Without question, the worst analytical results are those that are never acted upon. Imagine a company that invests significant capital (both human and financial) into an analytics project. Afterwards, the insights are ignored and never used to drive a decision. Why does waste like this occur regularly?
Often, the data scientists involved in the project did not properly communicate the results and benefits to the wider business audience.
People are hesitant to accept what they do not understand, especially if the results run counter to how the business operates; this reluctance can be compounded by a lack of knowledge of statistics and mathematics. It’s the data scientist’s responsibility to lower these barriers.
When I’m preparing to present analytics results to a client’s stakeholders, I remind myself that I’m not training them to be data scientists: They don’t need to understand the mathematical intricacies of the models; in fact, attempting to explain the nitty gritty details is frequently counterproductive to the goal of getting the business to act on the analytics.
Instead, it is much more important to translate statistics and mathematical models into clear recommendations that can be incorporated into the existing decision-making process. I never want to hear, “This analysis is fantastic, but what do we do with it?”
Does this mean that data scientists should be trusted implicitly? Absolutely not! It is the responsibility of the data scientist to explain the methodology and the level of confidence in the recommendations—but this explanation must be delivered at a level that’s appropriate for the audience.
There is a happy medium between “Trust me, I’m an expert” and quoting theorems from graduate-level courses. In my experience, a best practice is to keep the explanation at an elementary level, but to always be prepared to dive deeper if requested. This level of preparation may take extra time and effort, but better serves everyone in the long run.
Now that we’ve identified a major stumbling block, how do we overcome it?
Addressing the problem early and proactively is the best course of action. It is essential to involve all stakeholders at the beginning of any engagement; this enables the data scientists to properly address the business problem and translate results into actionable recommendations.
If stakeholders understand the nature of the recommendations from the beginning, future misunderstandings are much less likely. In my experience, clarifying assumptions and asking the right questions from the start go a long way toward ensuring that clients get results they can understand and use.