Analytics Centralization: An Efficient Structure

Comments (4)
  1. ateeq ahmad says:

    I agree completely with the author. I also would add that as Raghu Kashyap mentioned(thanks for using the hub and spoke nomenclature. I use it all the time), a central core feeding all the functions and prioritizing every knowledge need is critical for efficiency. Other, we get little groups not knowing what the others are doing. I have seen several different numbers for the same metric coming from different perspectives.

    The one point of caution I would like to sound is that a different department will be treated as a cost center and so the organization will have little appetite for it in lean times.

  2. I have had a very interesting and fruitful experience with the org structure. I called it “Centralized Decentralization” . If you have heard of “Hub and Spoke model” it is very similar to that. Although our model worked really well in the Big Data world. You can check out some of the presentations that I have done in the past where I talk about this.


  3. Aurovrata says:

    A very interesting article and one that does raise a number of pertinent questions. I think the article makes a very good start at tackling the central question, however, from my experience of working in a large company such as Renault-Nissan I would like to add my grain of salt ;). I quote from the article:

    “More importantly, in today’s organizations, data is hardly centralized with various data streams from different applications and business units that the organizations have acquired. A department knows its data the best and where exactly it lies in the complex web of information that organizations create over time.”

    To me this is the crux of the issue. How does one reconcile the disparate activities of a large company (often with an IT department, a finance/investment department, and a core business section) with the obvious gains of having a centralised system?

    In my opinion it is not possible to have a single central analytics team. There is a need for decentralisation as specificity is often too high to be reconciled to a common team. Moreover, a central team, organising, sharing, communicating and providing training and common goals can be highly effective to not only allow all analytical teams to work with common KPIs/templates so that a common dashboard becomes more effective, but I think also the need to share vision and communication goals to guide the development of each analytical teams so as to ensure that data captured and analysed serves a common goal that ultimately serves the greater purpose of guiding the company on its central vision and mission.

    (Sorry for any spellos, my power is about to run out and the network shut down, no time to reach through this comment….)

    1. Bhasker Gupta Bhasker Gupta says:

      Hi Vrata. Very interesting thoughts. I agree, each system has its own set of benefits and challenges, and it’s per the best wisdom of a company’s management to choose the structure that best aligns with the goal of competitive advantage through analytics. I have seen companies running decentralized model very successfully for a longish period, especially the mid to small size organizations and especially in organizations where analytics need or data collection is pertinent to a specific line of business.

      Yet, I am fervent supporter of a centralized model, for reasons I mentioned in the article. One reason in favor of central model that I did not mention in the article is – in a decentralized structure, there is not much of a career path available to the analysts. In the wake of dearth of any meaningful opportunities, it is observed that analysts start looking out for opportunities in analytics either in different operational units or a different organization altogether. In either case it is fatal to the line of business. This issue is not so much prevalent in the centralized structure.

Leave a Reply

Over 100,000 people subscribe to our newsletter.

See stories of Analytics and AI in your inbox.