Centralize Analytics- A few weeks ago, I was attending a data science conference when I started talking to the person sitting next to me. He said he used to be part of a centralized data science team in a large insurance company but mentioned how that group had been broken up and dispersed into the various lines of businesses.
That got me thinking, because throughout my career, across several different companies, this was a dilemma I’ve wrestled with: Should we centralize the analytics function, or should we distribute it among the various business teams? Unfortunately, many organizations, including some companies I have worked at or consulted in the past, go from one extreme to the other and fail to capitalize on the tremendous business boost that can be obtained from analytics. The tension between centralization and decentralization is a debate that’s been taking place for years, and it is made harder by power struggles and other human dynamics.
In this article, I will share how I believe what works best is a hybrid analytics organization where some functions are centralized and others are distributed across business areas. The key is identifying the spaces where collaboration needs to be fostered. I have been fortunate to guide the formation of such hybrid organizations and see firsthand the tremendous business value they can unlock from data.
Let us first start by understanding the drivers for centralization and decentralization for analytics.
The drivers for centralizations are usually control, cost and conformance. Control stems from a real or perceived weakness related to regulatory compliance or the presence of a talented and strong-willed executive who is passionate about a specific direction for analytics. Conformance is usually an issue in organizations that are going up the analytics maturity curve. Initially, the need is to standardize data sources; this is followed by a desire to conform data definitions, metrics and processes. Cost is always a concern when organizations believe that they are not getting the appropriate business value from data and analytics. Centralization helps an organization more completely understand its total cost and then attempt to reduce these costs by exploiting synergies and reducing redundant work.