Divisional BI, Not Enterprise BI

I concur, and I’ll respond to his post, but the response is a bit lengthy so I’m putting it here.

Anil Chitkara coined the term “Divisional BI”, to describe the way many of our customers have been successfully implementing BI.

I think many of the reasons for corporate centralized BI are rationalizations driven by high-costs. BI and the DW bases and projects that support it seem to be the ideal candidates for centralization – they require somewhat rare skills to do, and require expensive hardware and software.

Of course those are the same reasons why one finds specialists outside to do this sort of thing. But I digress…

We centralize things if we can’t afford to do them in a less centralized manner. So, thought experiment: what if BI cost much less? Then I think the natural size of business unit where BI would be deployed would be closer to divisions than corporate level.

Imagine you work for a company with 3 primary business units. One is a manufacturing arm, one is a financial services arm, one is a broadcaster. What could these possibly have in common where common BI infrastructure would make any sense? Clearly in an organization of this scope, BI should be pushed much lower in the organizational hierarchy. But what is the right scope?

One idea is to look at where the sub-divisional structure share lots of business characteristics – where a common set of BI analytics would work for multiple of the subdivisions. E.g., a transporation utilization BI solution would work for multiple divisions if they all ship goods by truck, for example. All divisions want to utilize trucking services efficiently.

These sub-divisions aren’t necessarily small. These can still be global businesses. The key is the common ground drives the place where the BI deployment makes sense.

All this requires that BI come down in price substantially. Otherwise businesses will continue to centralize it, and wonder why their business divisions are so underserved.

SaaS BI is the highest-leverage way to do BI, hence, most cost-effective, and so drives this more “divisional BI” trend.