Anyone looking for things to read on analytics these days certainly has a lot to chose from. There seems to be no end to books and articles that explain or explore this topic. I am going to wade into this crowded space with my own view on a particular aspect of this issue, namely, how analytics-bases companies actually work, and, more importantly, create value in settings as diverse as services, smartphone apps, or even medicine. This approach is one I have developed after many years of building, delivering and now investing in analytics-based service business, and I think it can help those not very familiar with this space better understand how analytics-driven companies really work. I call my framework the CIA Model of Analytics for two reasons. First, because the idea for this model came to me while reading about how intelligence services work but also, and more importantly, because in my model CIA stands for the following elements:
- Calculation: Mathematical operations on captured or produced data
- Interpretation: Machine or human insights and recommendations based on analysis of calculation outputs
- Action: Machine or human activity to implement or enable interpretation outputs
To begin, consider the ubiquitous car service, Uber. When you open your smartphone and tap on the Uber app three distinct and yet highly integrated things happen:
- The app connects to Uber’s computing network, which contains engines that conduct a series of calculations involving your location, destination, trip distance, etc. Once those calculations are complete, the Uber app gives you a pick up time and a fare.
- You then look at the time to driver arrival, fare, vehicle type, etc., and decide whether to accept the conditions of the Uber offer and, if you do, call the car.
- The drivers brings his vehicle to you and takes you to your destination.
Now, consider another example: shopping on Amazon.com.
- You type a short product description into Amazon’s search engine, which sets the Amazon network to work, calculating its best prediction of what you want.
- You examine the search results, weighing price and features until you, Amazon hopes, find the right product/price combination and make a purchase.
- Once the purchase is made, Amazon goes to work packaging, shipping and delivering your order.
There are many more examples I could give, but the pattern in all will be the same. In every case, the analytics-based business is following the same three-step model: calculations are made, someone or something interprets the results of the calculation, and then an action (or in some cases a re-action) occurs. The examples above are consumer-focused, but the same model applies in the enterprise world. A good example of CIA model at work in the business to business setting is a company called Entercoms.
Entercoms delivers service supply chain analytics-based services to a variety of global clients. For example, for one global manufacturer, it provides an analytics service that determines how many spare parts the client must buy and store in order to keep its worldwide service operations running effectively and efficiently. To deliver this service, Entercoms executes the following activities:
- A set of mathematical engines conducts a series of calculations of various data sets, such as parts demand, failure rates, machine usage, etc.
- A team of analysts interprets the results provided by the calculation engines, filtering the results through their knowledge of changing business and environmental conditions as well as any specific information their client provides. Once this analysis is complete, the team provides a series of recommended actions to their client.
- The client, working with the analytics team, acts on the recommendations in order to meet the overall operational goals.
Again, I could provide other examples of this same model at work in the enterprise context from companies such as Nielsen, Dun & Bradstreet, Bloomberg, etc., and all would follow the same model. It’s interesting to note that this three-part model is also found in intelligence services such as the CIA and NSA. Indeed, when reading over the recent revelations about the NSA’s domestic surveillance programs, I noted the same three-step pattern: massive calculation engines pour over telecommunications data, which is then interpreted by NSA analysts, which in turns leads to specific actions taken by the U.S. government or directed agents. One also finds the CIA model in fields such as medicine, when, for example, images are created by scanning equipment running complex calculation engines, the images are interpreted by radiologists, and surgeons act to remove a growth or diseased part of the body.
Returning to the business perspective, the CIA model is the framework I use to analyze where technology investments are being made. For example, if one looks at a lot of the funding moving into companies focused on the industrial internet, one sees a lot of focus on the calculation part of the value chain, i.e., improving the ability to get and operate on machine-level data. Less investment is flowing into the ability to better interpret that data and even less into better ways to act on whatever insights those new calculations may provide. This is an interesting phenomenon, since in my experience (which is shared by many executives in the analytics services world) the biggest issue in getting value from analytics investments is not the “C” or “I” layers but the “A” — actually being able to act on what one learns from all the math. In fact, one of the more challenging aspects this model highlights is the “signal loss” that occurs across the three phases. For example, in some cases great work at the C layer is lost because of poor interpretive skills at the I layer. In other cases, good insights are never implemented because of time or budgetary constraints at the A layer. Indeed, it’s not unusual to see losses in the 30-50% range from C to I and losses of 50-70% from I to A. This signal loss is a value loss as well, of course, which is why the CIA model can also have organizational implications. Indeed, Rahul Singh, co-founder of Entercoms believes that in growing an analytics company, it’s important to build your organization with the CIA model in mind:
The CIA model served as a good guide for building effective teams. People with strong ‘I’ skills were not necessarily the ones who were technically strong (the ‘C types’) but the ones who understood the broader environment under which operating decisions were made. The best ‘I’ staff are the ones most comfortable dealing with conflict in decision making.
Consequently, whether you are examining analytics services as management, client or investor, my experience suggests that understanding the role that each layer plays — and how effectively meaningful and useable information flows from C to I to A — is the best way to understand how value is being created or destroyed. Indeed, returning to our consumer examples, it’s interesting to ask how much of Uber’s valuation is in what part of the CIA model? At this state in its evolution, the highest value seems to be in the C part of Uber, but will that change over time? Similarly, in the case of Amazon, a great deal of its value was in C, but, as the company has evolved, better I (through user-rated reviews and better comparisons) seems to be the focus. Perhaps in the future, drone delivery will shift even more value to the A part of the business. Even more interesting is Google, which built its business on its I layer more than anything else, i.e., the ability to interpret relevance from its sophisticated algorithms, but now wants to move into the A world with inventions such as Google Glass and autonomous vehicles.
Overall, whether one is looking at a Google search engine, an Entercoms service team, an air traffic controller, or even a radiology department in a hospital, the CIA model provides an easy way to understand how analytics-based services create and deliver value. It’s a model I use constantly — not just in strategy but also in making organizational and technology investments decisions. For me, at least, it provides a quick and consistent way to analyze the fast-moving world of analytics.