Build Your Own Cognitive Computing Thingy


Have you ever been frustrated with the major tech companies' relentless efforts of seemingly invading your privacy? For example, Amazon's new phone and its app Firefly might be able to capture significant visual details about your whereabouts (VentureBeat), while Facebook's new feature for its smartphone app could be listening to what is going out around you (Forbes). And that's just the latest news. On the other hand, have you ever tried to envision what would be cool (valuable) for your smartphone to be capable of doing? Taking it a step further, if you would be responsible for building a service that improves its users' lives by providing the information or action that they need when they need it, how would you approach that task?

The sum of things that such service has to perform could be described by what IBM calls cognitive computing:

"Artificial intelligence meets business intelligence — Big Data growth is accelerating as more of the world's activity is expressed digitally. Not only is it increasing in volume, but also in speed, variety and uncertainty. Most data now comes in unstructured forms such as video, images, symbols and natural language - a new computing model is needed in order for businesses to process and make sense of it, and enhance and extend the expertise of humans. Rather than being programmed to anticipate every possible answer or action needed to perform a function or set of tasks, cognitive computing systems are trained using artificial intelligence (AI) and machine learning algorithms to sense, predict, infer and, in some ways, think."

Now, regardless of how you choose to call this thingy, as a Product Manager, you would need a way to meaningfully structure its functionality. And here, a framework that I put forth in April 2007 could provide valuable guidance. Based on a theory of needs that was most recently detailed in my book Spointra and the Secret of Business Success, the framework shows that you would need to develop three hierarchical core mechanisms (each could include devices, algorithms, as well as human interactions), which together deliver a value to its users — value points that, in aggregate, form a continuum.

The first mechanism (RANGE) gives the service and its users access to a particular pool of data. That's what Amazon and Facebook are primarily going for in the examples I mentioned at the beginning. The second mechanism (RELEVANCE) gives the service and its users access to a certain set of data that is relevant for the task at hand. A good example here is Google's search engine or algorithms. And the third mechanism (RECIPE) gives the service and its users a meaningful synthesis of the relevant data, whether it is materialized as information to be consumed or an action trigger for a particular device. An example of such mechanism is Wolfram|Alpha "computational knowledge engine," which for each information inquiry attempts to instantly assembles a unique response, displayed on a single page. However, as the framework shows, the service and its core mechanisms have to be continuously improved and periodically reinvented, or otherwise the value perceived by the bulk of its users will gradually decline.

For convenience, the original 2007 blog post is copied below, providing additional insights into this approach. Although my choice of terms to frame this type of service (information business, media business, or online business) might seem dated and narrow, the rationale stands. Over the years, many actions taken by various companies — for example, Google's universal search or its periodic search algorithm tweaks, or LinkedIn's Skills & Endorsements feature or its publishing platform — have been consistent with the elements and the dynamics described in the framework, providing evidence that it works.



Few business areas, if any, have been more affected by the Internet than the media. And by media I mean businesses whose activity consists of processing, managing, and distributing information (i.e., text, photo, video, music). This broader definition includes all of the traditional media and many web-based businesses. So, I thought that it would be appropriate to start my take on the business big picture by introducing an innovation model for the media business.

As you know, most media companies are businesses whose offerings are commoditized, or become commoditized very fast. And here is a way to prove this. Technology plays a major role in these companies' operations, which is a clear indication that a significant part of the operating processes are standardized. This occurs only when most of a company's offerings are delivered through the same, limited number of processes. Further, this means that these businesses thrive through the volume of transactions, as opposed to diversity. In other words, they sell or provide high volumes of only a few types of offerings, which indicates commoditized offerings (I will discuss commoditization in more detail in a future post).

Because, in media, it is rather impossible to innovate at the offering level, companies must innovate at the whole-offering level. And by whole-offering I mean all of a company's offerings plus the mechanism with which they are managed and delivered to the customers. (Although, in many cases, it is the main revenue generator, the advertising side of the media business is fully dependent on the information side of the business. More so, it addresses a different set of customers. Therefore, the whole-offering does not include advertising offerings; just information offerings.) Most media companies have just one whole-offering. In these cases, the whole-offering can be identified with the business. Some examples of whole-offerings include book distribution (, search engine (Google), encyclopedia (Wikipedia), video distribution (YouTube), and photo distribution (Flickr).

The logic behind the innovation model is relatively simple. On one side, there are the customers, who need information to better their existence. Fundamentally, humans follow three steps when employing information: search through the available information, gather relevant information, and then aggregate the relevant information according to a particular knowledge or algorithm in order to address an issue (i.e., entertainment, education). On the other side, there is the media business, which needs to process high quantities of a few types of information offerings, and then deliver them through one or more mechanisms (one type of offering may require a unique mechanism) to the customers. All I had to do was find a way to bring these two sides together. So, here's the result:

  • Whole-Offering = all of a media company's information offerings (i.e., text, photo, video, music) plus the mechanism (i.e., web-based application) used to process, manage, and distribute them to the customers.
  • Whole-Offering Functionality = a whole-offering's hierarchical levels of functionality (range >>> range+relevance >>> range+relevance+recipe), determined by the fundamental way in which customers employ information. Note: This hierarchy is a unique characteristic of the media business.
  • Range = the amount of information offerings that the whole-offering is capable of reaching.
  • Relevance = the actual relevance (to the customer) of the relevant information generated by the whole-offering.
  • Recipe = the degree to which the whole-offering's algorithm for aggregating the relevant information comes close to the algorithm that the customer would use in order to generate a solution to one of its issues (i.e., entertainment, education).
  • Whole-Offering Value = the level of importance of the customer issue that the whole-offering addresses.
  • Whole-Offering Innovation = an improvement of the whole-offering's value, achieved through the improvement of its functionality along the hierarchy mentioned above. Note: While small, incremental improvements may occur with a higher frequency, more significant improvements tend to move slowly up the functionality hierarchy following the advancement of the major technological era that unfolds at the time (see the Internet era examples below).
  • Operational Efficiency = a media company's capacity to generate and deliver its whole-offering(s) in the best possible manner with the least waste of resources. Note: While the lack of innovation endangers a business over time, the lack of operational efficiency poses an immediate threat. Therefore, operational efficiency must be a constant effort for any company (I will discuss this subject in more detail in a future post).

So, what is this model telling us?

Well, for primers, it explains the past. Of course, we can go back to various major technological eras like the printing press and the telegraph. However, as I mentioned above, it is useful to focus on one era and see how the innovation cycle gradually advances up the functionality hierarchy. So, if we look at the Internet-dominated recent past, it is clear that since its introduction, the Internet has rapidly increased the range of the (existent and new) whole-offerings to incredible levels. This explains the meteoric rise of companies like and Google.

The model also explains the present. Recent technological developments have allowed media businesses to significantly increase the relevance of their services (in addition to their range). The success of Wikipedia, MySpace, and other "web 2.0" whole-offerings/businesses is a result of these advancements.

And, finally, the model also indicates what will probably happen in the future. In this Internet era, the innovation cycle gradually advances to reach the recipe level. This suggests that the future is bright for the specialized communities, which can reach the recipe level with enough range and relevance that will ensure the critical mass of customers necessary for success. However, that's not to say that the successful whole-offerings of today will disappear. They will most likely find a way to thrive in the new environment. For example, Google will probably continue to excel at the range and relevance level. The acquisition of YouTube and other initiatives, like book-scanning, point in that direction. And, if nothing else, this can make Google a strong platform candidate for other whole-offerings. may run into difficulties because they sell mainly packaged information (i.e., books, CDs, DVDs), and customers most often want to access the pieces of information inside those packages. As a result, must, at a minimum, keep the pace with other companies' "un-packaging" initiatives (i.e., Google’s book scanning, Apple’s iTunes digital music store). Wikipedia too may have troubles advancing their whole-offering/business at the recipe level because of the heavy reliance on unpaid volunteers.


The article An Innovation Model for the Media Business was originally published on Cristian Mitreanu's Blog in April 2007. A different version of it (How to Innovate Your Online Venture) was published by MarketingProfs in May 2007. And a new graphic illustration for the framework (The Value of an Internet Business) was published on Cristian Mitreanu's Blog in December 2009.

Image 1: A slighly-refined version of the graphic illustration published in 2009 (The Value of an Internet Business). Late 2011.

Image 2: The illustration that appeared in the original blog post.


This article was first published here.

Cristian Mitreanu