Software vendors can learn something from Eskimos

Now that master data management is gaining attention as an issue, it is interesting to observe the stances of the industry giants. As perhaps might be expected, each claims to have an all-encompassing solution to the issue (though they curiously had no product offering at all in this area two years ago, so presumably must count as quick learners) – all you have to do is adopt their middleware stack. Oracle have their DataHub, SAP have MDME or whatever it is called this week, IBM have an offering crafted by their acquisitions of DWL, Trigo and Ascential, Microsoft is well, Microsoft. All of them seem to be missing a key point. Intent on expanding their own infrastructure footprint at the expense of their rivals, they do not seem to grasp that large enterprise customers simply aren’t in a position to move wholesale to one middleware stack. Large global companies have SAP and IBM Websphere, and Oracle, and Microsoft and have a huge base of deployed applications using these, so any solutions which says: “just scrap the others and move to ours” is really not going to fly for the vast majority of customers.

By contrast what customers need is software that can, in a “stack neutral” way, deal with semantic inconsistency of their business definitions, whatever technology these definitions reside in. Surely by now it is clear after the billions of dollars spent on ERP that “just standardize” is simply a doomed approach for companies of any scale. Large companies can just about manage a common chart of accounts at the highest level, but soon as you drill down into the lower level details these definitions diverge, even in finance. In marketing (which by definition has to respond to local markets) manufacturing and others there is even less chance of coming up with more than a small subset of common high-level business definitions. Just as the Eskimos are said to have fifty-two words for snow, you;d be surprised at how many different ways large companies can describe a product, or even something apparently unambiguous like “gross margin” (26 different ways in one company I worked for). Hence you need technology that can help resolve the semantic differences here, and support the workflow required to allow maintenance. For example, DWL was strong at customer data integration at the detailed level, but many types of MDM problems require complex human interaction. For example a new international product hierarchy does not just get issued; there are versions of it, people need to review, modify and finally publish a golden copy. Most MDM tools today simply ignore this human interaction and workflow issue.

I think IBM have the best chance of figuring this out of the big four, simply because unlike SAP and Oracle they don’t have an applications business to defend, while Microsoft has never really figured out how to deal with the complex end of large enterprises. IBM’s acquisitions in this area may have been multiple but they have been shrewd. Ascential was the strongest technology of the ETL vendors, while DWL and especially Trigo were well respected. Ironically, IBM may need yet another leg to their strategy, since they too have yet to really address the “semantic integration” problem that is at the heart of MDM.