3 Things Are Holding Back Your Analytics, and Technology Isn’t One of Them

Many firms continue to struggle with business analytics. This has nothing to do with technology. We’ve found three main obstacles to realizing analytics’ full value: the organization’s structure, culture and approach to problem solving. Structurally, analytics departments can range between two opposite but equally challenging extremes.

On the one hand are data science groups that are too independent of the business. These tend to produce impressive and complex models that prove few actionable insights. On the other hand, analysts who are too deeply embedded in business functions tend to be biased toward the status quo or leadership’s thinking. Culturally, organizations that are too data-driven (yes, they exist) will blindly follow the implications of flawed models even if they defy common sense or run counter to business goals.

Alternatively, organizations that rely too heavily on gut instinct resist adjusting their assumptions even when the data clearly indicates that those assumptions are wrong. The dichotomy continues when it comes to methodology. At one extreme, we see analytics groups that create overly complex models with long lead times and limited adaptability to changing inputs. On the other side of the coin, some teams create models that are too simplistic and fail to capture the nuances of the problems they’re trying to solve.

During the past decade, business analytics platforms have evolved from supporting IT and finance functions to enabling business users across the enterprise. But many firms find themselves struggling to take advantage of its promise. We’ve found three main obstacles to realizing analytics’ full value, and all of them are related to people, not technology: the organization’s structure, culture, and approach to problem solving.

Structurally, analytics departments can range between two opposite but equally challenging extremes. On the one hand are data science groups that are too independent of the business. These tend to produce impressive and complex models that prove few actionable insights.

Consider the experience of one retail financial services firm. There, the analytics function was comprised of employees who used specialized software packages exclusively and specified complicated functional forms whenever possible. At the same time, the group eschewed traditional business norms such as checking in with clients, presenting results graphically, explaining analytic results in the context of the business, and connecting complex findings to conventional wisdom. The result was an isolated department that business partners viewed as unresponsive, unreliable, and not to be trusted with critical initiatives.

On the other hand, analysts who are too deeply embedded in business functions tend to be biased toward the status quo or leadership’s thinking. At a leading rental car agency, for instance, we watched fleet team analysts present intelligence purportedly showing that the fleet should skew toward newer cars. Lower maintenance costs more than compensated for the higher depreciation costs, they said. This aligned with the fleet vice president’s preference for a younger fleet.

But it turned out that the analysts had selected a biased sample of older cars with higher-than-average maintenance costs among cars of the same age. An analysis of an unbiased sample (or the entire population) would have yielded a different result. (Of course there might have been other motivations to keep a younger fleet — customer satisfaction and brand perception, to name two — but cost reduction was not one of them.)

Culturally, organizations that are too data-driven (yes, they exist) will blindly follow the implications of flawed models even if they defy common sense or run counter to business goals. That’s what happened at a financial services firm where management was mulling a change to its commission structure.

Posted on 7wData.be.