conditions that aren’t ideal.

I know, that sounds daunting. The good news, though, is that I’ve spent the last year focusing on some of those things, and I’ve learned several important lessons that I’m happy to share.

1. Accessibility is a broad concept

Many people, like me pre-2016, think making your site accessible is synonymous with making it accessible to people who use screen readers. That’s certainly hugely important, but it’s only one part of the puzzle. Accessibility means access for everyone:

If your site takes ten seconds to load on a mobile connection, it’s not accessible.
If your site is only optimized for one browser, it’s not accessible.
If the content on your site is difficult to understand, your site isn’t accessible.

It doesn’t matter who’s using your website or when, where, and how they’re doing it. What matters is that they’re able to do it.

The belief that you have to learn new software or maybe even hardware to get started with accessibility is a barrier for many developers. At some point you will have to learn how to use a screen reader if you really want to get everything right, but there’s a lot more to do before that. We can make a lot of improvements that help everyone, including people with visual impairments, by simply following best practices.

2. There are permanent, temporary, and situational impairments

Who benefits from a keyboard-accessible site? Only a small percentage of users, some might argue. Aaron Gustafson pointed me to the Microsoft design toolkit, which helped me broaden my perspective. People with permanent impairments are not the only ones who benefit from accessibility. There are also people with temporary and situational impairments who’d be happy to have an alternative way of navigating. For example, someone with a broken arm, someone who recently got their forearm tattooed, or a parent who’s holding their kid in one arm while having to check something online. When you watch a developer operate their editor, it sometimes feels like they don’t even know they have a mouse. Why not give users the opportunity to use your website in a similar way?

As you think about the range of people who could benefit from accessibility improvements, the group of beneficiaries tends to grow much bigger. As Derek Featherstone has said, “When something works for everyone, it works better for everyone.”

3. The first step is to make accessibility a requirement

I’ve been asked many times whether it’s worth the effort to fix accessibility, how much it costs, and how to convince bosses and colleagues. My answer to those questions is that you can improve things significantly without even having to use new tools, spend extra money, or ask anyone’s permission.

The first step is to make accessibility a requirement—if not on paper, then at least in your head. For example, if you’re looking for a slider component, pick one that’s accessible. If you’re working on a design, make sure color contrasts are high enough. If you’re writing copy, use language that is easy to understand.

We ask ourselves many questions when we make design and development decisions: Is the code clean? Does the site look nice? Is the UX great? Is it fast enough? Is it well-documented?

As a first step, add one more question to your list: Is it accessible?

4. Making accessible sites is a team sport

Another reason why making websites accessible sounds scary to some developers is that there is a belief that we’re the only ones responsible for getting it right.

In fact, as Dennis Lembree reminds us, “Nearly everyone in the organization is responsible for accessibility at some level.”

It’s a developer’s job to create an accessible site from a coding perspective, but there are many things that have to be taken care of both before and after that. Designs must be intuitive, interactions clear and helpful, copy understandable and readable. Relevant personas and use cases have to be defined, and tests must be carried out accordingly. Most importantly, leadership and teams have to see accessibility as a core principle and requirement, which brings me to the next point: communication.

5. Communication is key

After talking to a variety of people at meetups and conferences, I think one of the reasons accessibility often doesn’t get the place it deserves is that not everyone knows what it means. Many times you don’t even have to convince your team, but rather just explain what accessibility is. If you want to get people on board, it matters how you approach them.

The first step here is to listen. Talk to your colleagues and ask why they make certain design, development, or management decisions. Try to find out if they don’t approach things in an accessible way because they don’t want to, they’re not allowed to, or they just never thought of it. You’ll have better results if they don’t feel bad, so don’t try to guilt anyone into anything. Just listen. As soon as you know why they do things the way they do, you’ll know how to address your concerns.

Highlight the benefits beyond accessibility

You can talk about accessibility without mentioning it. For example, talk about typography and ideal character counts per line and how beautiful text is with the perfect combination of font size and line height. Demonstrate how better performance impacts conversion rates and how focusing on accessibility can promote out-of-the-box thinking that improves usability in general.

Challenge your colleagues

Some people like challenges. At a meetup, a designer who specializes in accessibility once said that one of the main reasons she loves designing with constraints in mind is that it demands a lot more of her than going the easy way. Ask your colleagues, Can we hit a speed index below 1000? Do you think you can design that component in such a way that it’s keyboard-accessible? My Nokia 3310 has a browser—wouldn’t it be cool if we could make our next website work on that thing as well?

Help people empathize

In his talk “Every Day Website Accessibility,” Scott O’Hara points out that it can be hard for someone to empathize if they are unaware of what they should be empathizing with. Sometimes people just don’t know that certain implementations might be problematic for others. You can help them by explaining how people who are blind or who can’t use a mouse, use the web. Even better, show videos of how people navigate the web without a mouse. Empathy prompts are also a great of way of illustrating different circumstances under which people are surfing the web.

6. Talk about accessibility before a projects kicks off

It’s of course a good thing if you’re fixing accessibility issues on a site that’s already in production, but that has its limitations. At some point, changes may be so complicated and costly that someone will argue that it’s not worth the effort. If your whole team cares about accessibility from the very beginning, before a box is drawn or a line of code is written, it’s much easier, effective, and cost-efficient to make an accessible product.

7. A solid knowledge of HTML solves a lot of problems

It’s impressive to see how JavaScript and the way we use it has changed in recent years. It has become incredibly powerful and more important than ever for web development. At the same time, it seems HTML has become less important. There is an ongoing discussion about CSS in JavaScript and whether it’s more efficient and cleaner than normal CSS from a development perspective. What we should talk about instead is the excessive use of <div> and <span> elements at the expense of other elements. It makes a huge difference whether we use a link or a <div> with an onclick handler. There’s also a difference between links and buttons when it comes to accessibility. Form items need <label> elements, and a sound document outline is essential. Those are just a few examples of absolute basics that some of us forgot or never learned. Semantic HTML is one of the cornerstones of accessible web development. Even if we write everything in JavaScript, HTML is what is finally rendered in the user’s browser.

(Re)learning HTML and using it consciously prevents and fixes many accessibility issues.

8. JavaScript is not the enemy, and sometimes JavaScript even improves accessibility

I’m one of those people who believes that most websites should be accessible even when JavaScript fails to execute. That doesn’t mean that I hate JavaScript; of course not—it pays part of my rent. JavaScript is not the enemy, but it’s important that we use it carefully because it’s very easy to change the user experience for the worse otherwise.

Not that long ago, I didn’t know that JavaScript could improve accessibility. We can leverage its power to make our websites more accessible for keyboard users. We can do things like trapping focus in a modal window, adding key controls to custom components, or showing and hiding content in an accessible manner.

There are many impressive and creative CSS-only implementations of common widgets, but they’re often less accessible and provide worse UX than their JavaScript equivalents. In a post about building a fully accessible help tooltip, Sara Soueidan explains why JavaScript is important for accessibility. “Every single no-JS solution came with a very bad downside that negatively affected the user experience,” she writes.

9. It’s a good time to know vanilla CSS and JavaScript

For a long time, we’ve been reliant on libraries, frameworks, grid systems, and polyfills because we demanded more of browsers than they were able to give us. Naturally, we got used to many of those tools, but from time to time we should take a step back and question if we really still need them. There were many problems that Bootstrap and jQuery solved for us, but do those problems still exist, or is it just easier for us to write $() instead of document.querySelector()?

jQuery is still relevant, but browser inconsistencies aren’t as bad as they used to be. CSS Grid Layout is supported in all major desktop browsers, and thanks to progressive enhancement we can still provide experiences for legacy browsers. We can do feature detection natively with feature queries, testing has gotten much easier, and caniuse and MDN help us understand what browsers are capable of. Many people use frameworks and libraries without knowing what problems those tools are solving. To decide whether it makes sense to add the extra weight to your site, you need a solid understanding of HTML, CSS, and JavaScript. Instead of increasing the page weight for older browsers, it’s often better to progressively enhance an experience. Progressively enhancing our websites—and reducing the number of requests, kilobytes, and dependencies—makes them faster and more robust, and thus more accessible.

10. Keep learning about accessibility and share your knowledge

I’m really thankful that I’ve learned all this in the past few months. Previously, I was a very passive part of the web community for a very long time. Ever since I started to participate online, attend and organize events, and write about web-related topics, especially accessibility, things have changed significantly for me and I’ve grown both personally and professionally.

Understanding the importance of access and inclusion, viewing things from different perspectives, and challenging my decisions has helped me become a better developer.

Knowing how things should be done is great, but it’s just the first step. Truly caring, implementing, and most importantly sharing your knowledge is what makes an impact.

Share your knowledge

Don’t be afraid to share what you’ve learned. Write articles, talk at meetups, and give in-house workshops. The distinct culture of sharing knowledge is one of the most important and beautiful things about our industry.

Go to conferences and meetups

Attending conferences and meetups is very valuable because you get to meet many different people from whom you can learn. There are several dedicated accessibility events and many conferences that feature at least one accessibility talk.

Organize meetups

Dennis Deacon describes his decision to start and run an accessibility meetup as a life-changing experience. Meetups are very important and valuable for the community, but organizing a meetup doesn’t just bring value to attendees and speakers. As an organizer, you get to meet all these people and learn from them. By listening and by understanding how they see and approach things, and what’s important to them, you are able to broaden your horizons. You grow as a person, but you also get to meet other professionals, agencies, and companies from which you may also benefit professionally.

Invite experts to your meetup or conference

If you’re a meetup or conference organizer, you can have a massive impact on the role accessibility plays in our community. Invite accessibility experts to your event and give the topic a forum for discussion.

Follow accessibility experts on Twitter

Follow experts on Twitter to learn what they’re working on, what bothers them, and what they think about recent developments in inclusive web development and design in general. I’ve learned a lot from the following people: Aaron Gustafson, Adrian Roselli, Carie Fisher, Deborah Edwards-Onoro, Heydon Pickering, Hugo Giraudel, Jo Spelbrink, Karl Groves, Léonie Watson, Marco Zehe, Marcy Sutton, Rob Dodson, Scott O’Hara, Scott Vinkle, and Steve Faulkner.

11. Simply get started

You don’t have to go all-in from the very beginning. If you improve just one thing, you’re already doing a great job in bringing us closer to a better web. Just get started and keep working.

There are a lot of resources out there, and trying to find out how and where to start can get quite overwhelming. I’ve gathered a few sites and books that helped me; hopefully they will help you as well. The following lists are by no means exhaustive.

Video series

This free Udacity course is a great way to get started.
Rob Dodson covers many different accessibility topics in his video series A11ycasts (a11y is short for accessibility—the number eleven stands for the number of letters omitted).

Books

Blogs

Newsletters

Accessible JavaScript components

Resources and further reading

Introducing capsule networks - O'Reilly Media

Tue, 06 Feb 2018 12:00:00 +0800

Identification (source: Pixabay)

Capsule networks (CapsNets) are a hot new neural net architecture that may well have a profound impact on deep learning, in particular for computer vision. Wait, isn't computer vision pretty much solved already? Haven't we all seen fabulous examples of convolutional neural networks (CNNs) reaching super-human level in various computer vision tasks, such as classification, localization, object detection, semantic segmentation or instance segmentation (see Figure 1)?

Figure 1. Some of the main computer vision tasks. Today, each of these tasks requires a very different CNN architecture, for example ResNet for classification, YOLO for object detection, Mask R-CNN for instance segmentation, and so on. Image by Aurélien Géron.

Well, yes, we’ve seen fabulous CNNs, but:

They were trained on huge numbers of images (or they reused parts of neural nets that had). CapsNets can generalize well using much less training data.
CNNs don’t handle ambiguity very well. CapsNets do, so they can perform well even on crowded scenes (although, they still struggle with backgrounds right now).
CNNs lose plenty of information in the pooling layers. These layers reduce the spatial resolution (see Figure 2), so their outputs are invariant to small changes in the inputs. This is a problem when detailed information must be preserved throughout the network, such as in semantic segmentation. Today, this issue is addressed by building complex architectures around CNNs to recover some of the lost information. With CapsNets, detailed pose information (such as precise object position, rotation, thickness, skew, size, and so on) is preserved throughout the network, rather than lost and later recovered. Small changes to the inputs result in small changes to the outputs—information is preserved. This is called "equivariance." As a result, CapsNets can use the same simple and consistent architecture across different vision tasks.
Finally, CNNs require extra components to automatically identify which object a part belongs to (e.g., this leg belongs to this sheep). CapsNets give you the hierarchy of parts for free.

Figure 2. The DeepLab2 pipeline for image segmentation, by Liang-Chieh Chen, et al.: notice that the output of the CNN (top right) is very coarse, making it necessary to add extra steps to recover some of the lost details. From the paper DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs, figure reproduced with the kind permission of the authors. See this great post by S. Chilamkurthy to see how diverse and complex the architectures for semantic segmentation can get.

CapsNets were first introduced in 2011 by Geoffrey Hinton, et al., in a paper called Transforming Autoencoders, but it was only a few months ago, in November 2017, that Sara Sabour, Nicholas Frosst, and Geoffrey Hinton published a paper called Dynamic Routing between Capsules, where they introduced a CapsNet architecture that reached state-of-the-art performance on MNIST (the famous data set of handwritten digit images), and got considerably better results than CNNs on MultiMNIST (a variant with overlapping pairs of different digits). See Figure 3.

Figure 3. MultiMNIST images (white) and their reconstructions by a CapsNet (red+green). “R” = reconstructions; “L” = labels. For example, the predictions for the first example (top left) are correct, and so are the reconstructions. But in the fifth example, the prediction is wrong: (5,7) instead of (5,0). Therefore, the 5 is correctly reconstructed, but not the 0. From the paper: Dynamic routing between capsules, figure reproduced with the kind permission of the authors.

Despite all their qualities, CapsNets are still far from perfect. Firstly, for now they don't perform as well as CNNs on larger images such as CIFAR10 or ImageNet. Moreover, they are computationally intensive, and they cannot detect two objects of the same type when they are too close to each other (this is called the "crowding problem," and it has been shown that humans have it, too). But the key ideas are extremely promising, and it seems likely that they just need a few tweaks to reach their full potential. After all, modern CNNs were invented in 1998, yet they only beat the state of the art on ImageNet in 2012, after a few tweaks.

So, what are CapsNets exactly?

In short, a CapsNet is composed of capsules rather than neurons. A capsule is a small group of neurons that learns to detect a particular object (e.g., a rectangle) within a given region of the image, and it outputs a vector (e.g., an 8-dimensional vector) whose length represents the estimated probability that the object is present^[1], and whose orientation (e.g., in 8D space) encodes the object's pose parameters (e.g., precise position, rotation, etc.). If the object is changed slightly (e.g., shifted, rotated, resized, etc.) then the capsule will output a vector of the same length, but oriented slightly differently. Thus, capsules are equivariant.

Much like a regular neural network, a CapsNet is organized in multiple layers (see Figure 4). The capsules in the lowest layer are called primary capsules: each of them receives a small region of the image as input (called its receptive field), and it tries to detect the presence and pose of a particular pattern, for example a rectangle. Capsules in higher layers, called routing capsules, detect larger and more complex objects, such as boats.

Figure 4. A two-layer CapsNet. In this example, the primary capsule layer has two maps of 5x5 capsules, while the second capsule layer has two maps of 3x3 capsules. Each capsule outputs a vector. Each arrow represents the output of a different capsule. Blue arrows represent the output of a capsule that tries to detect triangles, black arrows represent the output of a capsule that tries to detect rectangles, and so on. Image by Aurélien Géron.

The primary capsule layer is implemented using a few regular convolutional layers. For example, in the paper, they use two convolutional layers that output 256 6x6 features maps containing scalars. They reshape this output to get 32 6x6 maps containing 8-dimensional vectors. Finally, they use a novel squashing function to ensure these vectors have a length between 0 and 1 (to represent a probability). And that's it: this gives the output of the primary capsules.

The capsules in the next layers also try to detect objects and their pose, but they work very differently, using an algorithm called routing by agreement. This is where most of the magic of CapsNets lies. Let's look at an example.

Suppose there are just two primary capsules: one rectangle capsule and one triangle capsule, and suppose they both detected what they were looking for. Both the rectangle and the triangle could be part of either a house or a boat (see Figure 5). Given the pose of the rectangle, which is slightly rotated to the right, the house and the boat would have to be slightly rotated to the right as well. Given the pose of the triangle, the house would have to be almost upside down, whereas the boat would be slightly rotated to the right. Note that both the shapes and the whole/part relationships are learned during training. Now notice that the rectangle and the triangle agree on the pose of the boat, while they strongly disagree on the pose of the house. So, it is very likely that the rectangle and triangle are part of the same boat, and there is no house.

Figure 5. Routing by agreement, step 1—predict the presence and pose of objects based on the presence and pose of object parts, then look for agreement between the predictions. Image by Aurélien Géron.

Since we are now confident that the rectangle and triangle are part of the boat, it makes sense to send the outputs of the rectangle and triangle capsules more to the boat capsule, and less to the house capsule: this way, the boat capsule will receive more useful input signal, and the house capsule will receive less noise. For each connection, the routing-by-agreement algorithm maintains a routing weight (see Figure 6): it increases routing weight when there is agreement, and decreases it in case of disagreement.

Figure 6. Routing by agreement, step 2—update the routing weights. Image by Aurélien Géron.

The routing-by-agreement algorithm involves a few iterations of agreement-detection + routing-update (note that this happens for each prediction, not just once, and not just at training time). This is especially useful in crowded scenes: for example, in Figure 7, the scene is ambiguous because you could see an upside-down house in the middle, but this would leave the bottom rectangle and top triangle unexplained, so the routing-by-agreement algorithm will most likely converge to a better explanation: a boat at the bottom, and a house at the top. The ambiguity is said to be "explained away": the lower rectangle is best explained by the presence of a boat, which also explains the lower triangle, and once these two parts are explained away, the remaining parts are easily explained as a house.

Figure 7. Routing by agreement can parse crowded scenes, such as this ambiguous image, which could be misinterpreted as an upside-down house plus some unexplained parts. Instead, the lower rectangle will be routed to the boat, and this will also pull the lower triangle into the boat as well. Once that boat is “explained away,” it’s easy to interpret the top part as a house. Image by Aurélien Géron.

And that’s it—you know the key ideas behind CapsNets! If you want more details, check out my two videos on CapsNets (one on the architecture and another on the implementation) and my commented TensorFlow implementation (Jupyter Notebook). Please don’t hesitate to comment on the videos, file issues on GitHub if you see any, or contact me on Twitter @aureliengeron. I hope you found this post useful!

Article image: Identification (source: Pixabay).

The Supersonic Parachutes Carrying NASA's Martian Dreams

Tue, 06 Feb 2018 12:00:00 +0800

A new generation of space scientists is using high tech materials to resurrect the long-neglected supersonic parachute.

Jim Toomey has a weakness for space memorabilia, but he doesn’t have unlimited funds. So he had to keep his desires in check as he perused an online auction of NASA relics in May 2012. The Bradenton, Florida, entrepreneur mustered the self-discipline to skip the $53,758 baggie from Apollo 15 that had once held lunar dust, as was well the Gemini 5 log book that went for $33,844. He instead scooped up what he terms “a whole bunch of space junk.” His haul included a dented bracket from a Space Shuttle’s tail fin ($240), a spacesuit’s heater cable ($240), and four reels of 16-mm film that were advertised as having something to do with Viking, the 1970s NASA program that landed America’s first two spacecraft on Mars ($360).

Toomey promptly donated all his NASA curios to the South Florida Museum, where he served as a trustee. Though the museum appreciated the gifts, it deemed them too offbeat for public display. Instead they wound up in a box on a backroom storage shelf, amid a sea of other forgotten artifacts.

Toomey didn’t give the space junk another thought until September 2015, when he received an odd phone call from the Boston auction house that had sold him the goods. The caller said the company had been contacted by an engineer from NASA’s Jet Propulsion Laboratory named Rob Manning who was desperate to locate the Viking films. According to Manning, NASA had discarded the reels by accident; they’d been left in a file cabinet marked for sale as government surplus during an office move in 1999.

Toomey agreed to speak to Manning and assured him that the objects had been preserved rather than melted down for their silver. During that conversation, a grateful Manning revealed why he and his JPL colleagues were so eager to view the films: They contained the only surviving footage of the August 1972 qualification test for Viking’s parachute, the contraption responsible for safely decelerating the program’s landers through the Martian atmosphere. Because that atmosphere is 99 percent thinner than Earth’s, Viking’s engineers knew their spacecraft would be plummeting at supersonic speeds as they neared the planet’s surface. The engineers had thus built a novel parachute that could endure such punishing conditions—a 2,200-square-foot expanse of white polyester with braided nylon suspension lines.

A parachute successfully landed a capsule at White Sands in 1972.

NASA

The August 1972 test had confirmed their success. A balloon first carried a 1,900-pound capsule to the edge of the stratosphere at New Mexico’s White Sands Missile Range. When the vehicle dropped free from the balloon, it fired rockets that cranked up its speed to nearly 1,600 miles per hour. At that point, a small explosive blasted open the compartment that held the parachute. An on-board camera captured the violence of that moment, as the polyester jerked and shuddered before mellowing into a stable hemisphere. The films that Toomey had bought at auction, and which NASA wished to reclaim, were the sole visual evidence of that engineering triumph.

NASA’s supersonic parachutes have barely evolved in the 46 years since that milestone above the New Mexico desert. The Viking model worked so well that the space agency kept using it for all of its Mars missions, thereby eliminating the need to spend hundreds of million of dollars on new tests. When the Curiosity rover alighted inside the Gale Crater in 2012, for example, it did so with the aid of a parachute that was essentially a larger version of the one that had decelerated the Viking 1 lander 36 years earlier.

But the trusty Viking parachute is finally on the verge of obsolescence. The aging design can’t provide the drag necessary to slow down payloads much heavier than a ton—a significant problem for NASA given that its longterm Martian dreams involve vehicles that will weigh more than 20 Curiosities combined. And so the engineers at JPL have been scrambling to come up with a much tougher replacement, a parachute that can help solve the Martian version of the last-mile problem for decades to come.

Rob Manning went searching for the lost Viking films as part of that crucial effort. He found them right as NASA’s supersonic parachute team was reeling from its second major setback in as many years, failures that underscored the maddening complexities of predicting how cloth and rope will behave at extraordinary speeds.

A NASA test vehicle was launched by balloon in 2015 from a tower at the US Navy's Pacific Missile Range Facility in Kauai, Hawaii.

Bill Ingalls/NASA

Parachutes may lack the obvious sex appeal of jets and rockets, but their aeronautical value wasn’t lost on the organizers of Operation Paperclip, the American intelligence venture that brought hundreds of German scientists to the US in the wake of World War II. Among the engineers they chose to spirit out of Germany were Theodor Knacke and Helmut Heinrich, the stars of the parachute section at the Graf Zeppelin Research Institute in Stuttgart. The two men were taken to Wright-Patterson Air Force Base in Ohio and put to work on military projects such as developing the braking parachute for the B-47 bomber. (Before the advent of thrust reversers, the high-speed bombers needed help slowing down when landing.) During the duo’s years at Wright-Patterson, Knacke invented the ringslot, a low-cost parachute that features circular bands of fabric separated by horizontal gaps. Heinrich, meanwhile, won a series of patents for methods of enhancing the stability of parachutes as they unfurl.

The advent of the American space program in the 1950s gave Knacke and Heinrich the chance to forge prestigious careers beyond Ohio. Though NASA briefly flirted with the idea of using a glider wing to make its manned vehicles steerable as they returned to Earth, parachutes won out as the most dependable and efficient means of slowing down spacecraft prior to splashdown. But there were few American engineers who specialized in parachutes, a technology that was widely thought to lack machismo. “You’d have these mechanical engineering guys look around at all the textiles and the sewing that go into it and they’d say, ‘Well, that’s women’s work,’” says Chuck Lowry, a Kentucky-bred legend of parachute design who cut his teeth on ejection seats and later became a key figure in the Apollo program. “When I was at North American Aviation, we had 100,000 employees and mine was the only badge that said PARACHUTES on it.”

The two German émigrés took advantage of this talent vacuum to become the fathers of American space parachutes. The entrepreneurial Knacke struck out for Southern California, where he founded Space Recovery Systems in 1957; the company manufactured parachutes for experimental rockets such as Blue Scout I, which was used to gauge the stresses of atmospheric re-entry. Five years later he joined Northrop as chief of the division that held the parachute contracts for the Gemini and Apollo programs. It was Knacke who persuaded NASA to select the famous three-parachute cluster for its Apollo capsules; he believed that a single large parachute would take a dangerously long time to open. (Toward the end of his career, which lasted well into the 1990s, Knacke wrote a book, Parachute Recovery Systems Design Manual, that’s acknowledged as the bible of his discipline.)

Heinrich, by contrast, resisted the siren call of industry to enter academia. In 1956 he became a professor of aeronautical engineering at the University of Minnesota, a position he would hold until his death 23 years later. NASA soon began pumping millions of dollars into Heinrich’s lab, much of it earmarked for one blue-sky project in particular: the development of a parachute robust enough to withstand supersonic flight. Though NASA still had yet to loft a human into space, it already had notions that a supersonic parachute might come in handy someday—if not on Earth, then perhaps on Mars or Venus.

The challenge was a daunting one: Parachutes are erratic beasts in even the calmest of conditions. “Parachutes have so much randomness to them,” Lowry says. “It’s not like the wing of an airplane, where it stays the same as it goes up and it goes down. A parachute is stowed in this bag and then it has to assemble itself into this shape. But there are a million other shapes it takes as it does that, and a million variables that affect that process of deployment and inflation. You just never know what’s going to happen as it captures the air.” And when that air is stratosphere-thin and rushing by at supersonic speeds, the list of things that can go wrong gets staggeringly long; there are countless potential pressure spikes that can distort the canopy as it struggles to achieve its final form.

Yet Heinrich was the ideal engineer to work on this problem because he’d solved it once before—albeit on a very small scale. Back in Stuttgart during the war, he’d made a mini parachute that the Luftwaffe used to control the descent of its mines and torpedoes. This so-called guide surface parachute, a dramatically cupped arrangement of eight triangular panels that measured four feet across, was notable for how little it oscillated when buffeted by strong winds. It was so stable, in fact, that it was able to survive wind-tunnel tests that mimicked the effects of traveling at Mach 3—more than 2,300 mph.

Heinrich’s research at the University of Minnesota focused on how the structural elements of parachutes—the vents and the stitching, for example—can be configured to absorb and dissipate the effects of supersonic shock waves. Though his labor would prove invaluable to NASA, he wouldn’t be the engineer to come up with the winning parachute design for Mars. That honor would instead go to one of his star students, a man named Clinton Eckstrom.

Soon after leaving his mentor’s lab, Eckstrom landed a job with G.T. Schjeldahl, a Minneapolis-based manufacturer of Mylar weather balloons. The company asked him to create a parachute for rockets that can reach the mesosphere; Schjeldahl wanted to get into the business of gathering meteorological data at 200,000 feet, where the atmospheric density happens to be similar to that on Mars. In 1964 Eckstrom completed his assignment by patenting what came to be known as the disk-gap-band parachute, or DGB. The design is primarily comprised of just two large swaths of cloth, separated by a sizable gap: a top shaped like an inverted saucer and a lower circular strip.

Impressed by the DGB’s reliability in high-altitude environments, NASA hired Eckstrom to work on its first series of supersonic-parachute tests, the Planetary Entry Parachute Program. Awash in limitless cash thanks to its pivotal role in the Cold War, NASA spent vast sums on the program and two follow-up initiatives. Eckstrom and his fellow engineers enjoyed the freedom to launch as many test vehicles as they wished, so they didn’t sweat their numerous semi-comical failures: In at least two atmospheric tests, for example, the parachutes’ bags, which were supposed to be fired clear during deployment, tore back through the open canopies like bullets.

By the early 1970s, NASA had narrowed its field of potential supersonic parachutes to two finalists. One was a fortified version of Eckstrom’s DGB, which had gained a reputation for orderly inflation; the other was a ringsail, a more elaborate successor to Knacke’s ringslot, which produced superior drag. NASA ultimately decided to play it safe and opt for the less complicated design: It was a DGB that passed muster at the White Sands Missile Range in August 1972, and it’s been part of all seven NASA missions to the Martian surface.

Despite the DGB’s spotless service record, there remain skeptics who believe the ringsail should have prevailed in the Viking competition. Their arguments resonate with the young engineer who’s been entrusted to create NASA’s next supersonic parachute, an endeavor he describes as “playing chess with the laws of physics.”

Within minutes of welcoming me into his office at JPL in Pasadena, California, Ian Clark has a swatch of ripstop nylon dangling from the corner of his mouth. “I’m going to have to go Hulkmania on this,” he says as he strains to tear the hardy fabric with his teeth. Once he finally manages to shred it, he hands me a fragment so I can examine the fibers that dangle from its edge—the matrix of a modern space parachute, which use ultra-light nylon in lieu of heavy Dacron polyester.

Clark knew next to nothing about parachutes when he started at JPL in 2009. As a doctoral student at Georgia Tech, he’d become an authority on so-called inflatable aerodynamic decelerators—giant inner-tube-like devices that are meant to be placed on the undersides of supersonic spacecraft, and which should theoretically act as brakes when they’re inflated in the atmosphere. (When fully filled with air, they resemble Victorian hoop skirts.) His original marching orders at NASA were to build an inflatable decelerator so effective that it could slow a Mars-bound vehicle to less than the speed of sound in a matter of minutes. Once that goal was reached, a supersonic parachute wouldn’t be necessary; a run-of-the-mill subsonic parachute would suffice to finish the braking job and prepare the spacecraft for landing.

Ian Clark leads the supersonic parachute team at the Jet Propulsion Laboratory in Pasadena.

Kenny Hurtado

But Clark wasn’t at JPL long until he realized that his plan to use inflatable decelerators instead of supersonic parachutes had been wildly optimistic. “You ran the math on trying to bump the parachute down to subsonic, and the inefficiencies associated with that started to manifest very quickly,” says Clark, a wiry 37-year-old who radiates a cheerful yet manic intensity. “You would need just these enormous inflatable drag devices, so enormous there’s no way they could ever be efficient.” By his second year at NASA, then, he became committed to a new strategy for slowing down future multi-ton—and perhaps manned—spacecraft as they approach Mars: coupling his inflatable decelerator with a revamped supersonic parachute.

Clark’s two-part concept became the foundation for the $200 million Low-Density Supersonic Decelerator project, which he was tapped to lead as principal investigator. In addition to building a 6,800-pound UFO-like vehicle equipped with one of Clark’s inflatable skirts, the supersonic decelerator team would also prototype a replacement for the Viking parachute. The plan was to fly the system three times at the Pacific Missile Range Facility in Kekaha, Hawaii. Those tests would be the first since 1972 to attempt to deploy full-scale supersonic parachutes in the Earth’s atmosphere.

Since supersonic-parachute research had been stagnant for 40 years, the team had to get creative when recruiting talent. One of the project’s unconventional hires was a Clara O’Farrell, an Argentinian engineer who spent her graduate career at Caltech studying jellyfish locomotion—specifically how the creatures propel themselves by squeezing out vortex rings. Though she’d long yearned to work on spaceflight, she feared there would be little demand for her arcane expertise. “When I graduated and started looking for a job in the industry, my background in fluid dynamics didn’t seem too useful since there aren’t many areas of space with fluids,” she says. “The whole space-being-a-vacuum thing sort of got in my way.” But O’Farrell found a home at JPL analyzing the physics of the supersonic decelerator’s parachute, a structure with many of the same amorphous qualities as a jellyfish’s body.

Each supersonic parachute must be carefully packed for testing.

Kenny Hurtado

Clark knew next to nothing about parachutes when he started at JPL in 2009.

Kenny Hurtado

Clark and his team tested 55 potential parachutes in a wind tunnel. They were able to weed out some of the candidates by attacking them with scisssors and then testing them anew; if a parachute couldn’t continue to perform after receiving a few slashes from a handheld blade, it stood no chance in the unforgiving Martian atmosphere. As worthy contenders emerged from the pack, Clark also dove into the literature on supersonic parachutes from NASA’s Nixon-era heyday. His archival research led him to admire the ringsail, the runner-up in the Viking contest. He liked that the design used thousands of stitched-together panels in its canopy, as opposed to the DGB’s reliance on a solid disk of cloth. Clark hypothesized that the ringsail’s more complex arrangement would make it less susceptible to failure, since tears can’t easily propagate from one isolated panel to the next. “The analogy I use is that the DGB is like a Toyota Camry—it’s very simple, but seemingly very reliable,” says Clark. “Whereas the ringsail, it’s much more finely tuned.”

Despite Clark’s enthusiasm for the ringsail, he wasn’t quite ready to ditch the disk-gap-band parachute entirely for the first supersonic decelerator test in June 2014. He instead settled on a hybrid: The lower portions of the parachute were constructed from small individual panels, but the top was a shallow broadcloth disk. Clark dubbed his invention the disk sail.

Though appealing on paper as a best-of-both-worlds solution, the disk sail fared poorly in Hawaii. The trial started off well, at least: The test vehicle reached an altitude of 120,000 feet while tethered to a balloon, then cut free and used its rockets to climb another 70,000 feet. At a speed of more than 2,800 miles per hour, the inflatable decelerator—the hoop-skirt contraption—expanded to its full girth in less than a second; the drag it produced cut the vehicle’s speed by around 30 percent. But when the mortar fired to open the parachute compartment, everything went awry: The Technora suspension lines didn’t even get a chance to straighten before the parachute was cut to smithereens by the environmental forces.

The test data convinced Clark that the disk sail’s top was doomed to become as flat and tight as a snare drum’s head when pelted with fantastic amounts of pressure. And so for the second test, scheduled for June 8, 2015, he elected to go full ringsail: He replaced the disk with a slew of bulbous panels that gave the parachute a more intricate, rounded shape.

To make sure they were on the right track, the team attached the ringsail to a rocket-powered sled—essentially just a quartet of high-powered engines affixed to a rail track—at Naval Air Weapons Station China Lake in California’s Mojave Desert. The parachute showed no signs of being fazed when the moving sled subjected it to 120,000 pounds of force—40,000 more pounds than the balloon test was anticipated to create. Clark was confident all would go smoothly on the next flight, and that his project would pull off a feat that hadn’t occurred on Earth in 43 years.

The second test once again began with a flawless performance by the airbag. The parachute’s white-and-orange fabric came gushing out as the vehicle hit Mach 2.4. For a few hundredths of a second, the ringsail opened just as everyone had hoped, with billowing edges that betrayed no hint of being warped by abnormal pressure. But when the parachute was 98 percent of the way to full inflation, a triangular slice of the canopy peeled away. The parachute hung there for a moment, taunting the engineers in the control room with its missing portion, before disintegrating into a jumble of flaccid ribbons.

As he watched the culmination of a half-decade’s worth of work turn into something that resembled used Kleenex, Clark vacillated between numbness and self-flagellation. The computer models, the wind-tunnel tests, the rocket sled—all of his team’s preparations had pointed toward a positive result. Yet there it was on the control-room screen, impotently flapping in the supersonic breeze. Clark knew then that he’d placed too much faith in math.

Four months after the second supersonic parachute disappointment, Rob Manning lifted Clark’s spirits with some good news: By sleuthing around the internet, he’d finally located the long-lost films of the last Viking parachute test. The South Florida Museum had agreed to give the priceless footage back to NASA, but Clark would have to pick them up himself and, as small recompense, deliver a public lecture at the museum’s planetarium.

As soon as he touched down back in California, Clark took the reels to a Hollywood company that restores damaged movies. The films were in surprisingly decent shape and the restorer was able to play them on a projector with minimal effort. Elated, Clark took an iPhone video of a segment that shows the parachute’s inflation in slow motion. That clip was what he’d been seeking ever since the last misfortune in Hawaii: visual evidence of what a supersonic parachute should look like when things go right.

Clark craved such evidence because he’d concluded that, despite his exhaustive research, the physics of supersonic parachutes are still too poorly understood to be of much use to an engineer. He had assumed he could predict how those parachutes would behave based on existing data from the 1960s and early 1970s. But the bewildering supersonic decelerator experience had taught him that there just haven’t been enough real-world tests to build an effective model. He thus realized that, at least until there are many more tests on the books, he must rely less on raw numbers and more on intuition—something that left-brained engineering types are typically loath to do. Picking apart the Viking films frame-by-frame was one of the few ways he could hone that sixth sense.

Clark also sought advice from several parachute graybeards, engineers who’d trained under Theodor Knacke and Helmut Heinrich and who still possessed an instinctive feel for how fabric reacts to extreme stress. Chuck Lowry, for example, counseled Clark to stick with the ringsail and use a webbing to reinforce the areas where the tear had occurred in the 2015 test. But Clark decided that he needed to maximize the odds that his next test would be a success; his top priority had to be accumulating fresh footage to analyze, as a small step toward making parachute design a more objective enterprise. And that meant accepting that his team should turn conservative and make their next parachute a DGB.

Clark would have to wait longer than expected to make and test that parachute. Citing a need to reallocate money to a program that repairs satellites, NASA slashed the supersonic decelerator’s funding by 85 percent in early 2016—a decision that led to the shuttering of the project before its planned third test. But Clark received a lifeline from Mars 2020, an upcoming mission to deliver a Curiosity-style rover to the Red Planet’s surface. The mission’s leaders were intent on sticking with the basic Viking design for their parachute, but they offered Clark the chance to fabricate his own version and, most important, test it in the atmosphere—an endeavor that would come to be known as the Advanced Supersonic Parachute Inflation Research Experiment, or Aspire.

Before Clark and a few handpicked collaborators got to fiddling with the Viking DGB, however, Mars 2020 wanted them to attempt something much more basic: Conduct an atmospheric test of the exact same parachute that Curiosity had used in 2012. If they could pull off that task, and thus end their two-test losing streak, they would then be given the resources to repeat the experiment with their own take on the design.

A Black Brant IX sounding rocket takes off from the Wallops Flight Facility on the Virginia coastline in October 2017.

Jamie Adkins/NASA

The first Aspire test—and perhaps Clark’s last shot at proving he could deploy a supersonic parachute—took place on October 4, 2017, at the Wallops Flight Facility, on the easternmost stretch of Virginia’s coastline. When a Black Brant IX sounding rocket reached an altitude of around 26 miles while screaming along at Mach 1.8, a built-to-print version of the Curiosity DGB came bursting out of its bag. The canopy snapped into shape and twisted counterclockwise in the ferocious wind. None of the engineers dared jinx the promising moment by celebrating prematurely, or even making the slightest sound.

“There was a lot of pressure on us to get this right,” Clara O’Farrell says. “The last time, it had been heartbreaking—we’d put all this effort into making changes, making the best parachute possible, and everyone’s hopes had gotten up before it failed.” This time, 20 seconds of silence passed before anyone in the control room felt comfortable enough to cheer.

It was by no means the glorious achievement the engineers had envisioned for themselves during the headiest days of the supersonic decelerator project; this DGB was just a retread from a past mission, not the parachute that would help land 10-ton vehicles safely on Mars. But they took pride in having accomplished something that hadn’t been done since before most of them were born. After so much frustration, they were finally a few inches past the starting line.

The next Aspire test, scheduled for March 20 at Wallops, will be far more ambitious than its predecessor. Clark and his colleagues have come up with a DGB that conforms to the basic Viking design but also features numerous clever tweaks that will hopefully increase drag without also increasing the odds of a catastrophe. The engineers worked with the company that sewed the parachute, Airborne Systems, to incorporate a new nylon that is three times stronger but only 50 percent heavier than the one Clark tore apart with his teeth for my benefit. And they also fortified the Technora suspension lines by changing the way their fibers are braided.

The footage that the forthcoming Aspire tests yield, shot by digital cameras capturing 1,000 frames per second at 4k resolution, could provide Clark with more insight into supersonic-parachute dynamics than the sum of all the vintage NASA documents he’s ever read. “At that level of detail I can see the entirety of the suspension lines, I can see inflections in the broadcloth, I can see shadows in the areas where the parachute is under stress and beginning to stretch out a little bit,” he says. (The 16-mm Viking films were shot at just 30 frames per second.)

Clark accepts that his efforts to comprehend the physics of supersonic parachutes could ultimately be for naught. Years of fleshing out his knowledge with increasingly expensive tests may lead him to conclude that decelerating, say, a 20-ton spacecraft full of astronauts on Mars requires a braking technology that doesn’t involve fabric—perhaps some combination of his inflatable decelerator and a fuel-powered engine. There are already some engineers who seem to be betting that the next age of Martian exploration will be parachute-free: SpaceX, for example, has discussed vague plans to use thrusters to land a large vehicle on Mars.

But supersonic parachutes will be given every chance to prove their viability for Mars because their efficiency cannot be beat. In the simplest terms, a hundred pounds of fabric will always provide exponentially more braking power than a hundred pounds of rocket fuel. And when designing a vehicle to make the 33.9-million-mile journey to Mars, saving half a ton of weight could spell the difference between a mission that ends in triumph and one that ends with a heap of twisted wreckage atop Olympus Mons.

Brendan I. Koerner (@brendankoerner) is a WIRED contributing editor. He wrote about theft from a Silicon plant in September.

Business Lessons from Reed Hastings (Netflix)- Part 1 – 25iq

Sat, 03 Feb 2018 00:00:00 +0800

Reed Hastings is the Chairman, CEO, President and co-Founder of Netflix. After graduating from Bowdoin College in Maine Hastings joined the Peace Corps. He then obtained a master’s degree in computer science at Stanford. Hastings was the founder of Pure Software which built tools for Unix software developers. Pure was acquired by Rational Software in 1997, the same year Netflix was founded.

“We’re figuring out every year how to use the Internet to make a great consumer experience. Every year is an experiment.” “We are investing heavily in [machine learning] because we want it to be: you turn on Netflix and there’s a row, there’s like four choices, and you just want to watch them all. To get to that consistent view is where we are targeting.” “We’re very experimental so if it works well we’ll do more. If it doesn’t work well, you won’t hear about it anymore.” “Netflix ‘gets me’ is the emotion we are looking for.” “If the Starbucks secret is a smile when you get your latte, ours is that the Web site adapts to the individual’s taste.” “We are just a learning machine. Every time we put out a new show we are analyzing it, figuring out what worked and what didn’t so we get better next time.” “It’s not Netflix that’s making the changes. It’s the Internet.

Netflix is a leading example of a business which understands the power of a connected and data-driven relationship with a customer. This relationship creates a powerful platform that enables a business to constantly experiment to discover, among other things, how to: (1) deliver more value to customers; (2) expand the market for existing products and services and create new products and services and (3) help customers find what will make them happiest. The goal with this process is to drive the best indicator of customer happiness which is actual usage of the product or service. As just one data point on how powerful this approach can be: more than 80% of the video people watch on Netflix is discovered through a recommendation engine. Wired describes the process:

Netflix uses machine learning and algorithms to help break viewers’ preconceived notions and find shows that they might not have initially chosen. To help understand, consider a three-legged stool. “The three legs of this stool would be: Netflix members; taggers who understand everything about the content; and our machine learning algorithms that take all of the data and put things together,” says Todd Yellin, Netflix’s vice president of product innovation.

People tend to put a lot of emphasis on Netflix using machine learning to decide what video content to create and not enough emphasis on how it is used to more efficiently acquire customers and turn them into fans of its services. The use of machine learning to improve customer acquisition and marketing has resulted in significant improvements in customer unit economics in many industries but Netflix is one of the best examples. For example, huge investments are now being made by businesses in systems which optimize every type and aspect of acquisition funnels. If one business does not use machine learning in this way and its competitors do so, that business is operating with a huge possibly fatal disadvantage. To illustrate how machine learning can be used to improve marketing, after making the initial decision to invest in House of Cards the most important use of machine learning by Netflix was to optimize marketing and distribution rather than production. By using machine learning approaches is it possible through experimentation to find the optimal approach to creating “fans” for Netflix’s services. Modern technology enables mass customization so that the nature of the trailer each person sees and when the video is released for viewing to be optimized to create the desired word of mouth.

Once a connected relationship exists with customers the ability of the service provider to use telemetry data respond to customer desires and behavior is greatly enhanced. Systems can be created to run experiments at lower cost and greater speed than has ever been possible before due to the proliferation of a vast array of cloud-based services. Once the experimentation process is automated, people like marketers, product managers and engineers have the ability to run experiments quickly without the direct involvement of a data scientist. The speed at which optimizing product and service experiments can be run provides a business with either a significant competitive advantage or the ability of that business to stay in business if competitors are doing so. As Charlie Munger says:

“The great lesson in microeconomics is to discriminate between when technology [like machine learning] is going to help you and when it’s going to kill you. And most people do not get this straight in their heads…. that there are all kinds of wonderful new inventions that give you nothing as owners except the opportunity to spend a lot more money in a business that’s still going to be lousy. The money still won’t come to you. All of the advantages from great improvements [in a technology like machine learning] are going to flow through to the customers.”

Can a business opt out of making machine learning investments? Well, it can’t do so for very long and stay in business.

The biggest cultural change that businesses are going through as a result of this technological shift is a transition from being organizations that measured themselves by shipping products and services to organizations driven by actual customer usage of services. I heard someone say recently that “everyone needs to become a data scientist” and I like that phrase a lot. Of course, systems and tools must be created to make this “we are all data scientists now” more of a reality. Creating these systems is essential since the alternative of employees of a business getting in line to wait for their turn to talk to a “real data scientist” will not scale.

Hastings does not talk about how research is used at Netflix to supplement data science, but I suspect that the emphasis on data about actual usage does not mean that they don’t conduct any research. For example, with a new product or service there is not yet any use to measure, so research can be valuable. Some blending of quantitative and qualitative data is probably happening at Netflix. This text from an article in The Guardian describes one aspect of decision making at Netflix that makes clear that human intuition is still a big part of how they make decisions:

Ted Sarandos once called himself a “human algorithm” because of his use of data rather than gut instinct to work out what viewers want, and he has described House of Cards as “generated by algorithm”: … he used data to work out how many subscribers loved both political thrillers and Kevin Spacey. “Big data is a very important resource to allow us to see how much to invest in a project but we don’t try to reverse-engineer,” he says, adding that commissioning decisions are more like “70% science and 30% art”.

What sort of data do they look at?

“…the focus is on the audience, and that there is no programming grid – or appointment linear-based television – that is typically used by traditional TV networks. Barometers of success are if the audience completes watching the show, the timeframe within which they finish watching a series if there is social media buzz by critics and fans.”

2. “What we do is try to learn and adapt. Rather than commit to one particular point of view, we will adapt.”

The truism is: the only constant is change. What is different about this truism now is that the pace of change is accelerating. When I hear people talk about a “Great Stagnation” I think: “What planet are these people living on?” and “What industry are they working in?” Developments in machine learning mean that systems can be created which allow adaptation to occur on a scale and at a speed never before possible. Prediction becomes less possible than ever (if it ever was possible) in this new environment due to rising complexity caused by connects systems and the ability to adapt quickly is vastly more important. Michael Mauboussin describes the challenge in this way: “Increasingly, professionals are forced to confront decisions related to complex systems, which are by their very nature nonlinear… Complex adaptive systems effectively obscure cause and effect. You can’t make predictions in any but the broadest and vaguest terms. … complexity doesn’t lend itself to tidy mathematics in the way that some traditional, linear financial models do.” A business like Netflix that has decided to focus on adaption and making decisions opportunistically at every step is trying to maximize optionality which allows them to handle complexity more effectively than their competitors. Nassim Taleb identifies the problem with a non-adaptive approach: “A rigid business plan gets one locked into a preset invariant policy, like a highway without exits —hence devoid of optionality.” The accelerating pace of change is creating a need for businesses to adapt faster and more efficiently. As Darwin famously said: ‘It is not the most intellectual of the species that survives; it is not the strongest that survives; the species that survives is the one that is able to adapt to and to adjust best to the changing environment in which it finds itself.”

3. “By 2011 we realized that many of the firms we were buying from were eventually going to want to run their own streaming service. We had no reliable supply. We had to go vertical since it was not going to be in their interest to sell to us over time.”

Hastings is saying that Netflix understands the dangers associated with “wholesale transfer pricing.” Eugene Wei has written specifically about how the concept applies to Netflix:

“Netflix had a great advantage when First Sale Doctrine permitted them to buy DVDs at the same wholesale price as any retailer since it capped their costs. But in the TV/movie licensing world, the content owner can constantly adjust their price to squeeze almost every last drop of margin from the distributor as you can’t find perfect substitutes for the goods being offered. Ask TV networks if they make any money licensing NFL, NBA, and MLB games for broadcast. Hint: the answer is no. In the digital world, transfer pricing can be even more of a cruel mistress.”

To illustrate what Netflix is trying to avoid with an analogy, imagine yourself walking into a store and wanting to buy an antique chest. Also assume that the owner by listening to what you said to your friend heard you say: “I simply must own it.” The merchant’s power to set the price combined with your desire to buy that chest means that you are in deep trouble in setting the price of the chest. Walking into that antique store and not needing to buy anything completely changes the negotiating dynamic on price. As a reminder:

Don’t fight a land war in Asia;
Don’t play cards with a man called Doc;
Don’t write really long blog posts and expect many people to read to the end; and
Don’t negotiate the price of something the seller knows you must have.

Netflix knows it must own a significant portion of its own content. It was forced to make a “burn the lifeboats” decision to move forward with owning content once they started streaming. The conventional wisdom was that tech companies which start dealing with Hollywood are inevitably eaten alive. And yet Netflix has made it work. But there were doubters at first. I remember when:

Hastings has made clear his view that a business can’t dabble in something like creating content or technology. He has said that at least one third of a company’s total resources must be committed in order for a business to have any financial success. He has also said that licensing content and trying to create a business around that is hard. Anyone and their dog can license and resell content to stream. Hastings believes: “It is easy to get into the streaming business– what is hard is making a profit.” Hastings tried licensing a few movies at Sundance and showing them on Netflix, but notes: “You can do the ‘money ball’ analysis about the content to figure how much it is going to get viewed. It is easy to see that it does not work.” Netflix had no choice but to start green lighting its own first party content “all in.” Business which dabble in something are setting themselves up for failure. Hasting’s view is that a business must jump in with both feet to be successful.

“There are scale economies in companies like Netflix. Those merit investing forward to get to scale. Call that normal business. Then there are rare businesses like LinkedIn and Facebook where there are network effects. The power of getting to scale is so great, and they are such extreme winner-take-all markets, that it’s worth doing anything to scale. You do crazy things because you have to grow 300% to maximize the opportunity. You do things that are sloppy. If you’re in a network effect business, you get more of ‘first is forever.’” “Netflix is trying to combine personalization, on demand content, smart TVs and all these great geeky things with incredible content. If we can combine them better that protects us against HBO that is very good a content but not so good at tech.” “You know often the right strategy for a challenger brand which is Amazon in the case of streaming video is to try many things, because they’re not sure, just copying Netflix is not going to typically get someone very far. The leader’s role is to keep the main thing the main thing. Our focus is on doing even better content, getting better partnerships, better mobile streaming. We just have to have the discipline to keep doing what we’re doing at many times the scale, and if we do that, things will work out really well for our global customer base and thus for our investors.”

Netflix benefits from “data network effects” since the more people who use Netflix, the more data they have and the better the recommendation systems get. Matt Turck writes:

“Data network effects occur when your product, generally powered by machine learning, becomes smarter as it gets more data from your users. In other words: the more users use your product, the more data they contribute; the more data they contribute, the smarter your product becomes (which can mean anything from core performance improvements to predictions, recommendations, personalization, etc. ); the smarter your product is, the better it serves your users and the more likely they are to come back often and contribute more data – and so on and so forth. Over time, your business becomes deeply and increasingly entrenched, as nobody can serve users as well. Data network effects require at least some level of automated productization of the learning. Of course, most well-run businesses “learn” in some way from data, but that’s typically done through analytics, with human analysts doing a lot of the work, and a separate process to build insights into the product or service. The more automation you build into the loop, the more likely you are to get a flywheel effect going. Google is a classic example of data network effect at play: the more people search, the more data they provide, enabling Google to constantly refine and improve its core performance, as well as personalize the user experience. Waze, now a Google company, is another great example, essentially a contributory database built on data network effects. There are also plenty of examples of data network effects found at the feature (rather than core business) level: for example, recommendation engines that are now everywhere from Amazon (products you’ll want to buy) to Netflix (movies you’ll want to watch) to LinkedIn (people you’ll want to connect with), and keep getting better with more users/data.

In addition to the demand-side economies of scale (network effects) just discussed, there are also supply-side economies of scale that benefit Netflix’s business. In other words, Netflix benefits from the size and scale of its operations, with cost per unit of output that decreases with increasing scale as fixed costs are spread out over more units of output. Many people have doubted this over that Netflix benefits from network effects and supply side scale economies over the years. Dredging a few headlines up is useful and entertaining:

“Why Comcast will Crush Netflix” from 2012 in Fast Company

“Lessons from the Strategy Crisis at Netflix” from 2016 in Strategy +Business

“Netflix: Sell, Sell, Sell” from 2017 in Seeking Alpha (of course the link no longer works)

As I have said many times, I will not do a valuation for readers of this blog of a stock and that includes Netflix. It would not be helpful for me to give stock “tips” in this blog since it encourages terrible behavior, particularly among people whom it can harm the most.

Netflix’s competitive position has produced a stock valuation that is quite attractive to say the least. The last time I checked NFLX was trading at ~265.

Even though I will not provide a stock tip people will be able to find them anyway from the usual mob of Wall Street sell-side analysts. Their views are depicted here in a graphical format:

Oh, you say you want a specific prediction from these sell-side analysts? Well there are many to choose from such as:

As two example of a sell side analyst predictions, MKM Partners analyst Rob Sanderson recently raised his 12-month price target on Netflix to $320 and BTIG analyst Rich Greenfield has a $330 price target. Opinions on the Netflix stock price are like belly buttons, everyone has one! It is easy to find one that fits with a preconceived notion on valuation unless it is to sell. There is one lonely analyst says NASDAQ and one other source I saw said there are actually three.

Some people reading this blog post will inevitably jump up and down about Netflix’s P/E multiple and other ratios. “The P/E ratio is too high!” they will exclaim. This jumping up and down is good exercise for them (it has cardiovascular and other health benefits). I will not take a position on whether the current NFLX stock price is too high or too low. I will say that I would make my analysis based on future absolute dollar cash flows and not current accounting earnings, especially since since it is a rapidly growing subscription business. I believe that accounting earnings are a lousy way to evaluate a business since I am an advocate of the approach suggested by Michael Mauboussin (the co-author of the book Expectations Investing): “Management can achieve earnings growth not only when it is investing at or above the cost of capital but also when it is investing below the cost of capital.” Mauboussin’s view is:

In making an assessment of Netflix these “up and down jumpers” that have love for PE ratios (CAPE!) should keep in mind while they are improving their fitness that Netflix’s streaming business = servers mostly from third party provider, people, customer data and algorithms. Netflix is fundamentally a software-based platform. In valuing a company like Netflix it is wise to recall that it does not make its growth investments in assets like video stores which can be depreciated over time under accounting rules. A software-based services company like Netflix is investing in assets like machine learning systems that create better sales funnels (lowers customer acquisition cost), machine learning systems (improves customer retention) and algorithms that reduce the amount of bandwidth required to stream video (increases gross margin and retention). Each of these types of software improves the unit economics of Netflix’s business and helps create a competitive advantage. BTW, Netflix is a software business. Is Hulu a software business? Is HBO?

Some readers of this blog may be asking about now: “Can’t you at least give us a little more guidance?” I can suggest that they look at the Netflix unit economics using tools I have discussed before on this blog. This will give them a bottoms up view that they can reconcile with their top down analysis in their search for greater forecasting accuracy. Here is one set of Netflix numbers based on some educated guesses about inputs into LTV like churn and customer acquisition cost:

Here is another version:

There are a massive number of combinations you can consider by changing the variables in blue in a spreadsheet. How much can Netflix raise prices and what would the impact be? What would churn be in the business if a recession started? Will gross margins change over time? How many subscribers can Netflix can generate? You can do your own math using your own assumptions and your own spreadsheet. Doing this sensitivity analysis is a lot of work, but investing isn’t supposed to be easy. If investing was easy, everyone would be rich.

The rest of my blog post on Reed Hasting and the end notes for this post will appear next week since this is getting to be too long already. I have the data that says that many of the people who start an essay like this will have quit by now. An essay of 7,000 words would have been a bridge too far. Thanks for reading if you are still here.

Like this:

Like Loading...

Lessons from Phil Knight about Business and Being an Entrepreneur

Categories: Uncategorized

The Top 20 Reasons Startups Fail

Fri, 02 Feb 2018 00:00:00 +0800

From lack of product-market fit to disharmony on the team, we break down the top 20 reasons for startup failure by analyzing 101 startup failure post-mortems.

After we compiled our list of startup failure post-mortems, one of the most frequent requests we got was to use these posts to figure out the main reasons startups failed.

Startups, corporations, investors, economic development folks, academics, and journalists all wanted some insight into the question:

“What are the reasons startups fail?”

So we gave those post-mortems the CB Insights’ data treatment to see if we could answer this question.

After reading through every single of the 101 postmortems, we’ve learned there is rarely one reason for a single startup’s failure. However, we did begin to see a pattern to these stories.

And so after sifting through the post-mortems, we identified the top 20 reasons startups failed.

Since many startups offered multiple reasons for their failure, you’ll see the chart highlighting the top 20 reasons doesn’t add up to 100% (it far exceeds it).

Following the chart is an explanation of each reason and relevant examples from the postmortems.

There is certainly no survivorship bias here. But many very relevant lessons for anyone in the entrepreneurial ecosystem.

It’s worth noting that this type of data-driven analysis would not be possible without a number of founders being courageous enough to share stories of their startup’s demise with the world. So a big thank you to them.

Not pivoting away or quickly enough from a bad product, a bad hire, or a bad decision was cited as a reason for failure in 7% of the post mortems. Dwelling or being married to a bad idea can sap resources and money as well as leave employees frustrated by a lack of progress.

As Keith Nowak writes in Imercive’s post-mortem:

“We were caught mid-pivot – half way between a strategy we knew wouldn’t work and one which we believed could be successful but was not able to be aggressively pursued. This was a very difficult place to be both professionally and personally. We were extremely frustrated at not being able to properly go after our new strategy and every day that passed without meaningful progress was one step closer to the failure of my first company. Even though we put everything we had into getting through this phase we were never able to make it through the pivot.”

Work-life balance is not something that startup founders often get and so the risk of burning out is high. Burn out was given as a reason for failure 8% of the time The ability to cut your losses where necessary and re-direct your efforts when you see a dead end was deemed important to succeeding and avoiding burnout, as was having a solid, diverse, and driven team so that responsibilities can be shared.

We often hear about startup entrepreneurs lamenting their lack of network or investor connections so we were surprised to see that one of the reasons for failure was entrepreneurs who said they did not properly utilize their own network.

As Kiko wrote,

“Get your investors involved. Your investors are there to help you. Get them involved from the start, and don’t be afraid to ask for help. I think we made the mistake early on of trying to do (and know) everything ourselves, perhaps out of insecurity over being so new to the business world. This is a mistake.”

Sometimes a startup can evolve from a simple idea to a world of legal complexities that can prove to be a core cause of shutting down.

As Decide.com wrote in their post-mortem,

“We received a notice from them informing us we weren’t compliant and unless we removed it they’d suspend our affiliate account. We weren’t making a lot of money but that account probably represented more than 80% our revenue.”

A couple of music startup post-mortems also associated the high costs of dealing with record labels and legal headaches as a reason for startup failure.

High-profile startup Turntable.fm wrote,

“Ultimately, I didn’t heed the lessons of so many failed music startups. It’s an incredibly expensive venture to pursue and a hard industry to work with. We spent more than a quarter of our cash on lawyers, royalties and services related to supporting music. It’s restrictive. We had to shut down our growth because we couldn’t launch internationally.”

Tying to the more common reason of running out of cash, a number of startup founders explicitly cited a lack of investor interest either at the seed follow-on stage (the Series A Crunch) or at all.

Location was an issue in a couple of different ways. The first was that there has to be congruence between your startup’s concept and location.

As Meetro wrote,

“We launched our product and got all of our friends in Chicago on it. We then had the largest papers in the area do nice detailed write-ups on us. Things were going great …The problem we would soon find out was that having hundreds of active users in Chicago didn’t mean that you would have even two active users in Milwaukee, less than a hundred miles away, not to mention any in New York or San Francisco. The software and concept simply didn’t scale beyond its physical borders.”

Location also played a role in failure for remote teams. The key being that if your team is working remotely, make sure you find effective communication methods, otherwise lack of teamwork and planning could lead to failure.

As Devver wrote,

“The most significant drawback to a remote team is the administrative hassle. It’s a pain to manage payroll, unemployment, insurance, etc in one state … for a small team, it was a major annoyance and distraction.”

There are many good ideas out there in the world, but 9% of startup post-mortem founders found that a lack of passion for a domain and a lack of knowledge of a domain were key reasons for failure no matter how good an idea is.

In their post-mortem, NewsTilt candidly spoke about their lack of interest in the domain they selected — writing:

“I think it’s fair to say we didn’t really care about journalism. We started by building a commenting product which came from my desire for the perfect commenting system for my blog. This turned into designing the best damn commenting system ever, which led to figuring out an ideal customer: newspapers…

“But we didn’t really care about journalism, and weren’t even avid news readers. If the first thing we did every day was go to news.bbc.co.uk, we should have been making this product. But even when we had NewsTilt, it wasn’t my go-to place to be entertained, that was still Hacker News and Reddit. And how could we build a product that we were only interested in from a business perspective.”

Pivots like Burbn to Instagram or ThePoint to Groupon can go extraordinarily well. Or they can start you down the wrong road.

As Flowtab’s post-mortem explains,

“Pivoting for pivoting’s sake is worthless. It should be a calculated affair, where changes to the business model are made, hypotheses are tested, and results are measured. Otherwise, you can’t learn anything.”

Discord with a cofounder was a fatal issue for startup post-mortem companies. But acrimony isn’t limited to the founding team, and when things go bad with an investor, it can get ugly pretty quickly as evidenced in the case of ArsDigital.

Phillip Greenspun writes:

“For roughly one year Peter Bloom (General Atlantic), Chip Hazard (Greylock), and Allen Shaheen (CEO) exercised absolute power over ArsDigita Corporation. During this year they

spent $20 million to get back to the same revenue that I had when I was CEO
declined Microsoft’s offer (summer 2000) to be the first enterprise software company with a .NET product …
deprecated the old feature-complete product (ACS 3.4) before finishing the new product (ACS 4.x) …
created a vastly higher cost structure; I had 80 people mostly on base salaries under $100,000 and was bringing in revenue at the rate of $20 million annually. The ArsDigita of Greylock, General Atlantic, and Allen had nearly 200 with lots of new executive positions at $200,000 or over …
surrendered market leadership and thought leadership”

Getting sidetracked by distracting projects, personal issues, and/or general loss of focus was mentioned in 13% of stories as a contributor to failure.

As MyFavorites wrote at the end of their startup experience,

“Ultimately when we came back from SXSW, we all started losing interest, the team was all wondering where this was eventually going, and I was wondering if I even wanted to run a startup, have investors, have the responsibility of employees and answering to a board of investors.”

If you release your product too early, users may write it off as not good enough and getting them back may be difficult if their first impression of you is negative. And if you release your product too late, you may have missed your window of opportunity in the market.

As a Calxeda employee said,

“In [Calxeda’s] case, we moved faster than our customers could move. We moved with tech that wasn’t really ready for them – ie, with 32-bit when they wanted 64-bit. We moved when the operating-system environment was still being fleshed out – [Ubuntu Linux maker] Canonical is all right, but where is Red Hat? We were too early.”

Ignoring users is a tried and true way to fail. Tunnel vision and not gathering user feedback are fatal flaws for most startups.

For instance, eCrowds, a web content management system company, said,

“We spent way too much time building it for ourselves and not getting feedback from prospects — it’s easy to get tunnel vision. I’d recommend not going more than two or three months from the initial start to getting in the hands of prospects that are truly objective.”

Similarly, VoterTide wrote,

“We didn’t spend enough time talking with customers and were rolling out features that I thought were great, but we didn’t gather enough input from clients. We didn’t realize it until it was too late. It’s easy to get tricked into thinking your thing is cool. You have to pay attention to your customers and adapt to their needs.”

Knowing your target audience and knowing how to get their attention and convert them to leads and ultimately customers is one of the most important skills of a successful business. But an inability to market was a common failure especially among founders who liked to code or build product but who didn’t relish the idea of promoting the product.

As Overto wrote,

“Thin line between life and death of internet service is a number of users. For the initial period of time the numbers were growing systematically. Then we hit the ceiling of what we could achieve effortlessly. It was a time to do some marketing. Unfortunately no one of us was skilled in that area. Even worse, no one had enough time to fill the gap. That would be another stopper if we dealt with the problems mentioned above.”

Failed founders seem to agree that a business model is important – staying wedded to a single channel or failing to find ways to make money at scale left investors hesitant and founders unable to capitalize on any traction gained.

As Tutorspree wrote,

“Although we achieved a lot with Tutorspree, we failed to create a scalable business … Tutorspree didn’t scale because we were single channel dependent and that channel shifted on us radically and suddenly. SEO was baked into our model from the start, and it became increasingly important to the business as we grew and evolved. In our early days, and during Y Combinator, we didn’t have money to spend on acquisition. SEO was free so we focused on it and got good at it.”

Bad things happen when you ignore what a users wants and need, whether consciously or accidentally.

Here’s what GameLayers wrote on their product UI,

“Ultimately I believe PMOG lacked too much core game compulsion to drive enthusiastic mass adoption. The concept of “leave a trail of playful web annotations” was too abstruse for the bulk of folks to take up. Looking back I believe we needed to clear the decks, swallow our pride, and make something that was easier to have fun with, within the first few moments of interaction.”

Pricing is a dark art when it comes to startup success, and startup post-mortems highlight the difficulty in pricing a product high enough to eventually cover costs but low enough to bring in customers.

Delight IO saw this struggle in multiple ways, writing,

“Our most expensive monthly plan was US$300. Customers who churned never complained about the price. We just didn’t deliver up to their expectation. We originally priced by the number of recording credits. Since our customers had no control on the length of the recordings, most of them were very cautious on using up the credits. Plans based on the accumulated duration of recordings make much more sense for us and the number of subscription showed.”

Despite the platitudes that startups shouldn’t pay attention to the competition, the reality is that once an idea gets hot or gets market validation, there may be many entrants in a space. And while obsessing over the competition is not healthy, ignoring them was also a recipe for failure in 19% of the startup failures.

Mark Hedland of Wesabe talked about this in his post-mortem stating:

“Between the worse data aggregation method and the much higher amount of work Wesabe made you do, it was far easier to have a good experience on Mint, and that good experience came far more quickly. Everything I’ve mentioned — not being dependent on a single source provider, preserving users’ privacy, helping users actually make positive change in their financial lives — all of those things are great, rational reasons to pursue what we pursued. But none of them matter if the product is harder to use.”

A diverse team with different skill sets was often cited as being critical to the success of a company. Failure post-mortems often lamented that “I wish we had a CTO from the start,” or wished that the startup had “a founder that loved the business aspect of things.”

Standout Jobs wrote in their post-mortem,

“…The founding team couldn’t build an MVP on its own. That was a mistake. If the founding team can’t put out product on its own (or with a small amount of external help from freelancers) they shouldn’t be founding a startup. We could have brought on additional co-founders, who would have been compensated primarily with equity versus cash, but we didn’t.”

In some cases, the founding team wished they had more checks and balances. As Nouncer’s founder wrote, “This brings me back to the underlying problem I didn’t have a partner to balance me out and provide sanity checks for business and technology decisions made.”

Money and time are finite and need to be allocated judiciously. The question of how should you spend your money was a frequent conundrum and reason for failure cited by startups (29%).

As the team at Flud exemplified, running out of cash was often tied to other reasons for startup failure including failure to find product-market fit and failed pivots,

“In fact what eventually killed Flud was that the company wasn’t able to raise this additional funding. Despite multiple approaches and incarnations in pursuit of the ever elusive productmarket fit (and monetization), Flud eventually ran out of money — and a runway.”

Tackling problems that are interesting to solve rather than those that serve a market need was cited as the No. 1 reason for failure, noted in 42% of cases.

As Patient Communicator wrote,

“I realized, essentially, that we had no customers because no one was really interested in the model we were pitching. Doctors want more patients, not an efficient office.”

Treehouse Logic applied the concept more broadly in their post-mortem, writing,

“Startups fail when they are not solving a market problem. We were not solving a large enough problem that we could universally serve with a scalable solution. We had great technology, great data on shopping behavior, great reputation as a though leader, great expertise, great advisors, etc, but what we didn’t have was technology or business model that solved a pain point in a scalable way.”

Foot In Mouth: 42 Quotes From Big Corporate Execs Who Laughed Off Disruption When It Hit

Fri, 02 Feb 2018 00:00:00 +0800

With the iPhone as a glaring example, and stretching through Amazon and e-commerce and more recently blockchain and bitcoin, many innovations have initially been met with derision by big company CEOs. History often proved them wrong.

Sometimes disruption is staring you right in the face and you can’t help but look the other way. That’s doubly true when you’re the head of a legacy business desperate to stay relevant.

Small players have historically changed the business landscape when the older guys aren’t aware enough or nimble enough to respond. Here are some examples of big CEOs and executives who minimized disruptive threats when they appeared. When asked about these smaller players at the time (and even now) some companies have remained woefully dismissive.

1. The iPhone and iPod

Perhaps the most glaring example of the tendency to dismiss innovative new players was when Apple’s iPhone hit the market in 2007. It was the first example of a smart mobile phone.

Onlookers were unsure what this device would compete with: Would it hurt mobile phone makers like Nokia or developers of handheld planners like Palm, who were also trying to enter into the phone space?

At the time, both companies refused to admit the threat. Here’s Palm’s CEO:

“We’ve learned and struggled for a few years here figuring out how to make a decent phone … PC guys are not going to just figure this out,” said then-Palm CEO Ed Colligan in 2006, after news that Apple was developing a phone. “They’re not going to just walk in.”

Here’s the head of strategy at Nokia:

“The development of mobile phones will follow a similar path to that followed by PCs,” said Nokia’s Chief Strategy Officer Anssi Vanjoki, in a German interview (translated through Google Translate). “Even with the Mac, Apple attracted a lot of attention at first, but they have remained a niche manufacturer. That will be their role in mobile phones as well.”

Microsoft CEO Steve Ballmer had this to say about the iPhone’s lack of a physical keyboard:

“500 dollars? Fully subsidized? With a plan? I said that is the most expensive phone in the world. And it doesn’t appeal to business customers because it doesn’t have a keyboard. Which makes it not a very good email machine.”

RIM’s co-CEO, Jim Balsillie, wrote off the iPhone almost completely:

“It’s kind of one more entrant into an already very busy space with lots of choice for consumers … But in terms of a sort of a sea-change for BlackBerry, I would think that’s overstating it.”

Of course, let’s not forget that phone manufacturers had been wrongly dismissive of Apple before. Here’s Motorola’s CEO on the iPod:

“Screw the Nano. What the hell does the Nano do? Who listens to 1,000 songs?” said Motorola CEO Ed Zander, speaking at a conference in 2006.

Even Apple founder Steve Jobs voiced hesitations about variations on the original iPhone, arguing that phones with larger screens would never catch on:

“You can’t get your hand around it,” Jobs said when Samsung and Google came out with phones larger than the 3.5” original iPhone. “No one’s going to buy that.”

Now, the iPhone X, released November 2017, is a full inch taller than its original iPhone predecessor. The iPhone X also features Apple’s biggest phone screen, measuring 5.8” on the diagonal — 66% larger than the first iPhone.

2. Amazon

Perhaps more than any other internet company, Amazon has embodied internet-based challenges to traditional business and commerce. From the moment it began shipping books and threatening brick-and-mortar book chains, CEO Jeff Bezos (pictured above) has set his sights on transforming industries with Amazon’s customer focus and obsession with efficiency and logistics. But also from the beginning, Amazon has been underestimated by incumbents.

Here’s IBM’s chairman minimizing how Amazon might transform retail and internet sales all the way back in 1999.

“Amazon.com is a very interesting retail concept, but wait till you see what Wal-Mart is gearing up to do,” he said [IBM Chairman, Louis V. Gerstner Jr.]. Mr. Gerstner noted that last year IBM’s Internet sales were five times greater than Amazon’s. Mr. Gerstner boasted that IBM “is already generating more revenue, and certainly more profit, than all of the top Internet companies combined.”

More recently, on a March 2016 earnings call, FedEx Executive VP Mike Glenn minimized the impact of Amazon’s decision to lease aircraft and buy its own trucks to get into the logistics and shipping business.

“While recent stories and reports of a new entity competing with the three major carriers in the United States grabs headlines, the reality is it would be a daunting task requiring tens of billions of dollars in capital and years to build sufficient scale and density to replicate existing networks like FedEx,” [Mike] Glenn said.

In August 2017, as news built around the wave of retail bankruptcies and the “retailapocalypse,” Footlocker’s CEO and chairman Richard Johnson claimed on an earnings call he wasn’t worried about vendors going directly to Amazon to sell expensive sneakers.

“We do not believe our vendors selling product directly on Amazon is an imminent threat. There is no indication that any of our vendors intend to sell premium athletic product, $100-plus sneakers that we offer, directly via that sort of distribution channel.”

In January 2018, Amazon launched its automated cashier-less Amazon Go store in Seattle. The store leverages technology like machine vision, AI, and sensor technology to allow customers to shop without checking out at a register: shoppers are simply automatically charged for their items via the Amazon Go app.

Following the opening of the Amazon Go store, Saks Fifth Avenue president Marc Matrick announced he was unconcerned about the threat automation poses to traditional brick-and-mortar establishments:

“When you think about the online versus the offline experience, we don’t need AI in our stores,” he said on CNBC. “We have ‘I’. We have living, breathing, 4,500 style advisors in our stores.”

3. cloud software

Way back before most enterprise software solutions were cloud-based, Salesforce was offering a new and different kind of CRM product. While it made waves early on, legacy software companies didn’t think too highly of the product.

“Microsoft will roll [Salesforce] over,” Thomas Siebel of Siebel Systems flatly told Bloomberg in 2003. “They get Zambonied.”

Execs have also said they’re not worried about Amazon Web Services (AWS), a cloud services platform that users can leverage to build their own apps. In October 2017, Oracle CEO Mark Hurd went on record to discuss his company’s cloud infrastructure offerings.

“I don’t really worry so much about [AWS], to be very blunt with you,” Hurd told CNBC. “We need to worry about ourselves. We’re in a great position.”

4. Netflix/video streaming

Video streaming service Netflix launched as a potential threat to Blockbuster with its DVD mailing service. Not only did it kill the video renting industry, it now has its eyes set on broadcast television. And leaders in both industries were reluctant to admit that Netflix was a real threat.

“Neither RedBox nor Netflix are even on the radar screen in terms of competition,” said Blockbuster CEO Jim Keyes, speaking to the Motley Fool in 2008. “It’s more Wal-Mart and Apple.”

Here’s Time Warner’s CEO:

“It’s a little bit like, is the Albanian army going to take over the world? I don’t think so,” said Time Warner CEO Jeffrey L Bewkes, when asked in 2010 what he thought about the company’s push toward licensed content.

Even more recently, traditional TV executives have tried to minimize the risk.

“The notion that [companies like Netflix] are replacing broadcast TV may not be quite accurate,” said Alan Wurtzel, NBCU president of research and media development, as he presented ratings of popular shows from streaming companies earlier this month. “I think we need a little bit of perspective when we talk about the impact of Netflix.”

Movie theater chain execs have also tried to sound brave in the face of the Netflix threat. Here’s Cinemark’s CEO on a February 2016 earnings call.

“We don’t have any concern there,” said Mark Zoradi, Cinemark CEO. “Netflix is a great service, it’s a great in-home service. They’ve had other movies. Netflix is very much a television network and not unlike what HBO and Showtime have done for years, they have some original product that goes out there. So it’s not playing in theaters, it’s playing on Netflix and we hope they have great success with it but I don’t see it as an issue relative to the theatrical business. It’s not one really that we talk about.”

5. Airbnb

In the next few years, Airbnb could overtake major hotel chains in total guest bookings, according to a Barclays report noted in Quartz, but hotels are still reluctant to speak openly about the Airbnb threat.

“Our guests don’t want the Airbnb feel and scent,” said Christopher Norton, EVP of global product and operations at the Four Seasons, speaking to Fast Company a couple of years ago. He went on to explain that his customers expect a “level of service that is different, more sophisticated, detailed, and skillful.”

“We have not seen a direct effect [from Airbnb] in any of our hotels … We don’t feel it’s having any impact on our results or that it has hit our radar as of yet,” said Richard Jones, senior VP and COO of Hospitality Ventures Management Group, in 2014.

6. Robo-advisors

Companies like Wealthfront and Betterment believe their computer algorithms will be better at investing customers’ money than personal financial advisers. But some finance professionals are not believers.

“[Wealth management] services require educated, credentialed, experienced advisors acting as fiduciaries on behalf of clients and actively engaged in a relationship with them,” said Jim Maurer, director of personal finance at Buckingham and The BAM Alliance, writing in an op-ed in CNBC, and he concludes, “I don’t see their services as competing with comprehensive wealth management.”

In a March 2016 note, Citigroup’s fintech analysts said robo-advisors may be all well and good, but that they won’t work for the wealthy clients that are the bread-and-butter of the big bank-wealth management industry.

“We see the advent of robo-advice as an example of automation improving the productivity of traditional investment advisers, and not a situation where there is significant risk of job substitution,” Citi analysts led by Ronit Ghose wrote in their report. “Higher net worth or more sophisticated investors will, in our view, always demand face-to-face advice.”

7. Wearables/smartwatches

The Apple Watch was released last year, and it’s still unclear whether or not it will become mainstream. All the same, wearables are more and more looking like they will have an impact on the traditional watch industry.

The Apple watch is an interesting toy, but not a revolution,” said Swatch executive Nick Hayek Jr., speaking to a Swiss newspaper.

Hayek was also quoted as saying in The Guardian that he doesn’t believe in a watch company handling healthcare data.

“I personally don’t want my blood pressure and blood sugar values stored in the cloud, or on servers in Silicon Valley … I cannot accept the responsibility of whether my device warns a customer in time before a heart attack.”

8. The Connected Car/Tesla

Google and Apple are exploring driverless car technologies. Carmakers, however, haven’t been so sure tech companies taking the automotive reigns is the future.

Daimler:

“If there were a rumour that Mercedes or Daimler planned to start building smartphones then they (Apple) would not be sleepless at night, said Daimler head Dieter Zetsche, speaking to Motoring.com.au last year. “And the same applies to me.”

Porsche’s CEO was more blunt:

“An iPhone belongs in your pocket, not on the road,” said Porsche CEO Oliver Blume said.

Jaguar Land Rover:

“We don’t consider customers cargo,” said Jaguar’s head of R&D, Wolfgang Epple, in 2015. “We don’t want to build a robot that delivers the cargo from A to B.”

General Motors:

“I think somebody is kind of trying to cough up a hairball here,” said former General Motors CEO Dan Akerson, when asked about the possibility that Apple was developing its own car. “A lot of people who don’t ever operate in [the car manufacturing industry] don’t understand and have a tendency to underestimate.”

General Motors, of course, later hopped on the autonomous vehicle bandwagon in a big way. And even when they talk about autonomous vehicles, big auto is keen to dismiss the work of newer competitors such as Tesla, which was founded as a software-driven, electrical vehicle company from the start.

Volvo:

“It gives you the impression that it’s doing more than it is,” said Trent Victor, senior technical leader of crash avoidance at Volvo, in an interview with The Verge. “[Tesla’s Autopilot] is more of an unsupervised wannabe.” In other words, Tesla is creating a semi-autonomous car that appears to be autonomous.”

9. Bitcoin/blockchain

Bitcoin has had its share of troubles — including the Mt. Gox scandal, price volatility, and an increasing concentration of mining activity in China — but even with these challenges it’s still the leading decentralized currency and has spawned financial industry interest in applications of its underlying blockchain ledger tech.

JPMorgan CEO Jamie Dimon opined plainly in late 2015 that bitcoin would not survive:

“This is my personal opinion, there will be no real, non-controlled currency in the world,” said Dimon. “There is no government that’s going to put up with it for long … there will be no currency that gets around government controls.”

Dimon was at it again two years later on September 12, 2017, criticizing bitcoin and calling it a fraud.

“It’s worse than tulip bulbs,” Dimon said of bitcoin. “It won’t end well. Someone is going to get killed. Currencies have legal support. It will blow up.”

Later in September 2017 Dimon also lashed out at the trend toward Initial Coin Offerings, whereby companies issue tokens, many of them based on the Ethereum blockchain, in order to fund blockchain-based products and applications. A few — such as Tezos, which raised $230M — have raised upwards of $200M.

We recently mapped over 100 companies that raised funds through ICOs. But Dimon wasn’t having any of it.

“It’s creating something out of nothing that to me is worth nothing,” he said. “It will end badly.”

Also in late September 2017, Ray Dalio, founder of hedge fund Bridgewater Associates, echoed Dimon’s comments.

“Bitcoin today — you can’t make much transactions in it. You can’t spend it very easily,” Dalio said on CNBC’s Squawk Box. “It’s not an effective storehold of wealth because it has volatility to it, unlike gold. Bitcoin is a highly speculative market. Bitcoin is a bubble.”

The executives made the comments at around a time when bitcoin dropped to sudden lows with bitcoin dropping close to a price of $3,000 per bitcoin. However, just a little over a week later, bitcoin was again over the $4,000-mark. Dimon’s comments also caused some head-scratching since it was reported some JPMorgan traders were buying bitcoin on behalf of clients as the price dipped.

10. New protein & meatless meat

The increasing momentum behind the meatless food industry — featuring lab-grown meats, seafood substitutes, and insect protein — spells change for the $90B global meat market. Meat incumbents are facing a rising tide of challenges, in the form of new technology from alternative protein/meatless startups, ethically-focused consumers, and environmental concerns.

However, many leaders in the mainstream beef industry are unconcerned about potential disruption of the industry.

“We’ll see what the marketplace says, but we feel good about the growth of beef and the consumer’s confidence in our product,” said Pete McClymont, executive vice president at Nebraska Cattlemen.

Ann Marie Bosshamer, executive director of the Nebraska Beef Council, agrees:

“Our focus truly is letting the consumer know what a nutrient-dense food beef really is,” she said when interviewed about the impact meatless alternatives might have on Nebraska’s cattle industry.

11. The personal computer

In what is perhaps the most famous failed CEO prediction, executives who built their careers on the mainframe and enterprise computers of the pre-PC era were dismissive of the central idea that they would become the bedrock for the success of Microsoft, Intel, and other PC tech giants: that every desk in an office would have a computer, and that every home would have one as well.

“There is no reason anyone would want a computer in their home,” said Ken Olsen, founder of Digital Equipment Corporation, 1977.

“I think there is a world market for maybe five computers,” said Thomas Watson, president of IBM, in 1943.

12. Google

Steve Ballmer is one of tech’s most outspoken CEOs. Early on in Google’s ascendancy, he made his disdain for the search giant known, up to and including discounting its very status as a company.

“Google’s not a real company. It’s a house of cards.”

13. Apple

Stan Shih of Acer claimed that Apple’s foray into computers was an aberration and that the company would fail to make a niche for itself.

“Apple is like a mutant virus, escaping from the traditional structure of the PC industry, but the industry will still eventually build up immunity, thus further blocking this trend, and we believe the size of the non-Apple camp will exceed Apple’s, because this is how the industry normally evolves.”

14. Mobile gaming

Mobile gaming is another industry legacy execs didn’t see coming. While app hits like Angry Birds have become huge, for a long time companies like Nintendo — though aware of the app’s monumental rise — didn’t see any lasting value there and hesitated to get involved.

These mobile games are “candidly disposable from a consumer standpoint,” said Nintendo president Reggie Fils-Aime, speaking in 2011.

15. Kindle

The Kindle was a bold bet for Amazon: from a start selling books, Bezos and the company apparently recognized the potential that ebooks had to offer. Making the ereaders in house was one of the best ways to assure that the burgeoning market could take off. Almost no one saw it coming, though. Not even the often-prescient Steve Jobs, who decried the venture.

“It doesn’t matter how good or bad the product is, the fact is that people don’t read anymore.”

16. The Television

Daryl Zanuck, co-founder of 20th Century Fox thought that the TV itself wouldn’t catch on.

“Television won’t be able to hold onto any market it captures after the first six months. People will soon get tired of staring at a plywood box every night.”

17. The Telephone

This one is a real throwback, but in 1876, Western Union Telegraph company was offered the chance to purchase the patent on Alexander Graham Bell’s telephone. William Orton, President of Western Union, failed to see the value of the newly invented device.

“What use could this company make of an electrical toy?”

Image credits, in order of appearance: Yutaka Tsutano, Steve Jurvetson, Clive Darra, Mike Petrucci, LGE, OuiShare, Philip Taylor, LWYang , Mark Doliner, fdecomite, Blake Patterson

Here’s our guide on how corporations can use startups to innovate faster.

'Fiction is outperforming reality': how YouTube's algorithm distorts truth

Fri, 02 Feb 2018 12:00:28 +0800

An ex-YouTube insider reveals how its recommendation algorithm promotes divisive clips and conspiracy videos. Did they harm Hillary Clinton’s bid for the presidency?

The methodology behind this story

Guillaume Chaslot, an ex-Google software engineer. Photograph: Talia Herman for the Guardian

It was one of January’s most viral videos. Logan Paul, a YouTube celebrity, stumbles across a dead man hanging from a tree. The 22-year-old, who is in a Japanese forest famous as a suicide spot, is visibly shocked, then amused. “Dude, his hands are purple,” he says, before turning to his friends and giggling. “You never stand next to a dead guy?”

Paul, who has 16 million mostly teen subscribers to his YouTube channel, removed the video from YouTube 24 hours later amid a furious backlash. It was still long enough for the footage to receive 6m views and a spot on YouTube’s coveted list of trending videos.

The next day, I watched a copy of the video on YouTube. Then I clicked on the “Up next” thumbnails of recommended videos that YouTube showcases on the right-hand side of the video player. This conveyor belt of clips, which auto-play by default, are designed to seduce us to spend more time on Google’s video broadcasting platform. I was curious where they might lead.

The answer was a slew of videos of men mocking distraught teenage fans of Logan Paul, followed by CCTV footage of children stealing things and, a few clicks later, a video of children having their teeth pulled out with bizarre, homemade contraptions.

I had cleared my history, deleted my cookies, and opened a private browser to be sure YouTube was not personalising recommendations. This was the algorithm taking me on a journey of its own volition, and it culminated with a video of two boys, aged about five or six, punching and kicking one another.

“I’m going to post it on YouTube,” said a teenage girl, who sounded like she might be an older sibling. “Turn around an punch the heck out of that little boy.” They scuffled for several minutes until one had knocked the other’s tooth out.

•••

There are 1.5 billion YouTube users in the world, which is more than the number of households that own televisions. What they watch is shaped by this algorithm, which skims and ranks billions of videos to identify 20 “up next” clips that are both relevant to a previous video and most likely, statistically speaking, to keep a person hooked on their screen.

Company insiders tell me the algorithm is the single most important engine of YouTube’s growth. In one of the few public explanations of how the formula works – an academic paper that sketches the algorithm’s deep neural networks, crunching a vast pool of data about videos and the people who watch them – YouTube engineers describe it as one of the “largest scale and most sophisticated industrial recommendation systems in existence”.

Lately, it has also become one of the most controversial. The algorithm has been found to be promoting conspiracy theories about the Las Vegas mass shooting and incentivising, through recommendations, a thriving subculture that targets children with disturbing content such as cartoons in which the British children’s character Peppa Pig eats her father or drinks bleach.

Lewd and violent videos have been algorithmically served up to toddlers watching YouTube Kids, a dedicated app for children. One YouTube creator who was banned from making advertising revenues from his strange videos – which featured his children receiving flu shots, removing earwax, and crying over dead pets – told a reporter he had only been responding to the demands of Google’s algorithm. “That’s what got us out there and popular,” he said. “We learned to fuel it and do whatever it took to please the algorithm.”

Google has responded to these controversies in a process akin to Whac-A-Mole: expanding the army of human moderators, removing offensive YouTube videos identified by journalists and de-monetising the channels that create them. But none of those moves has diminished a growing concern that something has gone profoundly awry with the artificial intelligence powering YouTube.

Yet one stone has so far been largely unturned. Much has been written about Facebook and Twitter’s impact on politics, but in recent months academics have speculated that YouTube’s algorithms may have been instrumental in fuelling disinformation during the 2016 presidential election. “YouTube is the most overlooked story of 2016,” Zeynep Tufekci, a widely respected sociologist and technology critic, tweeted back in October. “Its search and recommender algorithms are misinformation engines.”

If YouTube’s recommendation algorithm really has evolved to promote more disturbing content, how did that happen? And what is it doing to our politics?

‘Like reality, but distorted’

Play Video

3:13

How YouTube's algorithm distorts reality – video explainer

Those are not easy questions to answer. Like all big tech companies, YouTube does not allow us to see the algorithms that shape our lives. They are secret formulas, proprietary software, and only select engineers are entrusted to work on the algorithm. Guillaume Chaslot, a 36-year-old French computer programmer with a PhD in artificial intelligence, was one of those engineers.

During the three years he worked at Google, he was placed for several months with a team of YouTube engineers working on the recommendation system. The experience led him to conclude that the priorities YouTube gives its algorithms are dangerously skewed.

“YouTube is something that looks like reality, but it is distorted to make you spend more time online,” he tells me when we meet in Berkeley, California. “The recommendation algorithm is not optimising for what is truthful, or balanced, or healthy for democracy.”

Chaslot explains that the algorithm never stays the same. It is constantly changing the weight it gives to different signals: the viewing patterns of a user, for example, or the length of time a video is watched before someone clicks away.

The engineers he worked with were responsible for continuously experimenting with new formulas that would increase advertising revenues by extending the amount of time people watched videos. “Watch time was the priority,” he recalls. “Everything else was considered a distraction.”

Chaslot was fired by Google in 2013, ostensibly over performance issues. He insists he was let go after agitating for change within the company, using his personal time to team up with like-minded engineers to propose changes that could diversify the content people see.

He was especially worried about the distortions that might result from a simplistic focus on showing people videos they found irresistible, creating filter bubbles, for example, that only show people content that reinforces their existing view of the world. Chaslot said none of his proposed fixes were taken up by his managers. “There are many ways YouTube can change its algorithms to suppress fake news and improve the quality and diversity of videos people see,” he says. “I tried to change YouTube from the inside but it didn’t work.”

YouTube told me that its recommendation system had evolved since Chaslot worked at the company and now “goes beyond optimising for watchtime”. The company said that in 2016 it started taking into account user “satisfaction”, by using surveys, for example, or looking at how many “likes” a video received, to “ensure people were satisfied with what they were viewing”. YouTube added that additional changes had been implemented in 2017 to improve the news content surfaced in searches and recommendations and discourage the promotion of videos containing “inflammatory religious or supremacist” content.

It did not say why Google, which acquired YouTube in 2006, waited over a decade to make those changes. Chaslot believes such changes are mostly cosmetic, and have failed to fundamentally alter some disturbing biases that have evolved in the algorithm. In the summer of 2016, he built a computer program to investigate.

The software Chaslot wrote was designed to provide the world’s first window into YouTube’s opaque recommendation engine. The program simulates the behaviour of a user who starts on one video and then follows the chain of recommended videos – much as I did after watching the Logan Paul video – tracking data along the way.

It finds videos through a word search, selecting a “seed” video to begin with, and recording several layers of videos that YouTube recommends in the “up next” column. It does so with no viewing history, ensuring the videos being detected are YouTube’s generic recommendations, rather than videos personalised to a user. And it repeats the process thousands of times, accumulating layers of data about YouTube recommendations to build up a picture of the algorithm’s preferences.

Over the last 18 months, Chaslot has used the program to explore bias in YouTube content promoted during the French, British and German elections, global warming and mass shootings, and published his findings on his website, Algotransparency.com. Each study finds something different, but the research suggests YouTube systematically amplifies videos that are divisive, sensational and conspiratorial.

When his program found a seed video by searching the query “who is Michelle Obama?” and then followed the chain of “up next” suggestions, for example, most of the recommended videos said she “is a man”. More than 80% of the YouTube-recommended videos about the pope detected by his program described the Catholic leader as “evil”, “satanic”, or “the anti-Christ”. There were literally millions of videos uploaded to YouTube to satiate the algorithm’s appetite for content claiming the earth is flat. “On YouTube, fiction is outperforming reality,” Chaslot says.

A voter in Ohio. Trump won the election by just 80,000 votes spread across three swing states. Photograph: Ty Wright/Getty Images

He believes one of the most shocking examples was detected by his program in the run-up to the 2016 presidential election. As he observed in a short, largely unnoticed blogpost published after Donald Trump was elected, the impact of YouTube’s recommendation algorithm was not neutral during the presidential race: it was pushing videos that were, in the main, helpful to Trump and damaging to Hillary Clinton. “It was strange,” he explains to me. “Wherever you started, whether it was from a Trump search or a Clinton search, the recommendation algorithm was much more likely to push you in a pro-Trump direction.”

Trump won the electoral college as a result of 80,000 votes spread across three swing states. There were more than 150 million YouTube users in the US. The videos contained in Chaslot’s database of YouTube-recommended election videos were watched, in total, more than 3bn times before the vote in November 2016.

Even a small bias in the videos would have been significant. “Algorithms that shape the content we see can have a lot of impact, particularly on people who have not made up their mind,” says Luciano Floridi, a professor at the University of Oxford’s Digital Ethics Lab, who studies the ethics of artificial intelligence. “Gentle, implicit, quiet nudging can over time edge us toward choices we might not have otherwise made.”

Promoting conspiracy theories

Chaslot sent me a database of more YouTube-recommended videos his program identified in the three months leading up to the presidential election. It contained more than 8,000 videos – all of them detected by his program appearing “up next” on 12 dates between August and November 2016, after equal numbers of searches for “Trump” and “Clinton”.

It was not a comprehensive set of videos and it may not have been a perfectly representative sample. But it was, Chaslot said, a previously unseen dataset of what YouTube was recommending to people interested in content about the candidates – one snapshot, in other words, of the algorithm’s preferences.

Jonathan Albright, research director at the Tow Center for Digital Journalism, who reviewed the code used by Chaslot, says it is a relatively straightforward piece of software and a reputable methodology. “This research captured the apparent direction of YouTube’s political ecosystem,” he says. “That has not been done before.”

I spent weeks watching, sorting and categorising the trove of videos with Erin McCormick, an investigative reporter and expert in database analysis. From the start, we were stunned by how many extreme and conspiratorial videos had been recommended, and the fact that almost all of them appeared to be directed against Clinton.

Some of the videos YouTube was recommending were the sort we had expected to see: broadcasts of presidential debates, TV news clips, Saturday Night Live sketches. There were also videos of speeches by the two candidates – although, we found, the database contained far more YouTube-recommended speeches by Trump than Clinton.

But what was most compelling was how often Chaslot’s software detected anti-Clinton conspiracy videos appearing “up next” beside other videos.

There were dozens of clips stating Clinton had had a mental breakdown, reporting she had syphilis or Parkinson’s disease, accusing her of having secret sexual relationships, including with Yoko Ono. Many were even darker, fabricating the contents of WikiLeaks disclosures to make unfounded claims, accusing Clinton of involvement in murders or connecting her to satanic and paedophilic cults.

One video that Chaslot’s data indicated was pushed particularly hard by YouTube’s algorithm was a bizarre, one-hour film claiming Trump’s rise was predicted in Isaiah 45. Another was entitled: “BREAKING: VIDEO SHOWING BILL CLINTON RAPING 13 YR-OLD WILL PLUNGE RACE INTO CHAOS ANONYMOUS CLAIMS”. The recommendation engine appeared to have been particularly helpful to the Alex Jones Channel, which broadcasts far-right conspiracy theories under the Infowars brand.

The conspiracy theorist and talkshow host Alex Jones. Photograph: Brooks Kraft/Getty Images

There were too many videos in the database for us to watch them all, so we focused on 1,000 of the top-recommended videos. We sifted through them one by one to determine whether the content was likely to have benefited Trump or Clinton. Just over a third of the videos were either unrelated to the election or contained content that was broadly neutral or even-handed. Of the remaining 643 videos, 551 were videos favouring Trump, while only only 92 favoured the Clinton campaign.

The sample we had looked at suggested Chaslot’s conclusion was correct: YouTube was six times more likely to recommend videos that aided Trump than his adversary. YouTube presumably never programmed its algorithm to benefit one candidate over another. But based on this evidence, at least, that is exactly what happened.

‘Leading people down hateful rabbit holes’

“We have a great deal of respect for the Guardian as a news outlet and institution,” a YouTube spokesperson emailed me after I forwarded them our findings. “We strongly disagree, however, with the methodology, data and, most importantly, the conclusions made in their research.”

The spokesperson added: “Our search and recommendation systems reflect what people search for, the number of videos available, and the videos people choose to watch on YouTube. That’s not a bias towards any particular candidate; that is a reflection of viewer interest.”

It was a curious response. YouTube seemed to be saying that its algorithm was a neutral mirror of the desires of the people who use it – if we don’t like what it does, we have ourselves to blame. How does YouTube interpret “viewer interest” – and aren’t “the videos people choose to watch” influenced by what the company shows them?

Offered the choice, we may instinctively click on a video of a dead man in a Japanese forest, or a fake news clip claiming Bill Clinton raped a 13-year-old. But are those in-the-moment impulses really a reflect of the content we want to be fed?

Tufekci, the sociologist who several months ago warned about the impact YouTube may have had on the election, tells me YouTube’s recommendation system has probably figured out that edgy and hateful content is engaging. “This is a bit like an autopilot cafeteria in a school that has figured out children have sweet teeth, and also like fatty and salty foods,” she says. “So you make a line offering such food, automatically loading the next plate as soon as the bag of chips or candy in front of the young person has been consumed.”

Once that gets normalised, however, what is fractionally more edgy or bizarre becomes, Tufekci says, novel and interesting. “So the food gets higher and higher in sugar, fat and salt – natural human cravings – while the videos recommended and auto-played by YouTube get more and more bizarre or hateful.”

But why would a bias toward ever more weird or divisive videos benefit one candidate over another? That depends on the candidates. Trump’s campaign was nothing if not weird and divisive. Tufekci points to studies showing that “field of misinformation” largely tilted anti-Clinton before the election. “Fake news providers,” she says, “found that fake anti-Clinton material played much better with the pro-Trump base than did fake anti-Trump material with the pro-Clinton base.”

She adds: “The question before us is the ethics of leading people down hateful rabbit holes full of misinformation and lies at scale just because it works to increase the time people spend on the site – and it does work.”

Tufekci was one of several academics I shared our research with. Philip Howard, a professor at the Oxford Internet Institute, who has studied how disinformation spread during the election, was another. He questions whether a further factor might have been at play. “This is important research because it seems to be the first systematic look into how YouTube may have been manipulated,” he says, raising the possibility that the algorithm was gamed as part of the same propaganda campaigns that flourished on Twitter and Facebook.

In testimony to the House intelligence committee, investigating Russian interference in the election, Google’s general counsel, Kent Walker, played down the degree to which Moscow’s propaganda efforts infiltrated YouTube. The company’s internal investigation had only identified 18 YouTube channels and 1,100 videos suspected of being linked to Russia’s disinformation campaign, he told the committee in December – and generally the videos had relatively small numbers of views. He added: “We believe that the activity we found was limited because of various safeguards that we had in place in advance of the 2016 election, and the fact that Google’s products didn’t lend themselves to the kind of micro-targeting or viral dissemination that these actors seemed to prefer.”

General counsels for Twitter, Facebook and Google prepare to testify before the House intelligence committee hearing on Russia’s use of social media to influence the election. Photograph: Shawn Thew/EPA

Walker made no mention of YouTube recommendations. Correspondence made public just last week, however, reveals that Senator Mark Warner, the ranking Democrat on the intelligence committee, later wrote to the company about the algorithm, which he said seemed “particularly susceptible to foreign influence”. The senator demanded to know what the company was specifically doing to prevent a “malign incursion” of YouTube’s recommendation system. Walker, in his written reply, offered few specifics, but said YouTube had “a sophisticated spam and security breach detection system to identify anomalous behavior and malignant incursions”.

Tristan Harris, a former Google insider turned tech whistleblower, likes to describe Facebook as a “living, breathing crime scene for what happened in the 2016 election” that federal investigators have no access to. The same might be said of YouTube. About half the videos Chaslot’s program detected being recommended during the election have now vanished from YouTube – many of them taken down by their creators. Chaslot has always thought this suspicious. These were videos with titles such as “Must Watch!! Hillary Clinton tried to ban this video”, watched millions of times before they disappeared. “Why would someone take down a video that has been viewed millions of times?” he asks.

Quick guide

What you need to know about the Trump-Russia inquiry

How serious are the allegations?

The story of Donald Trump and Russia comes down to this: a sitting president or his campaign is suspected of having coordinated with a foreign country to manipulate a US election. The story could not be bigger, and the stakes for Trump – and the country – could not be higher.

What are the key questions?

Investigators are asking two basic questions: did Trump’s presidential campaign collude at any level with Russian operatives to sway the 2016 US presidential election? And did Trump or others break the law to throw investigators off the trail?

What does the country think?

While a majority of the American public now believes that Russia tried to disrupt the US election, opinions about Trump campaign involvement tend to split along partisan lines: 73% of Republicans, but only 13% of Democrats, believe Trump did “nothing wrong” in his dealings with Russia and its president, Vladimir Putin.

What are the implications for Trump?

The affair has the potential to eject Trump from office. Experienced legal observers believe that prosecutors are investigating whether Trump committed an obstruction of justice. Both Richard Nixon and Bill Clinton – the only presidents to face impeachment proceedings in the last century – were accused of obstruction of justice. But Trump’s fate is probably up to the voters. Even if strong evidence of wrongdoing by him or his cohort emerged, a Republican congressional majority would probably block any action to remove him from office. (Such an action would be a historical rarity.)

What has happened so far?

Former foreign policy adviser George Papadopolous pleaded guilty to perjury over his contacts with Russians linked to the Kremlin, and the president’s former campaign manager Paul Manafort and another aide face charges of money laundering.

When will the inquiry come to an end?

The investigations have an open timeline.

I located a copy of “This Video Will Get Donald Trump Elected”, a viral sensation that was watched more than 10m times before it vanished from YouTube. It was a benign-seeming montage of historical footage of Trump, accompanied by soft piano music. But when I played the video in slow motion, I saw that it contained weird flashes of Miley Cyrus licking a mirror. It seemed an amateurish and bizarre attempt at inserting subliminal, sexualised imagery. But it underscored how little oversight we have over anyone who might want to use YouTube to influence public opinion on a vast scale.

I shared the entire database of 8,000 YouTube-recommended videos with John Kelly, the chief executive of the commercial analytics firm Graphika, which has been tracking political disinformation campaigns. He ran the list against his own database of Twitter accounts active during the election, and concluded many of the videos appeared to have been pushed by networks of Twitter sock puppets and bots controlled by pro-Trump digital consultants with “a presumably unsolicited assist” from Russia.

“I don’t have smoking-gun proof of who logged in to control those accounts,” he says. “But judging from the history of what we’ve seen those accounts doing before, and the characteristics of how they tweet and interconnect, they are assembled and controlled by someone – someone whose job was to elect Trump.”

Chaslot and some of the academics I spoke to felt this social media activity was significant. YouTube’s algorithm may have developed its biases organically, but could it also have been nudged into spreading those videos even further? “If a video starts skyrocketing, there’s no question YouTube’s algorithm is going to start pushing it,” Albright says.

YouTube did not deny that social media propaganda might have influenced its recommendations, but played down the likelihood, stressing its system “does not optimise” for traffic from Twitter or Facebook. “It appears as if the Guardian is attempting to shoehorn research, data and their conclusions into a common narrative about the role of technology in last year’s election,” the spokesperson added. “The reality of how our systems work, however, simply don’t support this premise.”

After the Senate’s correspondence with Google over possible Russian interference with YouTube’s recommendation algorithm was made public last week, YouTube sent me a new statement. It emphasised changes it made in 2017 to discourage the recommendation system from promoting some types of problematic content. “We appreciate the Guardian’s work to shine a spotlight on this challenging issue,” it added. “We know there is more to do here and we’re looking forward to making more announcements in the months ahead.”

Content creators

A video by the Next News Network fed on debunked allegations against Bill Clinton. Photograph: YouTube

With its flashy graphics and slick-haired anchor, the Next News Network has the appearances of a credible news channel. But behind the facade is a dubious operation that recycles stories harvested from far-right publications, fake news sites and Russian media outlets.

The channel is run by anchor Gary Franchi, once a leading proponent of a conspiracy that claimed the US government was creating concentration camps for its citizens. It was the Next News Network that broadcast the fabricated claims about Bill Clinton raping a teenager, although Franchi insists he is not a fake news producer. (He tells me he prefers to see his channel as “commentating on conservative news and opinion”.)

In the months leading up to the election, the Next News Network turned into a factory of anti-Clinton news and opinion, producing dozens of videos a day and reaching an audience comparable to that of MSNBC’s YouTube channel.

Chaslot’s research indicated Franchi’s success could largely be credited to YouTube’s algorithms, which consistently amplified his videos to be played “up next”. YouTube had sharply dismissed Chaslot’s research.

I contacted Franchi to see who was right. He sent me screen grabs of the private data given to people who upload YouTube videos, including a breakdown of how their audiences found their clips. The largest source of traffic to the Bill Clinton rape video, which was viewed 2.4m times in the month leading up to the election, was YouTube recommendations.

The same was true of all but one of the videos Franchi sent me data for. A typical example was a Next News Network video entitled “WHOA! HILLARY THINKS CAMERA’S OFF… SENDS SHOCK MESSAGE TO TRUMP” in which Franchi, pointing to a tiny movement of Clinton’s lips during a TV debate, claims she says “fuck you” to her presidential rival. The data Franchi shared revealed in the month leading up to the election, 73% of the traffic to the video – amounting to 1.2m of its views – was due to YouTube recommendations. External traffic accounted for only 3% of the views.

Franchi is a professional who makes a living from his channel, but many of the other creators of anti-Clinton videos I spoke to were amateur sleuths or part-time conspiracy theorists. Typically, they might receive a few hundred views on their videos, so they were shocked when their anti-Clinton videos started to receive millions of views, as if they were being pushed by an invisible force.

In every case, the largest source of traffic – the invisible force – came from the clips appearing in the “up next” column. William Ramsey, an occult investigator from southern California who made “Irrefutable Proof: Hillary Clinton Has a Seizure Disorder!”, shared screen grabs that showed the recommendation algorithm pushed his video even after YouTube had emailed him to say it violated its guidelines. Ramsey’s data showed the video was watched 2.4m times by US-based users before election day. “For a nobody like me, that’s a lot,” he says. “Enough to sway the election, right?”

Daniel Alexander Cannon, a conspiracy theorist from South Carolina, tells me: “Every video I put out about the Clintons, YouTube would push it through the roof.” His best-performing clip was a video titled “Hillary and Bill Clinton ‘The 10 Photos You Must See’”, essentially a slideshow of appalling (and seemingly doctored) images of the Clintons with voiceover in which Cannon speculates on their health. It has been seen 3.7m times on YouTube, and 2.9m of those views, Cannon said, came from “up next” recommendations.

Chaslot has put a spotlight on a trove of anti-Clinton conspiracy videos that had been hidden in the shadows – unless, that is, you were one of the the millions YouTube served them to. But his research also does something more important: revealing how thoroughly our lives are now mediated by artificial intelligence.

Less than a generation ago, the way voters viewed their politicians was largely shaped by tens of thousands of newspaper editors, journalists and TV executives. Today, the invisible codes behind the big technology platforms have become the new kingmakers.

They pluck from obscurity people like Dave Todeschini, a retired IBM engineer who, “let off steam” during the election by recording himself opining on Clinton’s supposed involvement in paedophilia, child sacrifice and cannibalism. “It was crazy, it was nuts,” he said of the avalanche of traffic to his YouTube channel, which by election day had more than 2m views.

“Breaking news,” he announced in one of his last dispatches before the vote: the FBI, he said, had just found graphic images of Clinton and her aide in “sexually compromising positions” with a teenager. “It seems to me, with Bill Clinton’s trips to paedophile island a number of times, that what we have here is nothing short of the Clinton paedophile ring,” he declared.

Todeschini sits in his darkened living room in New Jersey, staring into his smartphone. “I’ll tell you what: the rabbit hole just got a couple of yards deeper.”

Contact the author: paul.lewis@theguardian.com.

A full description of the methodology Chaslot used to detect YouTube recommendations (and an explanation of how the Guardian analysed them) is available here.

17 Product Managers Who Will Own the Future of NYC Tech — and the 9 Frameworks They’ll Use to Do It | First Round Review

Thu, 01 Feb 2018 00:00:00 +0800

In October, 17 product managers sat around a table at First Round’s New York Office, ready to learn something new. They’d been hand-picked out of 800+ applicants for the First Round Product Program — the first seminar series of its kind for rising star PMs destined to shape the industry for years to come. Instructors included recent product luminaries ranging from Viacom SVP Product Mike Berkley to Ellevest CPO Alexandria Stried to Kit Founder Camille Hearst. But no one knew what to expect.

What happened next was kismet. We watched as the group turned into friends grabbing drinks after each seminar, and ultimately tight-knit colleagues who will follow and support each other through their careers. We also saw them rally around and adopt the frameworks, exercises, and tools shared by their teachers — digging in and applying them at their jobs right away.

To provide a window into the program’s content, we want to share these key lessons learned with the broader community — and give you a sneak peek at who these 17 remarkable product people are. We have conviction that they’ll be building the products and businesses that define NYC tech over the next five years. Get to know them (and the top tools they’ll be using to get there) now.

THE CURRICULUM

The Product Program’s nine sessions were each curated to take on an essential area of product leadership:

At First Round, we believe the best companies are most often built by extraordinary product minds. Even if you’re not a PM right now, you can benefit from adopting the habits and strategies that make talented PMs successful. Tohelp in this pursuit, we’d like to present the most tactical gems emerging from each session, you can click on any of the links above to skip directly to that section of this mini-manual.

1. Getting into the PM Mindset

The product manager role has many definitions, the most popular of which is CEO of the product — but more often it can feel like cat herder or project coordinator. There’s a duality to it. You need to be both a strategic whiz and an execution-oriented jack of all trades. It can be hard to embody all of this at once. To boil it down to manageable size, instructors Aseem Sood, Co-founder of Open Collective, and Steve Jacobs, Founder at Bright, offered this directive: A good PM fills in the gaps and gets out of the way.

Your job is essentially to go from the current state (a defined problem) to a future state of things (where the problem is solved). What gets you there are tactics. To build something truly great, you have to do this again and again, every time you add a feature or make a change.

Prioritization becomes critical. To figure out what you should be tackling first, Jacobs suggests calculating the significance of any given feature or product:

Significance = Magnitude x Number of People Impacted

In the equation, magnitude is a measure of how frustrating/painful/unbearable the problem being solved is. A magnitude 1 problem might cause mild annoyance, whereas a magnitude 3 problem might cause show-stopping frustration and anger. Stack rank what you need to do accordingly.

For each project you prioritize, define the current state and future state. Then list all the tactics you can think of to move toward the latter. As you start taking actions, continually question whether the tactic you’re trying creates more friction than the original problem. If the answer is yes, immediately shift course. Of course, there’s a lot more to the day-to-day of product management, but if you can internalize these few tenets, you have a significant advantage.

2. Figuring Out When to Build What

Knowing when to build or try something is vital. At a startup, it can be easy to fall into a fugue state thinking that everything is equally important and urgent. While the significance equation above can help on a high level, there are some more granular ways to tell if you’re building the right thing at the right time. Two instructors from Flatiron Health, Keith Cowing, Director of Product Management, and Geoff Calkins, VP of Product, gave advice for determining the exact sequence of events a PM should follow to accomplish what they want:

Time-Based Risk

When asking yourself, “Do we need this now?” consider time-based risk, and ask yourself these two specific follow-up questions:

Is there significant risk if we don’t do this soon? What is it? How likely are those things to happen?
What opportunities might we miss if we do not build this thing within a certain timeframe?

For example, maybe a competitor has launched a new version of its product that its customers don’t like as much. This gives you a window of time to entice some of them over to your product if you build and launch the right things — i.e. the features they're now missing.

Building Blocks First

The other follow-up question you should always ask is “How many other projects depend on this thing?” For example, a company might decide to take a year to work on infrastructure that will set them up for future success on a number of dependent projects. You want to enable as much for yourself as you possibly can while also hitting your other priorities.

3. Turning Product Vision into an Executable Strategy

'Product vision' is one of those nebulous terms that gets thrown around tech willy nilly. It should only be used to refer to a company’s North Star — what is it that they actually want to accomplish or solve? But it still remains totally useless without real plans for how to execute toward it. This session of the Product Program, taught by Clark Co-founder Sam Gimbel and Artnet CPO Chris Maliwat dove into how to take a product vision and turn it into smart, systematic action:

Structure your vision wisely. It should be forward looking — several years away at least. Whatever statement you use to define what you’re building, make sure your team is aligned around it, and that every single person understands how their day to day work is heading toward it. It should be useful for guiding everyday decisions on what to build, what to ship and what to prioritize.
Create 2-3 objectives that move you toward that vision. These should be goals (not tasks) accomplishable over one or a few quarters that will make measurable progress toward your North Star — like reducing churn or opening a new market. Arrive at these goals leveraging data coming out of customer conversations. What problems, liabilities and opportunities emerge? Create objectives around them, but not too many at a time.
Place bets under each objective. This is where you get into tasks. Nested under each hypothesis, place a number of bets — specific hypotheses you’re making to achieve those goals and the actions you’ll take to test them. Results should be as quantitatively measurable as possible. Spread out your risk so that you’re only taking on a few high-risk/high-reward projects at a time alongside likelier wins that move the needle.

Illustration by Diana Hsieh.

Following this template, you end up with a quarterly roadmap that has every action and each person’s work closely connected with the company’s direction and purpose. Plus, it sets you up to quickly test, measure and adopt the features that will make the biggest difference for you over time.

4. Effective Stakeholder Communication

Product managers sit at the nexus of many very different types of people, and — like a switchboard — they need to route communications accordingly and keep everyone informed about what’s happening in a language they understand. In this session, instructors Melody Koh, former Head of Product at Blue Apron, and Ellevest CPO Alexandria Stried listed all of the do’s and don’ts that ensure smooth communication with four main stakeholder groups.

Group 1: Executives and leadership

Your mission is to bring a time-deprived group of people up to speed very efficiently. To make that happen:

Do...

Send presentations, decks and other materials before every meeting. For very busy folks, print materials in advance of meetings and block off the first 5 minutes for review.
Validate every decision with data, showing that you’ve thought through all possible outcomes.
Be specific about the executives' desired participation. Outline clearly and succinctly what decisions you need their help making in real-time during the meeting.
Take notes and close the loop. Follow up immediately on open questions and requests.
Send high-level updates right after each meeting with action items, the status of tasks using something clear like red/yellow/green to indicate what's behind schedule, going okay, and going well.

Don’t...

Go into too much detail. Only share what’s important.
Surprise anyone with bad news. If the news is bad, reach out to folks 1:1 in advance so they have time to think. Then set a meeting with everyone to find resolution.
Show up unprepared. There’s no winging it.
Ignore room dynamics. You need to watch people’s reactions to information and adjust in real time. Defuse tensions as they arise. Slow down or speed up as needed.

Group 2: Your own team

These are the people you see and talk to all the time. But information still falls through the cracks, and you have to make sure everyone is moving forward.

Do...

Leverage efficient daily stand-ups. This is your time to nudge, get status updates and call out outstanding action items.
Review strategy/roadmaps regularly. Evaluate whether changes are needed weekly. What data motivates these changes?
Record and send out notes on key decisions and actions. This maintains alignment better than anything else.
Facilitate conversation around roadblocks. Don’t just observe that something isn’t getting done. Get to the bottom of it.
Reward team members often. Highlight and demonstrate the impact of their work. In particular, tell anecdotes about customer pain points that were alleviated.

Don’t...

Make decisions without engineering and design. Autonomous decision making has a destructive ripple effect. And never treat them as subordinates.
Send action items/requests without talking about them first. Don’t just assign things to people through a productivity tool. Chat about it in a 1:1 or stand-up.
Forget to update folks on roadmap or specs changes. This is particularly important after meeting with execs. Run through the changes with your team and explain why they were made.

Group 3: Internal and external partners

This might include internal teams like finance and legal, or external parties like brands you’re partnering with on campaigns. They have shifting needs that PMs need to catch and respond to fast.

Do...

Exhibit detailed understanding of their work and domain.
Use the right format at the right time with the right audience. Make sure you’re conveying information in the way people expect and like to digest it. Decks/whiteboarding/written reports/etc. This varies highly.
Leverage your teammates. Bring in engineering leads and others to make stakeholders feel like they’re hearing from the ultimate source.
Gently and continuously educate them. Partners sometimes don’t know the consequences of their actions. Take the time to diplomatically explain things — like a how a relatively simple-seeming change represents much more work for engineers.
Build relationships outside of work meetings. Go for 1:1 walks or coffees to build rapport.
Create transparency. Don’t rely on others to communicate to everyone. Publicly share and point to roadmaps available to everyone instead.

Don’t...

Forget who to loop in at what stage. Make a list of all the work expected and who needs to know what. Give everyone ample heads up.
Make stakeholders feel ignored. Even if you disregard feedback, follow up and explain.
Forget you have more insight than anyone else. Stakeholders don’t see your roadmap. Err on the side of over-communication.
Allow meetings to end without clarity. At the end of every meeting, state the primary conclusions and action items. Get universal agreement.
Forget to educate about timelines and tradeoffs. Proactively provide this info, don’t wait to be asked about it.

Group 4: Customers

As a PM, you should master talking to customers so their needs are well-represented — and to make the right impact for them.

Do...

Always start with the user problem. Don’t just accept a pain point exists. Ask why and understand the journey that creates that pain point.
Keep what’s important to them top of mind. Constantly refer to user studies and conversations as you speak to other stakeholders.
Treat email copy as a part of the product experience. It’s one of the first impressions you’ll ever make on a customer.
Generate empathy for yourself by reading through user feedback, attending user studies in person, shadowing sales and customer success, etc.

Don’t...

Leave product communications/messaging to the last minute. Messaging can often define product choices, and not the other way around. Start with this, don’t end with it.
Assume marketing will position the product themselves. You bring a ton of context they don’t have. Stay in the room.
Leave customer success in the dark about launch. They’re the front lines. Treat them accordingly.
Believe internal products require no roll out. Internal users are real users. They need to be onboarded, trained and kept engaged.

5. Create Compelling Product Messaging

It doesn’t matter that you’ve built a beautiful product if you don’t have a narrative that properly relays that greatness. Product Program Instructors Mike Berkley and Great Jones Founder and CEO Jay Goldklang, took this head on — explaining how PMs need to (and can) be the best storytellers at their companies

They recommend starting with one question:

What superpower do you want to give your user?

For example, the Amazon Kindle makes reading far more accessible to the average human. The iPhone lets us navigate to unknown places wherever we are in the world. Great products solve core human needs and are presented in a way that makes it easy for people to see, feel and understand their value deeply. This is how a strong product narrative moves people to action.

As a PM, it’s your job to ensure the entire team knows the story you’re trying to create for your users. As noted above, this should come first in your development process, not last. Goldklang relies on Will Carlin’s 5 C’s framework for telling strong stories. Before applying them, however, your goal should be to craft a story around a single user — not a group of users.

Context: Establish the setting and identity of the user you’re talking to.
Conflict: The problem your product attempts to solve for that user.
Conflict Escalation: Really visualize what it’s like for a user to encounter this problem. Draw out the emotions tied to the pain point and solutions that have been tried but failed. Really feel and describe the frustration, disappointment, etc.
Climax: Your product is introduced — what changes for the user?
Conclusion: Detailed description of the improvement in the user’s life.

Use this framework to create a story about your product. Share it in all-hands meetings, working sessions, cross-functional status updates, etc. And remember, no matter what you do, different versions of your story will emerge once it launches. To win, craft the story that is closest and most personal to your user. The more emotionally resonant it is, the more it will drown out competing perspectives.

6. Build Your Best Product Team

The product team you build will have a profound impact on what your company becomes. You have to hire people who aren’t just talented, but who are perfect for your particular business. Instructors Natalie Gibralter, Director of Product at Squarespace, and Kristen Donnino, Adaptly’s VP Product, shared a three-step exercise to ensure you’re on course for hiring your optimal team:

Build a strategic hiring plan
Define key competencies
Standardize your assessment of competencies.

This is an exercise every early-stage company should run prior to scale.

Develop a strategic hiring plan by determining who on your existing team should be a part of the hiring process (all relevant folks the role will interface with), and the concrete steps every candidate will take between application and hire. This will vary by company based on values and culture, but every strategic hiring plan should be designed to extract as much relevant information about a candidate as possible, while familiarizing them with the people they’d be working with. It should also optimize for only investing a lot of time in the best candidates.

Once you have a good plan outlined, share it with everyone you’d have on the interview loop. Let them poke holes in it and get into the details of the parts they’ll each play.

To define a list of vital competencies for a role, bring all stakeholders together in a room and have them call out all the skills and attributes they associate with great product managers. Once you have a healthy list, have everyone silently vote on which competencies they believe are the most important. Consider writing all of them on a whiteboard and give everyone 3 stickers they get to place next to their top choices. When Donnino and Gibralter ran this process with the Product Program class, “High Emotional Intelligence” won the day — with “attention to detail” and “data driven” as runners up.

You want to make sure everyone on your hiring team is aligned around the same top competencies so that people aren’t making judgment calls solely on their own criteria. After you have your list of qualities you’ll focus on, work as a group to clearly and explicitly define what they mean to you. The same word can mean different things to different people. Having a singular definition as a group helps ensure that candidates will be judged fairly and accurately.

To create questions and an evaluation rubric, you can use very similar methods. Again, lead your hiring team through a brainstorm and voting process. Phase 1: For each competency, everyone share their favorite interview questions to assess that trait in a candidate. Phase 2: Run another round of silent voting with stickers till you’ve got the best 3-5 questions. For example, the top two questions to assess emotional intelligence were:

Give me an example of a time you were misunderstood. How did you handle it?
How have you gotten your team out of a failure?

All of the questions you think up might be great, but you want to focus and maximize your time with a candidate. Work with your hiring team to agree on the best questions to ask, and keep them the same for every candidate interviewing.

You’re not done yet, though. You have to similarly decide what good and bad answers to each question sound like. Again, you can have people brainstorm what a great answer to a question might be, and then have the whole group vote to see which type of response emerges as the best. Clearly, candidates won’t be able to hit that exact mark verbatim, but you want to have a sense of what the most impressive answers will include or touch on. And you want everyone who will interview them to have that standard in mind.

Running this exercise is time intensive. You have to run several voting rounds to arrive at competencies, questions for each competence, and then the best and worst responses to each question. Sounds like a lot, but it’s incredibly worth it to have a standardized approach created collaboratively — one that can be recycled and reused again and again as hiring picks up pace.

7. Scale Yourself as a Product Leader

Every startup PM is in a race against time. Moving up depends on your ability to scale yourself and your abilities as fast as your companies are growing. To be smart about this, says instructor and ClassPass Head of Product Justin Chang, PMs should focus on scaling in four areas: decision making, velocity, collaboration and empowerment.

Decision making starts to slow down and crack at a certain point of growth. The warning sign is too many cooks in the kitchen and slowed pace. The antidote is the DACI framework:

Driver: The one person responsible for the project who drives process and keeps everyone aligned. Typically, this is the PM.
Approver: The person who approves the proposal/recommendation for the project. There can only be one approver. Even more senior people in the company hierarchy cannot overrule this person’s approval, once the approving authority has been granted.
Contributors: People working on the project team, providing input, producing work, etc.
Informed: People who are kept in the loop about the project and results, but not directly contributing, and not attending meetings.

According to Chang, this framework holds great power to improve morale, streamline activities, and reduce context switching for everyone.

Velocity of work starts to slow down as tech debt accumulates and teams grow. This can increase costs to a dangerous degree. To fix it, create durable teams around durable problems. For example, ClassPass organized its teams around Acquisition, Studio Experience and Machine Learning. To avoid scope creep and last-minute design changes, Chang recommends the following product development process:

Goal definition: Everyone included in your DACI framework should come together and emerge with a singular goal for the product.
Product definition: Align on scope of the project and what will be required to solve the problem at hand. What is and isn’t out of scope?
Design review: Be explicit about the type of feedback you want and don’t want.
Tech review: Make sure everyone has a chance to debate and buy into the technical approach.
Go/no-go: Review your checklist to make sure the rest of the org is operationally ready for a product/project launch — i.e. customer service has the bandwidth to answer questions, etc.

Collaboration starts to break at a certain company size. You realize it when people’s calendars are packed with meetings just trying to stay informed on what’s being built across the org. People lose track of the roadmap or are blindsided by product launches. Free people up and fuel effective collaboration with these three moves:

Make your product roadmap and product docs accessible to the entire company.
Hold Gate Meetings with everyone in your DACI group. Gate meetings force decisions that must be made to proceed. Put together a product doc with all the context for the decision and start the meeting with a 10-minute silent reading period.
Send decision emails to communicate to all possible stakeholders when big decisions have been made and why at every stage of the development process.

Empowerment at scale becomes important when teams get so big that people feel like they’re just executing on other people’s orders. This can be demoralizing. It shows up when you see people weakly rallying behind decisions and blaming others for failure. There are several strategies to combat this:

Present options instead of a firm decision. Give the people on your team as many choices as you can for how to solve a problem, what outcome they want to achieve, what metrics to measure, what success will look like.
Start milestone meetings with a background share. Go around the room and have everyone briefly state what they know about the project, wish they knew, and what they’re most excited about.

As General Patton would put it, “Never tell people how to do things; tell them what you need them to accomplish and let them surprise you with their ingenuity.”

8. Drive Product Development with Data

PMs use data to align stakeholders with roadmaps, track efficacy of what's been built, and prioritize what to build next. Yet, there’s constant pressure to do more with data and learn faster. To make sure you remain focused and efficient in this arena, instructors Xiaodi Zhang, 1stdibs’ CPO, and Anthony Schrauth, former CPO and CIO at Betterment, offer a simple set of rules:

You have to gather implicit data. Stop making excuses. Yes, you have other priorities, but you must find small ways to quantitatively measure user behavior patterns across your product. If you don’t, you’ll have no real visibility into how users will react to new features. You might implement something you thought would be awesome only to see a backlash. These can be little experiments, like seeing if someone will click on a link.

Don’t underestimate the importance of explicit data. So many startups lean on stats to validate their hypotheses without seeing the bigger picture. Protect yourself against this by taking in qualitative feedback shared directly by your users. Give it equal if not more weight.

Always go to your customers when you observe them. You want to see how people are using your product in their natural habitat. Craft questions based on their unique context about how they’d use the product in specific situations. If they’ve developed any workarounds, take special note.

Find the right users for your questions. At B2B companies, product managers often find themselves engaging with the C-suite at their customers. Their feedback will be different than that of other employees who might be the primary users of your product. Determine who is the most relevant user of your product, and pose the questions to them directly.

Find a meaningful metric for your performance. When you poll your users on how much they like your product, you need to make sure you’re measuring the right thing. What metric will indicate that you’re succeeding with people the way you want to? Net Promoter Score is a common choice — measuring how likely it is that someone will recommend your product. But that’s not universally appropriate.

For example, at Betterment, the team used NPS Score for a while, but then noticed that it was fluctuating with the ups and downs of the stock market. They augmented it with a Customer Effort Score — a measure of whether the company made it easier to perform certain tasks.

9. Going from PM to Founder

In many ways, product management is the ideal springboard for founders. It’s a position that affords you opportunities to go deep in areas that will serve you when running your own business, namely negotiation, P&L and forecasting, legal, hiring and operations. But before you can get into all of that, you need to be sure you’re choosing the right idea to work on. During the final session of the Product Program, instructors Camille Hearst, Co-founder and CEO of Kit, and Hannah Donovan, Co-founder and CEO of Trash TV, sat down with the PMs to talk idea selection.

How do you know if an idea is worth pursuing? They recommend evaluating each one according to Marty Cagan’s Four Big Risks:

Value: Do people want this? When you talk to prospective users, do they see value in what you’re building?
Usability: Can people figure your solution or product out intuitively?
Feasibility: Can you and an eventual team build what you have in mind within a realistic time frame with the resources you can realistically get?
Viability: Is there a clear business model and path to making money?

Before you set out after an idea, make sure you can check each of these boxes and confidently explain your answers to each of these questions to possible investors.

Meet NYC’s Product Leaders of Tomorrow

All of the frameworks and lessons shared above are now tools for the next generation of New York-based product leaders and founders. As we said, the 17 PMs who attended the Product Program were selected out of more than 800 candidates. They are (in our opinion) the people to watch in the NYC tech community. And now that they’ve connected with each other, they’re unstoppable. We highly recommend getting to know them now.

Alex Powell, Product Manager at Greenhouse, grew up in New Jersey (proudly, despite what everyone else says). He decided to stick around the Garden State for four extra years at Princeton before starting his career as an immigration paralegal — back when he thought he was going to go to law school. Luckily, Greenhouse saw his potential and hired him for their support team. He made the shift into product as the first PM aside from the founder. When he’s not at work, he’s quoting obscure TV shows and movies.

Alexander Tsu, Associate Product Manager at Foursquare, oversees enterprise data products and a developer platform with over 120,000 users. It’s his job to make sure that developers can tap into Foursquare’s massive database of 93 million locations to build incredible new products. He studied economics at Dartmouth and computer science at the University of Chicago. Outside of work, he spends his time watching the Golden State Warriors, drinking boba, and singing off-key karaoke.

Ara Cho, Product Manager at Nomad Health, recently joined the team to focus on building out an online recruiting platform that connects hospitals and clinicians. She developed her product chops at OnDeck, helping more small businesses get loans and expanding the company's markets in Canada and Australia. In the process, she learned all about working with international users. She has a BS in business administration from Berkeley and an MBA from Harvard. Outside of work, she’s a volleyball fanatic, heavy consumer of YouTube, and takes advantage of New York to achieve her step goals every day.

Brittney Gwynn, Director of Product at Simple Contacts, is focused on acquisition, conversion, and building technology that delivers care directly to patients at their convenience. She started her career at Google, where she developed an interest in discovering gaps in user experiences and anticipating product usability flaws. She later joined SoundCloud as a PM, where she helped launch new consumer subscription products in several markets. Brittney loves product because it touches all aspects of the business, from marketing and engineering to supporting the entire company vision, all while solving tangible user problems and ideating next level, aspirational user journeys. She graduated from Howard University, is a native Detroiter, loves to travel, and serves as the NYC City Lead for Black Tech Women.

Daniel Layfield, Product Manager at Codecademy, splits his time between the product team and managing external partnerships with governments, tech companies and nonprofits. Codecademy is in a phase of rapid growth, so his role is a combination of finding the right business opportunities and scaling the organization. Prior to jumping into tech, he spent five years in management consulting and two years in banking. He left that world to start his own company in programmatic advertising. Outside of work, he plays rugby for the New York Rugby Club and the New York Knights.

Diana Hsieh, Product Manager at Cockroach Labs, loves the collaborative relationships that naturally form between PMs and other departments because she gets to learn something new from her peers every day. She spends much of her time sitting with engineering, speaking to groups of customers, and catching up on technical papers. Before Cockroach, she spent two years at a venture capital firm doing diligence on enterprise software startups, but started her career as an investment banker. While she enjoyed the business side, she wanted to get more hands-on with technology. She joined Cockroach so she could play a part in building the future. She graduated with a degree in economics from MIT and is working on her master’s in computer science from Georgia Tech.

Don Richard, Product Manager at Verizon, is fascinated by opportunities for tech to fundamentally change how we learn, communicate, and level the playing field for the less fortunate. He currently works on Verizon’s Fios TV and internet service, building mobile commerce and platform products. Prior, he ran a mobile commerce and payments consultancy called Coin Labs, and led partnerships for customer communication startup LiveNinja. He has a degree in marketing from the University of Florida (where he also minored in entrepreneurship) and his JD from the University of Richmond School of Law. In his spare time, he hangs out with his wife, beagle and goldendoodle.

Ida Fey, Product Manager at AbleTo, builds technology that connects people suffering from serious medical conditions or life stressors with clinical therapy. She oversees the patient experience and is responsible for both the clinical and technological aspects of the product. Before working in healthcare, she hopped between marketing consulting and software. She graduated from the University of Michigan with a BFA in Art Therapy and Photography. Building products has allowed her to meld her experiences and focus on the human impact of technology design. Outside of work, she spends her time gallivanting through art museums, hanging out with her dog, traveling, and staring through a camera lens.

Isaiah Greene, Product Manager at Slack, is on the company’s Search, Learning and Intelligence team, where his focus is to help build an intelligent layer on top of the product to make life simpler, more pleasant and more productive for users. After graduating with his computer science degree from the University of Pennsylvania, he spent five years at Google — first as a software engineer and then as a visiting CS instructor at Morehouse College, and later as a PM. Switching to product management allowed him to combine multiple loves: building products that improve people’s lives using technology, problem solving, vision setting, exploration, communication, and ideation.

Jacob Effron, Product Manager at Flatiron Health, helps community oncologists improve how they run their practices. He’s passionate about using technology to increase healthcare access and quality, and enjoys wearing many hats day to day. Before Flatiron, he was a healthcare and economic development consultant at McKinsey. After one particularly exciting project on mobile-enabled business models, he moved to Tanzania to work with a solar startup selling small systems to off-grid households. He graduated from Yale with a BA in economics and history. Having grown up in Boston, he has serious love for clam chowder and all Boston sports teams. In NYC, he enjoys playing pickup basketball and singing karaoke.

Jisha Jacob, Product Manager at AppNexus, focuses on supply management and inventory quality products, working closely with engineering, data science and design. Before becoming a PM, she was on the company’s service team, first as a product specialist, then as an implementation consultant onboarding new clients, and finally as a solutions engineer. After graduating from Stanford, she worked on her own startup, The Active Generation (TAG), to excite and empower students to live healthier lives, and was selected for General Electric’s summer incubator. Today, she’s earning her master’s in CS and machine learning at Columbia, and spends her spare time writing young adult science fiction about a futuristic hacker school for kids.

Maddy Moelis, former Product Manager at Zola, has always been fascinated by consumer psychology and the power of empathizing with customers to solve real problems. This led her first into consumer insights and then product management. She’s currently working on a pre-launch startup in the consumer space. Before going out on her own, she led product at diversity recruiting platform Jopwell, was a PM at wedding company Zola, and managed the consumer insights program at Warby Parker. Originally from Westchester, she graduated from Wharton with a BS in marketing, and is an avid fan of the Giants, Knicks, Mets and Rangers.

Marley Spector, Product Manager at Dropbox, just started her new role on the Review team after spending three years at Hired, where she built products for job seekers and hiring managers. Previously, she was in the rotational analyst program at JP Morgan Chase. She studied engineering and economics at Swarthmore College, where she was a pitcher and outfielder on the softball team. When not working, she boxes, plays tennis, and loves finding great sushi restaurants in NYC and SF.

Rafael Balbi, Product Manager at LiveAuctioneers, helps thousands of sellers globalize their operations to sell one-of-a-kind items around the world. At the moment, he’s focused on seller tools and optimizing user funnels. In the last two years, he’s worked on ad units, a payment processor, fraud, and introducing a fresh take on live bidding. Prior, he was the first product hire at Maker’s Row, the marketplace for designers to produce goods domestically. He has his BS in economics from Syracuse University. He spends his time outside of work doing yoga, boxing and salsa dancing. His long-term goal is to lead development of inclusive products that’ll bring the world closer together.

Sean Echevarria, Senior Manager of Product and Research at Walmart, is on the People Experience team. He focuses on developing people products to address the entirety of the talent journey — from learning and applying to Walmart to onboarding and employee engagement across all brands and segments. Previously part of the Talent Experience team at Jet.com, Sean is a graduate from North Carolina State University where he received his BS in Mechanical Engineering and fell in love with tech through its Engineering Entrepreneurship Program. Outside of the office, you can find him hosting events through The UX Lab, a meetup he co-founded that now has 6,000 members across the U.S. — and The Product Lab, a series of talks from product leaders for up-and-coming product managers.

Victoria Kennedy, Product Manager at Hustle, works as head of the Platform team. Before this role, she worked at prestigious digital design agency Huge, focusing on consumer digital products. While there, she worked on the Today Show’s new app, a new learning platform for Girls Who Code, and an ecommerce site for an international fashion brand. Previously, she was a consultant at ThoughtWorks, and worked at both a gaming startup and ecommerce startup. Professionally, she’s most interested in civic technology. Outside of work, she’s a dancer, mainly West African, loves cooking, and looks forward to the day she can learn to garden.

Willie Maddox, Product Manager at Google, works on Google Classroom to improve learning experiences for students and teachers. Before this role, he worked at YouTube and at Microsoft improving entity-based search for Windows and Bing. He believes product is the optimal fit for him since he didn’t want to have to choose between design and engineering. When not at work, he’s usually sharing his opinions on better designed/engineered solutions in everyday life over cocktails. He also loves watching historical documentaries on PBS and generally tries to listen to as much music as possible while cycling and running around the city.

We were so thrilled to work with all 17 of these remarkable product minds. But the coolest part for First Round was seeing them forge bonds of their own, building off each other’s ideas in class, and talking for hours afterward over drinks. We believe community is a vital component to success in our industry, so we’re always looking for opportunities to forge connections that will bolster and educate people throughout their careers.

We plan to keep this going with a San Francisco Product Program launching later this year. We’ll let you know when applications are open, but keep your eyes peeled around summertime for an announcement. And if you have any questions in the meantime, send a note to the fine folks who design and run these programs Anita Hossain (anita@firstround.com) and Whitnie Low Narcisse (whitnie@firstround.com).

Many thanks to Maddy Moelis, Alex Powell, Isaiah Greene, Jacob Effron, Marley Spector, Jisha Jacob, Brittney Gwynn, and Diana Hsieh for compiling the notes for this post. You can read even more detail on the advice shared above on Medium here.

This type of cognitive load ...	is responsible for ...
Intrinsic	Actual learning of the material
Germane	Building that new information into a more permanent memory store
Extraneous	Everything else about the experience of encountering the material (e.g., who’s teaching it, how they teach, your comfort level with the material, what the room is like, the temperature, the season, your physical health, energy levels, and so on)

Tech2Pocket

SpaceX’s historic Falcon Heavy successfully launches | TechCrunch

Mozilla announces an open gateway for the internet of things | TechCrunch

Quizlet – The World’s Largest User Generated Learning Platform

What I Learned about Management and Culture from Growing Uber’s Asia Business from Zero to Billions | First Round Review

MANAGEMENT

Great managers stay involved in key decisions.

Keep close watch on yourself.

Leverage stories wherever possible.

Leaderboards are crazy effective.

One-on-ones need to be a two-way street.

Skip-level 1-on-1 meetings are a critical tool.

Make most meetings about decisions.

Make managers accountable (for the good and the bad).

Let long-term perception shape present decisions.

Evaluate managers with upward feedback and culture surveys.

Be there in the tough, dig-deep times.

Make a “management view” or “operational view” of the P&L accessible to all managers.

Align around communication between teams.

Don’t lose touch with HQ — and if you’re in HQ, get out to the field.

Fortify your cross-functional communication to anticipate major problems.

ORGANIZATIONAL EFFECTIVENESS

Every meeting must have a clear goal. Every decision must have a clear owner.

Big changes and launches require cross-functional awareness.

Cross-functional awareness should never require unanimity in decision making.

Create a decision-making framework and stick to it.

Reporting relationships work much better when they're local.

Never assume healthy information flow.

Choose a single point of contact by function.

Put every employee on the front line.

Get your management team together quarterly.

CULTURE

Know and respect your team’s sacred spaces.

Make vacation a cultural norm.

Proactively address your blind spots.

Transparency and consistency build trust.

Appreciate functions that are often overlooked.

Make it an operational discipline to do more with less.

AI Just Learned How to Boost the Brain's Memory | WIRED

Related Video

Neuroscientist Explains One Concept in 5 Levels of Difficulty

My Accessibility Journey: What I’ve Learned So Far · An A List Apart Article

Accessibility is exciting

1. Accessibility is a broad concept

2. There are permanent, temporary, and situational impairments

3. The first step is to make accessibility a requirement

4. Making accessible sites is a team sport

5. Communication is key

Highlight the benefits beyond accessibility

Challenge your colleagues

Help people empathize

6. Talk about accessibility before a projects kicks off

7. A solid knowledge of HTML solves a lot of problems

8. JavaScript is not the enemy, and sometimes JavaScript even improves accessibility

9. It’s a good time to know vanilla CSS and JavaScript

10. Keep learning about accessibility and share your knowledge

Share your knowledge

Go to conferences and meetups

Organize meetups

Invite experts to your meetup or conference

Follow accessibility experts on Twitter

11. Simply get started

Video series

Books

Blogs

Newsletters

Accessible JavaScript components

Resources and further reading

Introducing capsule networks - O'Reilly Media

So, what are CapsNets exactly?

The Supersonic Parachutes Carrying NASA's Martian Dreams

Related Video

The First Supersonic Flight Is Still Astounding 70 Years Later

Bob Mueller’s Investigation Is Larger—and Further Along—Than You Think | WIRED

Mueller Time

Related Video

Inside the Fake News Factory of Macedonia

Setting Up A New Phone – AVC

Ich Bin Ein Tweeter

Handling Hate Speech