Analytics - centralized or distributed

Bella posted an interesting comment today - here - on my post about the opportunity for analytics in which she commented on Tom Davenport's POV in Competing on Analytics. Her comment about the pros and cons of centralizing analytics made me think that it was worth a post. She feels that Tom's proposal to centralize analytics is a classic consultant mistake (at least that's how I read it). Interestingly the same kind of "centralize analytics" message came through in the recent TDWI report.

It seems to me that a centralized analytics group is a consequence of widespread use of analytics more than it is a driver of it. Certainly most large organizations that use analytics in many areas tend to have a centralized analytics group. However, like Bella, I agree that most organizations are going to be better off focusing on developing analytic expertise in specific areas of their business. Not, perhaps, in a specific business process but certainly in support of a specific group of business processes. The data understanding, business know-how and focus on improvement necessary will all be easier to develop with a strong focus. The time, I think, to focus on a centralized analytic group is as you expand into multiple areas and start to really focus on a standard deployment infrastructure and toolset. The "E" in Enterprise Decision Management is not meant to imply that you must do everything enterprise-wide so much as that you must consider decisions an enterprise asset and decision management an enterprise skill.

Thought anyone?

Technorati Tags: analytical competitor, analytics, davenport, decisioning, EDM, Enterprise Decision Management, predictive analytics

Comments

[Alberto]

the way I look at it is that there is a centralized enterprise data mining tool, but that the analytics group should be decentralized. The advantage of a centralized enterprise data mining tool is economies of scale. The main issue with this method is that the enterprise must make a strategic investment to do this right and it will not be develop over night.

The advantage of having a decentralized analytics group is that it allows areas of specialization to flourish. The issue with this approach is that the enterprise must make an strategic investment in selecting and training the staff that will be using the enterprise data mining tool.