I recently attended IBM’s analyst summit on business analytics. Since last year’s event was largely a preview of Cognos 10, which was several years in the making, I wondered what this year’s event would be about. IBM focused much of the attention on predictive analytics, strengthened by its acquisition of SPSS. My colleague Robert Kugel covered another theme from the event in his post on Cognos Planning.
I’ve noted recently that predictive analytics are on the rise, and based on its products and plans, apparently IBM agrees. However, our benchmark research on analytics shows only 13% of businesses using predictive analytics today, so clearly obstacles remain to widespread adoption. We’ll be investigating all of them as part of our 2011 research agenda, but I suspect one of the obstacles is training. Predictive analytics are not always easy to use. IBM has two major approaches to this issue. First, it has packaged some of these capabilities in its performance blueprints and plans to introduce more blueprints in the future. Second, IBM continues to enhance and refine the SPSS Decision Management product, which was designed specifically to make predictive analytics both easier to use by business users and repeatable as part of an analytic process. Decision Management includes wizard-like capabilities for these seven activities:
- connect to data
- define global selections
- define desired outcomes
- define operational decisions with rules and models
- optimize outcomes
As a way to simplify development of predictive analytics models, IBM’s refinement of Decision Management automates data preparation and evaluates several types of models to pick the best for the task at hand. If you are concerned, as I am, that automated model development is a bit like giving sharp knives to children, it is good to know that the models created in Decision Management can be reviewed and modified if necessary in SPSS Modeler and then returned to the creator. But while that may help, I’m still skeptical about automated model development and would like to see more controls and governance over the models that are created. As Robert Kugel mentioned, IBM is also pushing the notion of business processes. Decision Management supports automated processes with its own set of business rules capabilities as well as integration with the more sophisticated rules capabilities of IBM’s ILOG business rule management product.
IBM spokespeople described a range of experiences in working with customers in a variety of industries to deploy predictive analytics. They categorized the predictive analytics opportunities into three broad categories: customer analytics, operational analytics and threat and risk analytics. Our research cited above shows a greater prevalence of predictive analytics in customer contact centers and in supply chain operations than in other lines of business. Predictive customer analytics can help companies acquire, grow and maintain customers. One example IBM gave was a communications company doingsurvival analysis to understand and prevent customer churn. In operations, IBM described an application that uses real-time monitoring data for a fleet of automobiles to reduce repetitive repairs. And in threat and risk analytics, IBM has published a case study on how the police department in Richmond, Va., is using predictive analytics to reduce crime.
Perhaps IBM’s most sophisticated application of predictive analytics is Watson, the computer produced by IBM Research to compete on the Jeopardy quiz show. Unfortunately, you cannot purchase Watson as a product, although some of IBM’s marketing may lead you to believe otherwise. It cost tens of millions of dollars. IBM has not announced any plans to offer it as a product, but the technology behind Watson has given IBM insight into predictive analytics that could prove to be a differentiator. Or it could turn out to be the analytics equivalent of Maxim freeze-dried coffee. At one time, freeze-dried coffee was rocket science – derived from work at the U.S. space program – much as many people today consider predictive analytics rocket science. Kraft spent heavily to develop a market for freeze-dried coffee, and then with a much more modest investment Nestle introduced Taster’s Choice, riding on the coattails of Kraft’s hard work, and overtook Maxim within two years. Kraft eventually removed Maxim from the market, but Taster’s Choice survives to this day. I have no specific reason to believe predictive analytics effort will suffer the same fate at IBM, but there is a precedent. Witness the PC market that IBM helped develop.
In any case, IBM made a big investment in predictive analytics when it acquired SPSS, and it continues to further its investment as it brings these capabilities into the mainstream of its business analytics activities. The individual products such as Cognos, ILOG and Netezza still need further integration as part of a single unified IBM stack, but the company is pushing the envelope for predictive analytics. I don’t know if IBM is getting a return on its investment, but it is certainly advancing the market for predictive analytics; its efforts should lead to better business outcomes, which will be good for everyone.