I'm a partner in the advanced analytics group at Bain & Company, the global management consulting firm. My primary focus is on marketing analytics (bio). I've been writing here (views my own) about marketing, technology, e-business, and analytics since 2003 (blog name explained).

Email or follow me:

« Privacy vs. Security Survey Interim Results #prism #analytics | Main | Please sponsor my 2013 NLG #autism ride: 2007 Ride Recap »

June 16, 2013

Organizing for #Analytics - Seven Considerations

We're now in the blood-sugar-crash phase of the Analytics / Big Data hype cycle, where the gap between promise and reality is greatest.  Presenting symptoms of the gap include complaints about alignment, access to data, capacity to act on data-driven insights, and talent.  This September 2012 HBR blog post by Paul Barth and Randy Bean of NewVantage Partners underscores this with some interesting data.

Executives' anxiety about this gap is also at its peak.  Many of them turn to organization as their prime lever for solving things. A question I get a lot is "How should we organize our analytic capabilities?"  Related ones include "How centralized should they be?", and "What should be on the business side, and what belongs in IT?"  

This post suggests a few criteria for helping to answer these questions.  But first, I'd like to offer a principle for tackling this generally:

Think organization last, not first.

A corollary to this might be, "Role is as role does."  Too much attention today is paid to developing and organizing for analytic capability.  Not enough attention is paid to defining and managing a portfolio of important business opportunities that leverage this capability.  In our work with clients, we focus on building capability through practice and results.  Our litmus test for whether we're making progress is a rule we call "3-2-1": In each quarter, the portfolio of business opportunities we're supporting with analytic efforts has to yield at least three "news you can use" insights, two experiments based on these insights, and one "scaling" of prior experiments to "production", with commensurate results.  (The specific goals we set for each of these varies of course from situation to situation, but the approach is the same.)

Approaching things this way has several benefits:

  • You frame "Analytics" and "Big Data" requirements in terms of what you need to solve specific challenges relevant to you, not in terms of a vendor's list of features;
  • You stay focused on the result, and not the input, so you don't invest past the point of diminishing returns;
  • By keeping cycles short and accountable to this rule, you hedge execution risk and maximize learning;
  • Your talent recruitment, development, and organization are done in the context of explicit opportunities, and thus stay flexible and integrated around concrete business results and not abstract concepts for what you need;
  • The results-oriented management of the capability helps build confidence that the overall ROI expected will be achieved.  Momentum is strategic.

Now, two critiques that can be made of this approach are, first, that it's too ad hoc and therefore misses opportunities to leverage experience beyond each individual opportunity addressed, and second, that it ignores that most people are "tribal" and that their behaviors are shaped accordingly.  So once you've got a decent portfolio assembled and you're managing it along, here are some organizational considerations you can apply to help decide where folks should "live":

  • For the business opportunities you're faced with, how unique is "local knowledge" -- that is, intimate knowledge of the specific market dynamics or operational mechanics that generate the data and shape the necessary analytics -- to each of them?  The more so, the more it will make sense to place your analysts in the groups responsible for those areas.
  • To what extent does the type of analysis you are pursuing require a certain degree of critical mass? It's hard for a single person or even small groups to manage and mine a Big Data capability, and if you sprinkle Big Data analysts throughout your firm to support different groups, you overwhelm each of them and under-serve the opportunity. Plus, each of them ends up with different Hammers Looking For Nails based on the particular tools and techniques they learn, rather than picking the best ones for different jobs.
  • How important is enterprise leverage to the business case for your capability?  If it is, centralizing your analysts so that purchasing efficiencies and idea sharing and reuse are maximized will be more important.
  • Are you concerned about objectivity?  When analysts get embedded deeply with business teams, there's a risk they can "go native", either because they fall in love with the solutions they're part of developing, or because of pressure, subtle and otherwise, to prove these solutions work.  This phenomenon is well-documented in scientific fields, even with peer review, so it's certainly more problematic in business.  
  • Are you, for whatever reason, having trouble keeping your analysts and their efforts aligned with your key priorities? For example, if one group needs to quickly get a product into market to grab its share of a high-growth opportunity, and then evolve it from there, and your analysts work in a group whose norms and objectives are more about "perfect" than "good enough", you may need to move folks, or get different folks in place.
  • How's your analyst-marketer relationship? If they're talking and working together productively, and the interpersonal karma is good, you can worry less about whether their boxes on the chart are closer or further apart.
  • Finally, which of these four "C's" describes the behavior you're trying to encourage: communication, coordination, collaboration, or control?  At the communication end of the spectrum, you just want folks to be aware of each other's efforts.  Coordination, for example, can mean "Hey, I'll be running my test Tuesday, so could you wait until Wednesday?"  Collaboration may require formal re-grouping, but it might only be temporary.  Control can be necessary for effective execution of complex projects.  The more analytic success relies on such control, rather than being satisfied by the "lesser" C's, the more you may solve for that with organization.

In our work we'll typically apply these criteria using scoresheets to evaluate either or both the specific business challenges we're solving for or the organizational models we're evaluating as possible options.  Sometimes we just use "high-medium-low" assessments, and other times we'll do the math to help us stay objective about different ways to go.  The main things are to keep attention to organization in balance with attention to progress, and to keep discussions about organization focused on the needs of the business, rather than allowing them to devolve into proxy battles for executive power and influence.


TrackBack URL for this entry:

Listed below are links to weblogs that reference Organizing for #Analytics - Seven Considerations:


The comments to this entry are closed.


Books by
Cesar Brea