ActiveInsight

engaging in context - why Only 13% of organizations in the business technology sector find direct mail to be a very effective marketing tactic...

2011-07-12T13:39:00.000-07:00

Going back to a recent survey by Sherpa regarding good old direct email campaigns. (see here) I've had a few observations. The first figures that catches your eyeis that the majority of 57% find direct email campaigns only, and I quote: "Somewhat effective". this is understandable since we all used this technique and we know that indeed it sometime works but may not be worth the effort and may have some negative effects that cant be measured regarding spamming your prospects with a constant flow of emails (that tend to repeat themselves after a while). 21% found direct email campaigns to be "Not effective" and 22% found them as "Very effective".

The most interesting piece of information to me however was this:"Only 13% of organizations in the business technology sector find direct mail to be a very effective marketing tactic". Sherpa's Jen Doyle explains that this might be due to the fact that these companies sell complex products where emails may be less effective or relevant.

Now, we (activeinsight) have quite a few customers in the B2B hi-tech sector so I'd like to add another explanation based on the same logic: Yes, these companies usually sell complex products with multiple features and these products require the prospect's attention. If the prospect receives an email out of nowhere while he or she is doing something else the chances of them diving into the details is very small. Its like talking to someone who isn't really listening, your trying to catching them when they are "Out of context".

If your solution is more then a bit complex, you need your prospects to be in context and in focus, with all of their attention on their complex problem and how your product might solve it.

Most emails sent via marketing automation software urge prospects to read a relevant piece of content or visit a landing page. Why not offer it on your web-site, to specific prospects based on their digital body language. This way you offer specific content to a segmented audience based on the prospect intent and potential. you capture your prospects attention at the right time, when they are in context and tuned to your message.

Optimizing Online Campaigns

2011-05-27T09:05:00.000-07:00

Lately, I’ve been seeing a lot of B2B companies putting time and effort into product or sector specific online campaigns. But while hopes were usually high, results were not always as positive.

As organizations strive to focus their marketing efforts on specific and measurable goals they often launch focused online campaigns aimed at a specific product region or market sector. These campaigns often use a combination of an email campaign, SEO and PPC ads focused on driving relevant traffic into a specific mini-site. These mini-sites are in turn built with a predefined conversion goal such as a trial version download or setting an on-site technical assessment .

Dedicated online campaigns often fall short of their expectations due to two main problems:
1. Online prospects may not always follow the “ideal” mini-site flow or funnel
2. Later stages of the campaign’s funnel, usually taking place on the main website are neglected or overlooked.

Lets take, for example, a semi-fictional software company that wishes to raise adoption of their software development products.The company, Star SW, has defined a full-featured, free, 1-user version of its product and is launching a focused campaign aiming at early stage programmer. Star SW believes that students and young programmers would like using their product and hopefully raise adoption for Star SW products with current or future employers.

After running the campaign for 3-weeks, Star SW noticed 2 problems: the first is low conversion rates on the mini-site, although clicks per visit were relatively high. the second problem was that only a few of the programmers that downloaded the software actually went through with the first few product examples necessary to get started and acquainted with the software.

Using behavioral targeting and real-time reactions in dedicated campaigns and mini-sites can be extremely effective as it can automatically detect visitors that are not following the desired path and funnel them back on track. behavioral targeting can also be used to strengthen the campaign’s message on the company’s main website and followup on relevant steps taken after the preliminary goal was accomplished.

Lets briefly review how StarSW could have used behavioral targeting to improve the results of their campaign.

Optimizing Mini Site Conversions
After reviewing visitor click-streams to the mini site, Star SW’s Online Marketing Manager, Matt, realized that while visitors were expected to download the product after 3-4 clicks on the mini site, many of them browsed more then 7 pages and many of them visited the mini-site twice. Matt realized that these visitors were hesitating before downloading and were looking for more information and education regarding the product. Using ActiveInsight, Mat defined 3 real-time reactions proposing additional educational material to hesitating visitors. Since Mat didn’t want to interfere with the flow of visitors that do download the product after 2-4 clicks, as expected, he set the “educational” reactions to appear only after the 5th click or the second visit. He further added regional targeting to offer additional educational material in specific languages. So, Matt leveraged behavioral targeting to improve both top-funnel and mid-funnel conversions on his mini site.

Improving Campaign Efficiency
Matt also recognized another problematic area in the funnel. After users downloaded the software they were redirected to technical resource pages on Star SW’s technical WIKI. The problem was that the existing WIKI was somewhat overwhelming for newcomers as it provides a wealth of information for Star SW’s various products and is not focused on the 1-user edition. While Mat thought its important for new users to recognize the rich technical information available for them it is best that they initially receive a few simpler resources, mainly instructional “quick start” videos that have been previously prepared and uploaded to youtube .
Mat used ActiveInsight to detect visitors that were redirected to the WIKI from the mini-site download page or from the links in the actual product help module and present them with the “quick start” instructional videos aimed to ease their initial usage of the product. Making the WIKI adaptive to different types of visitors made it more efficient and supported the goals of that specific campaign without effecting every day usage by existing customers.

To summarize, using behavioral targeting to boost conversion rates can be especially easy in focused online campaigns since targets are clear and the effect of improvements based on behavioral reactions is substantial.

Accelerating Marketing Automation

2011-01-27T09:04:00.000-08:00

Marketing Automation software has become very popular with B2B organizations. Most of these organizations use marketing automation software to nurture and (hopefully) generate leads from their online traffic. Since Marketing automation tools focus their engagement and nurturing efforts on email campaigns, their effectiveness is limited and while they cost quite a bit and require substantial implementation efforts, they still miss out on a significant portion of relevant prospects. The two main obstacles in the effectiveness of Marketing Automation tools are:
1. Engaging and converting top-funnel, anonymous prospects.
2. Interacting and engaging prospects at the right time, while they are still on-site and focused on you.

Why do these obstacles exist?
Well, to start with, If you are a B2B organization, most of your online prospects are anonymous visitors or what’s referred to as top funnel prospects. These prospects do not identify themselves and you have no previous knowledge about their interests or potential. All you have is their online behavior and demographics. Real-time behavior and demographics can actually provide you with a wealth of knowledge regarding prospect interests and potential but can only be useful if you can react to these prospects automatically and in real-time – that’s right, while they are still in your website . Once top-funnel anonymous prospects leave your web site you have no practical way of reaching them. You lost your prospects.

The second obstacle is engaging prospects with the right message and at the right time. In many cases you might have the prospect’s email address in his cookie, generated during previous visits (after downloading a white paper or registering for an event). While this case seems to be the classic scenario for marketing automation tools (email + previous knowledge + current behavior = email response) it has several shortcomings. The main one being that you are trying to engage your prospects while they’re already doing something else. They lost focus, and are doing something else. You didn’t deliver the relevant message (be it a targeted marketing message or a sales engagement offer) when they were actually on your website but rather a few minutes, hours or days after that, when your prospect has already browsed your competitors site and is probably busy with other tasks. You had their attention and you lost it. You knew what you wanted to tell them (a relevant use-case, latest product news, invitation to a local conference, etc.) but you couldn’t.

That’s why behavioral targeting and real-time reactions are so effective. It provides the ability to target and engage top-funnel, anonymous prospects based on current interests and potential and the ability to use current and previous knowledge to auto-engage mid-funnel prospects at the right time, when they are focused on their problem and reviewing your solution.
ActiveInsight actually complements Marketing Automation tools by providing real-time behavioral targeting for targeted marketing, sales intelligence and qualified lead generation. It can react to specific prospects whose interests and perceived potential are high and display specific product/region/campaign specific messages to improve conversions.

That’s why Real-time Behavioral Targeting is your Marketing Automation Accelerator.

For more information about Behavioral Targeting for B2B visit http://www.activeinsightsaas.com

Don't just analyze, React. Turning behavioral insights into action

2010-10-20T07:03:00.000-07:00

One of the most amazing features of the human brain is its ability to recieve and analyze large amounts of data and react accordingly (ex. fight or flight). As analytical information systems evolved over time it seems that emphasis has always been on the ablity to process and analyze data rather then react to it. The result is a lot of information about what happened and even why it happened but little ability to effect and improve what is actually happening now.

So we always have explenations and excuses and perhaps ideas for improvements but we lack the ability to effect the present in order to change the future outcome.

Well, as you may have guessed, I think this should change. I think we should focus on what our applications can do with data rather then just displaying it in various manners. Dont get me wrong, the analysis offered by tools in the BI, web analytics and fraud detection fields is important but as a step towards action, not as a goal of its own. We should focus on predictive and reactive analysis to leverage the value of the data and the events around us.

I belive in "Turning Insight into Action" - and that is the heart of thinking Reactive. Nowadays, when there is so much information flying around, changing in an ever growing pace we should all think of how to turn this information and it analysis to actions that would help us achieve our goals whther they are selling more, securing better or improving service.

A report is really valuable if you have the means to turn its conclusions into actions while the relevant objective can still be reached. Here's an example: If I review daily reports of online trafic to my web site and can detect visitors location, organization and the pages they have viewed but cant interact with them (preferably automatically) to achieve my goals (get leads, sell, share...) based on these facts its not worth much since they are out of my web site by the time I read the report. They may or may not return and I might not recognize them if they do. Valuable data is waisted due to lack of ability to react on time.

Think Reactive! Dont waist your time on interesting insight that you cant turn into effective action.

Complex Event Processing for SIEM (Security Information and Event Management)

2010-04-15T05:09:00.001-07:00

As the security and regulations world evolved, it finnaly got its own BI application in the shape of SIEM products - specialized engines that can gather massive information loads, correlate security events and store all of this data efficiantly while still providing regulation driven retrieval abilities. While this market is considered quite stable with several giants like RSA, IBM, Cisco, CA and Arcsight leading the market I dont think the main problem has been solved yet.

SIEM applications work quite well as long as they get data from IT element that create clear and detailed logs, this is very easy to achieve when dealing with network elements like routers and switches, security elements such as firewalls and IDS's or even basic IT infrastructure like servers and storage devices BUT most APPLICATIONS, especially those developed in house or heavily configured since bought do not usually create detailed logs. That is why Application Level information and Events do not usually get into the SIEM application.

It is like guarding all the windows and doors of your house but paying no attention to your vault. Money is stolen from applications not from a router, internal fraud is done in applications not in the firewall and that should be the main guarding point or atleast one of them.

See Gartner aknowledges application level auditing and event detection as a critical and yet unsolved issue in their 2008 SIEM Magic Quadrent (available here), I think its the second line in the document, the probles is that wide-spread.

Only when SIEM applications recieve the relevant application data and events will they truely be able to detect internal frauds by correlating application level events with other, peripheral data and events (employee John Smith usually connects to the network from port X is now connected through the external FW and downlaoding sensitive customer reports).

So whats the problem? Why arent SIEM products taking care of this? well, because its a complex issue that differs with each customer: the standard approach would be to use existing logs, only they are not very detailed. Configuring applications to create detailed logs (that include specific parameters and relevant logic like action type, user name, transaction type and amount) is not always easy and has a huge toll on application performance as each server will now write this data to disk resulting in high I/O loads and storage costs. The mechanisms to gather all of these logs and filter the relevant information will result in another, maintenance heavy, mechanism which will not be evan close to real-time.

Other problems arise when application data is not clearly paramaterized (ex.the parameters have no names but rather codes) and must be matched and prepared before feeding the SIEM application.

The only solution to such issues are non-intrusive tools that can detect relevant application events, run basic filters and transformation and feed the SIEM applications as they expect. This still leaves the issue of mapping parameters and entities of different application but this is a one time effort that may rely on mapping tables to make it manageable. Such tools can offers real-time, event driven feeding for SIEM and offload this work from the application server creating one, central management point for application event routing. Thats waht we provide inActiveInsight.

Remember Jerome ? This case as well as others demonstrate the importance of application level event detection and auditing and why SIEM applications must recieve real-time application events to truely maximize their potential for internal fraud detection and support regulatory compliance to the fullest. See here for more details.

Why Real-time Reactions Make the Difference

2010-04-15T05:08:00.001-07:00

Here's a true story: A couple of years ago, I used to borrow DVD's from a local DVD machine, after borrowing about a dozen I got a phone call from the local Block-Buster store that owns the machine, they said that they see i have been borrowing about a movie a week and offered me a 35 movie membership ticket that would save me 50%.Do you want to guess what happened next? Was I offended? Was I angry that someone saw what I was doing?
NO, I was quite happy and bought the membership card. Their offer was relevant, focused and on time !

One of the main changes that the online channel has brought along is a change of pace. Everything is at our finger-tips, we jump from one subject to another in a flash, from one web site to another in a click - our concentration span is narrowed. Heck... I will probably check my email 4 times and surf to 2-3 sites while writing this post.

This short concentration span and our accelerated pace of life requires immediate responses - when we surf to a site looking for info, service, products or whatever we expect to be served immediately. If you (the site owner) have something special to tell me or sell me, something relevant to ME (based on my behavior or history) you better have it ready and serve it hot. I won’t be here in 2 minutes and probably won’t remember what I wanted. You have to hit the iron while it’s still hot. By the way, this is what inbound marketing is all about - Its what you do when the customer is at the door.

These days, companies usually have some historical data, specific to users or more general in nature that helps them decide what to offer the customer or how important it is to provide him with VIP service. The big problem is - How do you leverage this information in time?

How do you make efficient use of what you know about me and what I am doing now?

The answer lies in real-time reactions. If you can react in real-time to who I am + my history with your company + my current behavior then you have a much better chance at selling me you product, improving the service I get etc.

If you take a moment and think about it, this idea of real-time reactions is not so innovative, we are quite used to it in the offline world. When you enter a shoe store, the salesman will detect your entrance followed by your behavior in the store. He will react with the relevant offer when it is relevant (hopefully :-). If he sees you are interested in running shoes and points to the current New-Balance sale he increases his chances in closing a deal. The "real" salesman can also detect which customers are behaving more potentially and react to them. This way he maximizes his resources.

But our friend, the shoe salesman, doesn’t usually have the privilege of a customer history database or high-potential behavioral patterns. Web sites, however, can be connected to CRM and BI systems, so the potential is there. The question is, can they use this information in time? And can they correlate it with what the user is doing right now?

There many applications that deal with customer history and define what should be offered to them and how they should be treated. Various campaign management and BI systems offer features such as Next Best Offer and Next Best Action. What they usually lack are two critical abilities to leverage those abilities in the online world: Detecting that a customer is in our web site to actually deliver the offer to him and detect his current behavioral pattern to verify whether that his behavior aligns with what we knew/thought or perhaps shows that the offer should be different. Thats what we had in mind when we developed ActiveInsight - providing real-time, value-based reactions to online behavior.

Just think of yourself and your own experience as a customer - how often do you get irrelevant offers in sms, Email or phone? How often do you get relevant offers too late? And what an impact is it when you get a relevant offer on time...

Complex Event Processing for Smart Grids

2010-04-15T05:06:00.000-07:00

I was unaware of the smart grid until several months ago when I began meeting potential partners for ActiveInsight and was introduced to this new concept. we provide real-time detection of patterns, an important function for the smart grid, so several companies approached. I must say I was amazed once I realized how our lives would change in a few years.

The Internet has changed our lives, this complex web connecting billions of people around the world has changed economic models, broken knowledge barriers and provided many challenges and success stories but soon, our beloved web will lie in the shadows of a much larger web - the Electrical smart Grid, a grid that will eventually connect all electrical and other utility devices to form a correlated mesh of networks that provide higher utilization of our natural resources.

The smart grid will deliver electricity from suppliers to consumers, and sometimes the other way around, using digital technology and logic to save electricity, reduce cost, increase reliability and cut pollution and natural resource waist. Essentially, in Smart Grid implementations, the electricity distribution network is connected to digital communications and data-gathering tools, by getting current and/or real-time information on electricity demand, problems and such operators can run the grid more efficiently. Consumption peaks can be detected and supplied or at least controlled, misuse can be detected based on deviation from normal patterns or based on known patterns of misuse. Every node on the grid will be able to act as a sensor and receive configuration commands. Think of all your electrical appliances at home, off your future electrical car, of the millions of solar collectors spread around the globe, of all the electrical transformators and other devices - all inter-connected, sending messages and receiving commands.

What all of this means is that all of the knowledge and experience from the IT and Internet world, especially in the areas of Networking, BI (Business Intelligence), Management Systems, Event Driven Architecture, Extreme Transaction Proccessing and many other, yet to be fully tested fields, like Statistical and Behavioral Detection and Reaction will be deployed in the smart grid, supporting the major back-bone of the world.

The major challenges after connecting the main nodes to a network would be handling the massive data amounts, making sense out of them and trying to react in time. These are challenges that the IT world has not always been able to answer in simpler and smaller deployments like the Internet or even large enterprises but will have to step up and answer.

Unlike the web, which has developed independently and became quite large and significant before commercial companies and others joined in, the Smart Grid is a reality being pushed forward by various bodies, both political (like President Obama and others) and commercial companies (see Cisco or IBM). It is also expected that the smart grid will leap forward, skipping several of the steps the Internet took, due to existing knowledge, the overwhelming motivations and perhaps the Internet itself that may be part of the Smart Grid as well.

I find this field extremely interesting and challenging and I am very happy to see that we are begining to take some responsibility for our worls, hopefully, just in time.

Event Stream Processing in the Cloud

2010-04-15T04:53:00.000-07:00

While preparing my presentation for the upcoming "World Summit of Cloud Computing" (wow, what a title..) I thought it might be a good idea to share some of the concepts of Event Stream Processing and its relation to cloud computing.

Generally speaking, Event Stream Processing means being able to receive events and deal with them, now this may sound quite simple BUT it becomes very challenging once event sources, event quantity and event frequency grow in number. ESP usually consists of one or more of the following functions: Event Processing, Event Correlation and Event Aggregation, some refer to the ladder as Complex Event Processing CEP, although CEP is a sub-case of ESP - ESP deals with event streams and sometimes you want to "look" at several events together, CEP.

So why is event stream processing suddenly become important? The answer is quite simple: The number of events and sources is growing very, very rapidly. Think about it, mobile phones, GPS devices, digital TV's, RFID, surveilance cameras, smart home devices, security devices, cars, sattelites, meters, stocks, news, wiki's and on and on and on... This is the digital ERA and everything around us is going "BIP". But while the instrumentation and digitalization of our world has been growing very fast the ability to filter large amounts of events a quickly and detect the valuable events and patterns is still a challenge.

This is the point where the cloud comes in, cloud computing can enable elastic, on-demand event stream processing that can at least help in dealing with the massive loads of events the will arrive in various peaks. I wont go into cloud computing itself there are about a billion posts and web sites on that (even I posted a sentimental post about the cloud that raines money) but I will address a few issues regarding running ESP on the cloud.

First of all, the challenges of correlating and aggregating events remain the same, a good ESP framework (like ours :-)) must be able to:

Process high-frequency event streams cross-correlate events
Cross-correlate these events - correlate different events from different sources to the same context and/or correlate the same event to more then one context. For example, an event sent by a car (CarId, Location, Speed, Direction) could be related to an insurance related context (is this car being stolen), a traffic safety context (is my son driving the car too fast) and a commercial context (sending a coupon to a restaurant 10km ahead)
Aggregate events to their relevant contexts
React to event patterns in the various contexts (i.e. send the coupon or update the insurance company or interact with the online customer)

Secondly, to truly harness the power of the cloud computing for ESP, you must be able to scale-out and not just up . Although the cloud may seem to have endless processing power, behind every cloud infrastructure vendor are the same physical servers we are familiar with which means you cant scale up with no limits AND scaling up wont always create the expected affect since one machine is still just one machine. So a cloud based ESP must be able to dynamically add (and remove) server nodes and allow these changes to be recognized by the event stream clients to make use of the added node's processing power.

In addition to the above usage of ESP in the cloud, cloud based ESP can also be used as an on-demand service or as an enhancement of SaaS application that supports various multi-tenancy issues, local regulations and value-added services. Ill adress these issues in a future post.

Hello World!

2010-04-15T04:36:00.000-07:00

This is the official blog for ActiveInsight, an open source framework for distributed complex event processing (CEP). We have two websites: our Wiki, at www.activeinsight.org and our commercial site at www.activeinsight.net .

I would like to start with a brief introduction to the ActiveInsight framework, preceeding posts will cover practical implementations of ActiveInsight, interesting CEP, ESP and distributed event processing issues as well as thoughts of our team members regarding market status, direction and future vision.

Active Insight is an open source platform for real-time, value-based detection and reaction to events and patterns. It offers a distributed (cloud ready) event processing runtime with an embedded pattern engine to support event aggregation and correlation. Active Insight simplifies the development of distributed event processing using the plain old java object (Pojo) approach where events and event processors are plain java objects wired by Spring dependency injection.

As the world becomes digital the amount and frequency of event streams produced by cell phones, web sites, GPS devices , utility devices, stock markets etc. grows dramatically and requires new types of processing.

Harnessing the cloud for event-pattern detection offers a robust yet flexible way to handle fluctuating event-stream peaks and multiple event sources as well as presenting some new challenges in multi-tenancy applications and services.

ActiveInsight can be integrated with various front and back-end applications such as various web servers, logging systems, CRM, BI, Statistical etc. See examples of event and pattern detection power integrated with Tomcat, Syslog, R-server, ESB and others.