We are living in a period where certain buzz words are repeated again and again. While there are always buzz words being repeated at any point in time, it seems that there are more buzz words active today than usual. Companies, from startups to well-known technology behemoths, are falling all over themselves to make sure that everyone knows they are a player in some of the spaces covered by the most popular buzz words.

The valuations being placed on many small companies built around buzz words today seem way out of whack. Companies that are bleeding money in a crowded field are getting not just one, but often several, rounds of funding. In addition, the valuations being placed on these companies during fundraising or when they are acquired looks outlandish by any measure except perhaps when compared to the heydays of the Internet boom.

So just what is this CBIMBDMLAAS tongue twister and why is it interesting? The new company will be focused on Cloud-Based In-Memory Big Data Machine Learning Analytics as a Service (CBIMBDMLAAS). I challenge readers to find another premise to build a business around that captures as many of the hot trends in the market today as that term does. Just being able to say that mouthful with a straight face is almost certainly worth a first round of funding in the low millions of dollars today as long as even a cursory business plan and light prototype is used to support it. I will have those soon (I promise), but I need your money first to develop the idea further.

If you hurry and contact me in the next seven days, you can put money in early while the company valuation is still just under $100 million. That will be a huge bargain in a few quarters when the valuation shoots up tenfold as the first versions of the product win a few customers! Of course, we’ll have to lose money on those initial customers even as we continue to spend heavily to build the products. That’s ok, however, because it will provide you with the opportunity to put in even more money to help us ride out the losses until the valuation gets even higher. Slots are limited and it is first come, first serve so act quickly. You’ll kick yourself for missing this opportunity later.

Just in case anyone is ready to pull out their checkbook, I need to make sure that you are aware that I am not actually starting a CBIMBDMLAAS company. However, I honestly think that if I did start one, I would get funding easily. There are certainly some amazing new possibilities that big data, the cloud, and analytics combine to provide us. However, we need the market to settle down and pursue those opportunities rationally. Buzz words don’t make a profit and having a plan on paper isn’t the same as having a product in hand.

I congratulate those who have been able to cash in on recent trends by starting and selling a buzzword focused company for big bucks. However, the current state of things can’t and won’t continue forever because history has proven to us that such things never do. Perhaps some of the current wave of startups will beat the buzzer and cash out, but I suspect that there will be many without a chair when the music stops. Additionally, companies that pay the exorbitant prices to provide a chair today may just realize their error in paying so much for so little.

With all that said, I still can’t help but wonder what would happen if I did start my CBIMBDMLAAS company…

Bill Franks is an IIA Faculty Member and the Chief Analytics Officer at Teradata Corporation.
Read more from Bill Franks on the IIA Faculty Blog.

Terms come in and out of vogue on a regular basis. In recent years, the use of the term Machine Learning has surged. What I struggle with is that many traditional data mining and statistical functions are being folded underneath the machine learning umbrella. There is no harm in this except that I don’t think that the general community understands that, in many cases, traditional algorithms are just getting a new label with a lot of hype and buzz appeal. Simply classifying algorithms in the machine learning category doesn’t mean that the algorithms have fundamentally changed in any way.

Many startup companies, particularly in the cloud, are touting machine learning capabilities. In some cases, the algorithms are hidden behind a user interface so that users may not know what is happening under the hood. Users may believe that a new capability or algorithm that is closer to artificial intelligence is being used. However, would those same users be excited if they knew that they are buying a very early and immature version of yet another tool to create a decision tree?

Perhaps I have an outdated view, but I have always thought of machine learning as being closer to artificial intelligence than data mining. I want a machine learning algorithm to adjust itself dynamically and learn how to apply new rules. This is distinct from an iterative algorithm like a k-means cluster analysis. It can be argued that a clustering algorithm “learns” after each pass and adjusts dynamically. However, the rules are set in advance and don’t change. Once the first iteration of a k-means process has begun, the final answer is set in stone even if we don’t know the answer yet. Everything that happens after starting the first iteration can be manually duplicated if desired. A k-means algorithm uses fixed rules and the algorithm never learns to do something differently.

Like k-means clustering, many algorithms being tagged with the machine learning label today are more iterative in nature than adaptive and learning in nature. I first came across the difference between artificial intelligence and a complex set of rules in high school. For a science fair project, I programmed my computer to play the game Isolation. Isolation is played on an 8 x 6 grid. Players move their piece to an open space and then punch out any space on the board. The idea is to get your opponent trapped on an island with no moves to make before you are trapped.

As I played the game, I realized that a strategy of choosing a space with a lot of options on the next two or three moves, as well as the next move, would usually beat moving to the space where the most options existed for only the next move. My computer program took advantage of this. The program identified every possible space it could move to. Then, the program determined for each of the spaces how many moves were possible on the next move beyond the current move. I believe the program examined the options on the third move as well. Whichever available space had the largest number of options across the next several moves was the one the computer would pick.

When I took my program to the science fair, it beat most people. Since many people hadn’t played the game, it wasn’t surprising to me because a moderately skilled player will beat a novice in most games. However, many people thought my computer was truly intelligent, especially since it even had three difficulty levels. The only difference between the difficulty levels was the probability that the computer would randomly select a space instead of picking the best space. While people perceived that there was a lot of intelligence behind the program, there really wasn’t.

The point is that with some simple, recurring rules I was able to create a program that could beat most people in a strategy game. However, the computer really wasn’t thinking or learning. It was simply following predetermined, iterative rules that I had provided. There is an old saying that any sufficiently sophisticated technology is indistinguishable from magic. I am beginning to wonder if any sufficiently complex rules-based algorithm is indistinguishable from true artificial intelligence or adaptive machine learning.

I have no issue if the market wants to label algorithms that are based on iterative rules as machine learning. I do wonder, however, how many people are just following the hype and do not understand that what they think is an algorithm that is learning and adapting is really just a set of complex rules.

Bill Franks is an IIA Faculty Member and the Chief Analytics Officer at Teradata Corporation.
Read more from Bill Franks on the IIA Faculty Blog.

How do you strike a balance between data sufficiency and trying to get the data as perfect as possible?

Tom: It’s important to understand what decision is being made with the analytics and how precise the answer has to be. Then align the precision of the data with the required accuracy of the decision. For example, customer profitability data that has known minor flaws can still be used for a high-level analysis to determine if one group of customers is more profitable than another.

Marc: As we discussed on the webinar, you can iterate in data preparation to bring it to the necessary level of accuracy. It also helps to have good conversations with business clients about data sufficiency and expectations.

Bob: Data needs to be at a threshold of completeness and quality to get started, but it’s often a fairly low threshold. It’s more pragmatic to improve data as you go to reach the point of sufficiency. Why is the burden always on the data preparers? The burden of proof should be on those who are demanding that the data be perfect. Reversing that burden can help move things along.

How do you best enable business leaders to champion analytics?

Tom: It has been my experience that senior leaders who champion the use of analytics really understand what analytics can and can’t do, and they appreciate the potential of analytics in their organizations. We brought in an outside expert to speak to our Board of Directors and executive leadership about how analytics could be used to differentiate our company in a competitive industry. As a starting point, I’d buy the business leader a copy of Competing on Analytics.

Marc: I think of this in terms of how to help the business leader be successful. An engaged business leader who has results in his/her organization is positioned to socialize the success stories and champion analytics.

Bob: Seconding both points, initial education plus personal success plus organizational momentum make for an enthusiastic champion.

When setting up a new analytics group, would you prefer to populate it with people with deep business knowledge and teach them analytics or with people who have deep analytics knowledge and teach them the business?

Tom: This is a great question. I think the right answer is a combination of people with deep business and data knowledge and experience in the organization and others who bring deep analytics skills, often from other organizations. The “right” mix of these two types of people depends on the organization. If the company has a highly complex operating structure and the data is decentralized and “dirty,” I would lean more towards the group with strong business and data knowledge. If the data is more pristine and the organization is less complex, I’d lean more in the direction of people with strong analytical skills.

Bob: On the one hand, given the talent shortage, many companies would like to get their hands on as much analyst talent as possible. On the other hand, research and experience from our colleagues at Talent Analytics suggest that for most everyday analytics applications (those that don’t need deep data scientist skills) you can get pretty far by starting with quantitatively oriented business problem solvers and teaching them some analytical tools and methods.

Can you go into more detail on agile methods for analytics projects, including the first steps?

Marc: Agile is about close collaboration and rapid deliverables. I would lay out these two things to consider before jumping in:

• Communication. Define a regular communication process about the project with participants and stakeholders, be it daily, semi-weekly or weekly, and a format that does not attempt to solve issues but only articulates them so they can be addressed by the right people.
• Cadence. Then figure out the cadence of delivery – do we commit to delivering some chunk of work every 2 weeks, or 3 weeks, etc. With the cadence defined, the analytics team can think through how to divide their work to make measurable progress within the delivery cycle. For example, should we work on data understanding in a two week cycle, then build a preliminary model incorporating key variables in the next? If the project tasks become too big, they have to be broken down work into meaningful chunks.

Bob: I’d just add that agile is about speed and flexibility, but it’s not a free-for-all. It’s also about the manageability that Marc describes. I call the goal “agile with discipline.”

Have you used “friendly competition” in an analytical project? I am part of a team which is considering the approach.

Bob: I suggest putting two teams on the same problem as an experiment if that approach is compatible with your culture. The best solution may draw from both teams’ work, and the competition itself may drive teams to consider more options and be more innovative.

Marc: I think you have to try a few of these and find out what works in a particular organization. Some of this gets back to culture and the types of activities that enable people to practice the art of the possible. If the approach proves valuable, it can help move teams and organizations to a more data-driven approach.

Tom: I don’t like the friendly competition approach because it can create more problems than it solves.  When your analysts start competing, they may try to make others look bad, which isn’t good for team building or morale. You also have duplication of effort which most organizations can’t afford. A slightly different flavor of this approach is to have an informal peer review process where analysts from different groups meet at the start of a project to discuss the pros and cons of different analytical approaches and methodologies. The analysts can state their case for their recommended approach, and then the team works together to reach consensus on the approach that will be used for the project. You get the benefits of a challenger model without the bad feelings and duplication of work.

Is Excel here to stay? If not, how do we move beyond it?

Marc: My view is that, like it or not, Excel is here to stay. Microsoft has enabled all kinds of analytic functions in Excel – chi-square, z-test, etc. – and I suspect they will only add more.

Tom: In my opinion, Excel is here to stay. Like any tool, Excel has the most value when the analysts are well trained in how to use it most effectively. We’ve also had some success training the recipients of the analysis to use Excel. You can use Excel as a delivery vehicle for the analytics; for example, provide the business users with a pivot table that enables them to drill down and explore the data further on their own.

Is analytics viewed as a core competency of your company?

Tom: I think analytics would be viewed as a core competency in parts of KeyBank, but not in other departments or lines of business. It’s been my experience that the analytics center of excellence needs to demonstrate value and show measurable results before analytics is viewed as a core competency.

Marc: At Deere we are moving in that direction. As someone responsible for analytics, I would like that to be the case. And when I put myself in a business role, I would say that the company increasingly views analytics as a core competency.

Who should be running the platform – the analytics team or IT?

Tom: The model that has worked best in organizations I’ve worked in has IT owning the analytics platform, but the analytics center of excellence is very involved in design and prioritization of projects to enhance the various data warehouses, marts, and tools used for analysis.

Marc: Ultimately IT should run the platform. I would suggest that anyone leading a team that uses more than desktop tools like Excel and SAS desktop should consider having a data engineer who can work with IT on platform issues. On my team we have an individual in a consulting role working between IT operations and our decision science and modeling staff. This person sits as a virtual participant on analytics teams in order to understand and communicate the application or model deployment needs. This person is a former administrator and programmer with analytical curiosity and passion for results who was looking for a new role.

Robert Morison is the Lead Faculty Member for Enterprise Research at IIA.  Read more from Bob in the IIA Faculty Blog.

It wasn’t too long ago that many people espoused the decline, if not death, of the SQL language and relational database technology in general. As a level set, remember that relational technology stores data into rows and columns and that the way to access relational data is through Structured Query Language (SQL). For a couple of years, there was a full frontal assault on relational approaches from the Hadoop and non-relational crowds. The overhead of placing data into pre-defined rows and columns was deemed too great, compared to storing data within a non-relational environment.

In non-relational environments, users are free to use a wide range of programming languages to analyze data in any format. The data is typically simply stored in files with no assumed format or relationships. This approach does have its merits, but it also has its limitations.

In case you hadn’t noticed, a huge flip-flop has occurred. Many of the same people and organizations that were recently dismissing the entire concept of relational environments and SQL are now racing to … wait for it … add SQL-style interfaces on top of non-relational platforms like Hadoop! Let’s first take a look at how the flip-flop came about and then discuss why it is a good thing.

One big and mistaken assumption in the case against relational technologies is that relational technologies are not flexible and can’t handle unexpected questions or poorly formatted data. Therefore, a non-relational platform is required to be nimble. It is important to distinguish between an inherent shortcoming of a relational system and a shortcoming in how that system is implemented. That distinction is critical to understanding the flip-flop.

It is true that many organizations, particularly the large ones, not only had a large number of relational systems in place, but also locked the systems down very tightly. It was in fact difficult for users to ask new questions or to gain access to enough computing resources. However, this was due to the policies laid on top of relational technology as opposed to the technology itself. It is entirely possible to load and query data in a relational environment that isn’t in 3rd normal form, that hasn’t been formally modeled, and that isn’t yet clean. I spent years doing this.

The concept of an analytic sandbox or discovery environment centers on freeing users from traditional IT-imposed access limits and allowing them to explore and experiment with data in a relational environment. Granted, not all types of data can be handled in a relational system, but most common business data sources can be.

Like any solution, relational approaches are very good for many problems and are not as good for others. The same can be said about non-relational environments. Analytic professionals like me have always used a mix of environments because it isn’t about one approach being better or worse, but about which fits a given problem best. To me, SQL is the new kid on the block because when I started out, SQL did not exist! Over time, I added SQL processing into the mix where it made sense. It ended up making sense a huge proportion of the time, but not all of the time.

Recently, some organizations have tried to do too much with non-relational platforms. In many cases, this has led to inefficient processes that take more time to create, manage, and process than standard SQL approaches. Luckily, most of those who were looking to put up SQL’s tombstone have come around to their error.

It is terrific for the industry that the flip-flop around relational technologies has occurred. Having a mix of capabilities is a good thing and it isn’t a zero-sum game where only one approach can win. Facebook realized that trying to implement SQL-style processing outside of an environment built for it was wasting time and money to reinvent something that already existed and worked just fine. As a result, Facebook added a large relational environment into its mix because certain types of processing just work better that way.

I’ll be participating in a virtual event March 27 called Data Discovery In Action. Feel free to register here at no cost. The focus of the event will be on how to combine various processing paradigms and analytic techniques to maximize the ability of your organization to discover and deploy new high impact analytics. There will be discussion of both relational and non-relational approaches, which is how it should be!

Many of us who have spent years developing advanced analytic processes were surprised to see relational technologies and SQL getting beat up so badly. It never made sense to kill SQL and I’ll forgive those who were misguided in their attempts to do so. After all, it can’t help but sting a little to have to pull an about face and execute a flip-flop like politicians are known to do. But, sometimes executing a flip-flop is the right thing to do.

Bill Franks is an IIA Faculty Member and the Chief Analytics Officer at Teradata Corporation.
Read more from Bill Franks on the IIA Faculty Blog.

The other major difference is the pace of change for the two organizations and their customers. AT&T went through a massive set of changes around the turn of the 20th century, as it was building out the telephone network and users were growing at an exponential rate. In the 21st century, however, it seems to be drifting. Its only fast-growth business is AT&T Wireless, which was spun out into a separate company in 2001.

Google, however, is sitting on a volcano. Its markets and products are growing at a dizzying pace. Google Analytics, a tool for monitoring website traffic and digital marketing activity, is a great illustration of the changes taking place in the marketing world, and of Google’s success at adapting to them.

Some figures for changes in the recent past that were described at the conference include:

• 2.7 billion Internet users—up from 1.5 billion 3 years ago ;
• 4 billion videos watched on YouTube every day, up from 2 billion three years ago;
• Mobile media consumption up 500% in 3 years;
• 300% increase in video ads in the past year;
• 200% increase in “programmatic buying” (algorithm-based matching of ads with websites and users) of digital ads in the past year.

No numbers can easily describe it, but there is also incredible change in the marketing technology environment. CRM, web analytics, campaign management, social media management, and database marketing are all merging and recombining. All of these software domains are trying to master multi-channel, multi-screen consumer activity.

Google, for its part, is certainly keeping up with the pace. They paid me to speak, but no speaking fee could make me say something about them I don’t believe. Over the past year the company has launched 70 sub-products or features within the Google Analytics family. Product managers have decided they might be moving too fast for customers, so they only announced 14 new capabilities at this session. Google Analytics has gone from being a free, but not terribly impressive, offering to being a strong competitor for Adobe and IBM in this market. I doubt that AT&T ever moved quite so quickly with a single product family.

Given the rapid changes in digital marketing, Google could do nothing less if it wants to stay on top of them. However, all of this rapid evolution is tough on customers and their marketing quants and technologists. Forget actually implementing anything—it’s a full-time job just keeping up with what’s available in the marketplace. I have a lot of respect for the web analytics professionals at the conference, although the change in their worlds is just beginning. There will be not only continued evolution in technology, but also in the organizational structures and capabilities required to pull off an integrated view of customers in this crazy era.

It’s an exciting time to be a marketer, and I’m sure it’s an exciting time to be at Google too. But even the management team of Google Analytics is surprised at the level of change. Paul Maret, the founder of Urchin—the company acquired by Google that became Google Analytics—and now its VP of Engineering, said that he can’t believe the pace of change. So imagine how the rest of the world feels!

Originally published in WSJ’s CIO Journal.
Read more from Tom Davenport on the IIA Faculty Blog.

I call myself “ringleader” because the research gave me the opportunity and pleasure of polling and working with a variety of IIA faculty to develop and synthesize the list of obstacles. But that was only half the fun. The other half was having the occasion to explore what’s different about analytics – what special twists do analytics initiatives introduce?

Some of the basic obstacles are common to business change initiatives generally. For example, the organizational resistance to new work methods can be especially formidable when the analytics change people’s decision-making prerogatives.

Other obstacles are common to efforts to build organizational capability. For example, the challenge of increasing analytical talent plays out against a backdrop of rising demand and shortage of supply across the board.

And other obstacles have much in common with information technology projects in general. These include the danger of front-loading effort on the preliminaries (typically data and tools) and thereby delaying business results and losing momentum for analytics.

As I step back from this exercise and look across the obstacles, three themes jump out for me:

1. Analytics projects are inherently experimental, and “agile” is the only way to conduct them. Get some good data, prototype a model, fold in more data, test and learn as you go, and iterate until you have a workable solution or declare a dead end. That approach is obvious to professional analysts but often not to the business people they work with, who may still be accustomed to hands-off specifying of what they want. But the whole team must be agile. Even if the business domain and its analytics are unfamiliar, the last thing you want to do is try to “tame” the initiative with conventional sequential project management.
2. No matter how much experience and ability your enterprise possesses, you can’t go it alone with analytics. So much is happening in the marketplace as analytics tools and techniques and applications multiply and mature. “Ecosystem” has become an overused buzzword, but it’s the right word here. An enterprise has to leverage the ecosystem of analytics services to keep its capabilities current and its talent sufficient. One of our obstacles is the failure of the “human network” – lack of connection among analysts both within an organization and into the marketplace and professional ecosystems.
3. Timing is tricky. As business demand for analytics grows, the supply of analytical capability – technical talent, business talent, data and technology platforms – must advance in coherent fashion. Not in fits and starts, which robs some projects of needed resources. And not overbuilt up front, which delays business results and analytical business maturity. Key foundational capabilities, such as formal model management, should be introduced right on time. The elusive ideal is for supply to be just a step or two in advance of anticipated demand. And to have an extra gear in reserve (e.g., an on-demand source of supplemental talent) in case business appetite and ambition suddenly spike.

For the record, our nine obstacles cover:

• Leadership
• Data
• Platform
• Projects
• Models
• Business Adoption
• Talent
• Human Network
• Ambition and Pace

For each, we briefly discuss the nature of the obstacle then list the warning signs on the one hand and prevention or remedy actions on the other. IIA’s ERS clients can access the report in the member library. If you’re interested in reading the report, but not yet a member, you can contact IIA here.

You can also learn more about ways in which you can navigate the challenges frequently posed in analytics by attending our upcoming webinar on March 4, entitled Succeeding with Analytics: Overcoming Common Obstacles. You sign up and join in by registering here.

Read more from Robert Morison in the IIA Faculty Blog.

• A heavy equipment manufacturer is making a big bet on sensors in its products, and is trying to determine what are the best uses of analytics on the sensor data to provide value to customers;
• A large bank has established a new organization to develop new products and services based on data, and to make analytical capabilities more broadly available throughout the bank;
• A highly-ranked healthcare provider is spending enough money on improving its data and analytics infrastructure that it plans to describe the initiative in some detail to its board of directors;
• At a home appliance manufacturer I visited about a year ago, no one seemed very interested in analytics; now a centralized analytics group has been established, and at the Symposium we discussed what the company could do with analytics from Internet-connected appliances;
• An insurance company is using experimental design approaches to test most if not all of the company’s new products before they go into wide release.

These seem like signs that analytics have become one of the most important arrows in any business’s quiver. If analytics and analytical people are intimately involved in new products and services, new strategies, new organizational structures, and new programs discussed by boards of directors, they have arguably achieved escape velocity.

There was another indication of analytical success at the Symposium. I led one session—there were a variety of them throughout the day, and the emphasis was on participation, not presentation—on how to get beyond reporting and descriptive analytics in organizations. I came prepared with a long list of things organizations could do to ease demand for reporting, and increase the demand for predictive and prescriptive analytics.

Fortunately I hadn’t gone too far into this list when I realized it was the wrong one. Instead of encouraging demand, I needed to be talking with these companies about limiting it, or at least getting more supply. Many people said their problem was an inability to meet the demand for “true” analytics. While several attendees said they still struggled with self-service reporting and other approaches to offloading that type of demand, their primary focus was how to prioritize projects and meet the needs of perfectly worthwhile business initiatives with an analytical component.

I left the meeting thinking that while we’ve mastered some of the “startup” challenges with analytics, we really have to deliver now. I got the feeling that the attendees got a lot of value from the discussions, which were directly addressed at their challenges and opportunities. Unlike so many meetings, the “experts” and the “faculty” didn’t dominate the dialogue at all. There are a lot of ways in which organizations can get value from IIA, but the Symposium is among those that I enjoy the most—especially when I am surprised by success!

Read more from Tom Davenport on the IIA Faculty Blog.

Watch the replay of our Feb. 19 webinar to learn:

• What is analytics maturity?
• Why is it important to measure maturity?
• How can you get a reliable assessment?