SPSS and the predictive enterprise

SPSS has been talking about the idea of the predictive enterprise for some time and, although they do not necessarily use the same terminology, so have a number of other suppliers in the business intelligence market. That is, they are talking about companies that do not just want to report on historic data but want to know what will happen if current trends continue and, based on that information, can take decisions appropriately. The predictive enterprise is slightly misleading because the emphasis is really on the decisions that companies can make as a result of predictions rather than the predictions per se.

Nevertheless, while this sounds good, few vendors have done more than spin a weave of words around existing products. However, exceptions are starting to appear. For example, OutlookSoft’s Predictive Planning in the corporate performance management space and, the subject of this article: Predictive Enterprise Services, from SPSS.

To someone who has historically done a lot of work in the application development space, Predictive Enterprise Services comes as something of a surprise. This is because it is essentially a repository where, in this case, you can check models in and out, version them, reuse them, deploy them into third-party applications, and so on. Which sounds an awful lot like what any self-respecting development environment has been doing for the best part of a decade.

Mea culpa but I have to admit that I had never even considered the possibility that these sorts of facilities were not already available for managing models. It just seemed so obvious that I assumed it. Moreover, it is not as if companies in this space do not have repositories—they certainly do—but at least according to SPSS (and I have no reason to doubt them) no-one else offers the sort of comprehensive facilities that Predictive Enterprise Services does.

There are a couple of additional points to make. The first is that Predictive Enterprise Services is open, so you can use it to manage data mining and predictive models based on third-party products and you can also use it to manage other relevant processes such as ETL (extract, transform and load) jobs. The second point is that, according to SPSS, Predictive Enterprise Services is nearly always implemented in environments where there is a large scale scoring requirement, where SPSS holds an advantage over at least some of its competitors because it can do all scoring in-situ and without having to extract data into a dataset.

So, why should data mining be so late in learning the benefits of model management in a repository? The short answer is that this technology has historically been marketed at relevant business analysts and user departments and there has been relatively little IT involvement. That has changed—there is an increasing demand for elements of data mining to be built into other applications and there is a much greater involvement with generic IT requirements—hence the demand for model management, as exemplified by Predictive Enterprise Services.

Predictive Enterprise Services is certainly a product that large-scale users of models and scoring should seriously look at and, given its support for non-SPSS products, this is true regardless of the product or tool that you currently use to build those models.