There are several factors that need to be considered when centralizing your company's analytics function including size of the organization, availability of analytics talent, the subject matter expertise of talent, and the organizational culture. The centralized model can be effective but not always as effective in some situations because some situations call for a more distributed structure. Based on the literature and the experiences of some organizations that have considered the centralization approach, we have put together some of the key findings that could help inform your company's future efforts in determining the right level of centralization or right structure of analytics function.
Below are some of the key factors and trade-offs that need to be considered as well as some recommendations for another possible approach because what organizations are learning is that this a decision that is very situational and that there is no one-size-fits-all solution.
When analysts are distributed and embedded in smaller divisions and silos, there is the possibility for the lack of coordination and collaboration resulting in redundant analyses and having to continually reinvent the wheel. In a centralized analytics function, like-minded analytics people are brought together which thus facilitates a cross-pollination of analytical techniques and knowledge. The duplication of analytics work is reduced and there is an improved and a more standardized operating protocol for the incoming data requests and for how the work will be done. However, in a centralized analytics function, because of the protocols in place and the standardized prioritization of projects, there tends to be a higher level of bureaucracy which could limit the agility of the analysts and the ability to do timely and flexible analyses.
In a decentralized analytics function, analysts embedded in divisions where they have subject matter expertise tend to have a deeper personal investment in the projects of the organization they work for. Because of their level of involvement, they tend to have a better sense for the nature of the project and the policies and they understand better how the work they do aligns with the overall strategy and purpose. There is the potential for analyses to be richer because analysts have a high degree of ownership and involvement. In a centralized analytics function, there tends to be a lower degree of ownership of the business area and analyses could potentially be rigid. Analysts in a centralized function also to have a consultant’s mindset where they take on the project, do the analytics work, and then move on to the next project.
And because making the right decision depends on the
situation, it may not be a matter of choosing one over the other but rather
considering a third alternative where there are embedded subject matter expert
analysts in the various units across the organizations and with an increased
effort on the focus on cross-disciplinary and cross-community
collaboration. For example, at Intuit, a company that realized the power of analytics across the different departments,
a cross-disciplinary approach seemed more reasonable than centralizing. What companies can consider is having a Center of Excellence in Data Analytics with an emphasis on cross-disciplinary collaboration among the embedded data analysts in their respective divisions where their expertise are needed. There can still be a central data analytics function but this time recognizing the distinct work that is being done in each division. Analysts can remain in their domain of expertise but with a renewed focus on collaborating with other data analysts across the division, allowing for the cross-pollination of knowledge and techniques.
And for HR analytics, this is an important consideration to make because HR analysis can be very different from the other analytics functions throughout the company. HR analysis is not exactly the same as web traffic analysis or marketing analytics. Does it make sense for HR analytics teams to be under the company's centralized analytics department? Stats is stats but it's not always as straightforward when you're dealing with job satisfaction, retention, and other human behaviors unique to HR's domain of expertise.
How companies decide what to do with those darned data scientists. http://venturebeat.com/2013/12/04/how-companies-decide-what-to-do-with-those-darned-data-scientists/
How to Organize Your Analytical Talent. http://www.analytics-magazine.org/january-february-2010/165-executive-decisions-how-to-organize-your-analytical-talent
To Centralize Analytics or Not, That is the Question http://www.forbes.com/sites/piyankajain/2013/02/15/to-centralize-analytics-or-not/
Which is the best model for Institutionalizing Analytics: Centralized, Decentralized or Federated? http://www.mu-sigma.com/analytics/blog/?p=164
comments powered by Disqus