Cognos, SPSS and Predictive Analytics

SPSS and IBM’s Cognos group recently made an announcement of a partnership to integrate their products. I don’t have any more details than the press release but take this announcement as an indication that IBM/Cognos reached the logical conclusion that Predictive Analytics is distinctly and uniquely different from Business Intelligence/Performance Management and that customers want/need it. It is an interesting partnership as it helps Cognos’ customers move from capturing and reporting data to analyzing and using it more effectively – to “extend the value of Predictive Analytics outcomes to decision-makers” as the press release puts it. I have sometimes referred to this as “predictive reporting” – bringing predictive analytic techniques to bear on the reporting/BI environment. Some describe predictive analytics as just a form of BI but I think there are serious differences between BI and predictive analytics. Anytime a major BI player acknowledges the power and value of predictive analytics I take heart but I also worry that this will further blur the distinction between BI, predictive analytics and Enterprise Decision Management or EDM.

The crucial distinction between BI and EDM is that the former helps you understand your business, the latter helps you execute business. If BI is the bridge between your data and your strategies, EDM is the bridge between your strategies and your customers or transactions. At the risk of simplifying two sophisticated disciplines, here are some other differences in approach between BI and EDM:

• Where BI solutions provide information access and insight on customers as groups, EDM uses customer-level insight to identify the ideal action to take in a particular transaction.
• Where BI is a back-room, offline operation controlled by knowledge workers, EDM is embedded in operational systems and processes. Some processes will call out to an EDM system for decisions during “live” transactions; other processes operate offline. In both circumstances, the decisioning must be tightly coupled with the operational system.
• Where BI analytics traditionally synthesize past performance, the analytics used in EDM frequently predict future behavior. Predictive analytics play a role in BI as well of course, but are more dominant in EDM systems, this announcement not wirthstanding.

What this partnership does do is show the truth of a 2006 PricewaterhouseCoopers survey reported in DM Review that essentially said that companies thought their data should give them a real competitive edge but found it actually did not. The prime reason for this was an inability to mine the data. This came ahead of consistency, accessibility, timeliness and accuracy even though, if you read the BI press, you might think that issues of consistency (single version of the truth) or accessibility (buy everyone a reporting tool) were critical! What was even more enlightening, however, was that deriving value required four key components:

1. Mining data, to produce insight presumably
2. Delivering data assets, or insight based on these data assets (given 1 above)
3. Integration with business processes
4. Real-time information

Clearly Cognos is attempting to address these issues through its partnership with SPSS.  The announcement also reminded me of a TDWI webinar by Wayne Eckerson on Predictive Analytics based on a survey TDWI conducted of its members (here’s the report’s executive summary and you can get the report here).  This showed a clear and growing interest in predictive analytics amongst BI/DW practitioners. The survey responses are exactly the kinds of things that I am sure drove the SPSS/Cognos partnership – a desire to move beyond reporting and into prediction in areas like marketing, forecasting, churn and fraud. The survey noted that the ROI of predictive analytics is much higher that for other BI applications, though it comes with a higher investment also.

I believe that BI/DW professionals need to look beyond predictive reporting and understand the power of rules and analytics in combination to automate 95-99% of operational decisions – EDM, in other words. When I talk about “predictive analytics” on this blog or when Neil and I discussed it in the book, we mean this – the creation of executable models that make a useful prediction about a transaction or customer. It is not clear how this partnership relates to the broader question of how IBM sees executable predictive analytic models but, as SPSS’ products support this kind of predictive analytics also, Cognos customers taking advantage of the SPSS partnership will be well positioned. In the end, EDM is about analytically-driven processes and while these include predictive analytics, predictive analytics can also be used in decision support systems and reports (I discussed the differences between decision support systems and EDM systems with Dan Power some time ago). A partnership between a major BI player and a major data mining/predictive analytic vendor can only be progress.