Skip to content

Delivering The Right Level Of Analytical Detail

After a long, difficult analytical project is completed, how you deliver and position those results is as critical to success as all of the work done to get to the results. I’ve written in the past about not saying too much and taking your final presentation seriously. Here I will focus on a simple concept to follow to successfully get your business sponsors to act on your findings.

Your Interests Are Likely Not Your Sponsors

Analytics and data science professionals love algorithms and data. That’s why we do what we do. However, most of the business sponsors of our projects really don’t care much for the technical details. They simply want to know what to do differently based on the results. Do not make the critical mistake of trying to educate a business sponsor on the ins and outs of the modeling approaches you chose and the theories behind them. The sponsor almost always wants to you to worry about those things for them, not for you to teach them about those things.

This is a mistake that many people who practice analytics and data science for a living make, especially early in their careers. I recall how eager I was when I was fresh out of graduate school with my new statistics degree to explain the statistical methods I used to anyone who would listen. Luckily, I noticed quickly how often and badly that approach backfired and forced myself to move away from it. It was clear to me that not everyone shared my enthusiasm about statistics. Why this wasn’t obvious from the start to myself, and why it isn’t obvious to many of my colleagues either, I cannot say!

Separating The Practical From The Theoretical

Let’s consider a simple example of a classic marketing response model where the goal is to predict which customers are most likely to respond to an offer. There are multiple approaches that can be used to generate such a model including decision trees, logistic regression models, and neural networks, among others. There are plenty of technical details and caveats for each of those approaches. There are also a lot of nuances in terms of how to best interpret and compare results from a technical perspective. Those details and nuances, while fascinating to an expert, are completely uninteresting to most business professionals who would sponsor such a project.

In the end, each of the modeling methods will come up with a probability that any given customer will respond. That’s what the business sponsor cares about and it is a very easy concept to understand. Don’t waste time explaining a bunch of technical details behind the models to someone who doesn’t care. Instead, focus on how easy it is to interpret and act upon the results. From a practical perspective, it is as easy as this:

  • Every customer will have a probability of response
  • The projected profitability of each customer based on that response prediction will be computed
  • A decision can then be reached about where to draw the line in terms of who gets the offer

For all the theoretical complexities in building the models, the business sponsor really doesn’t need to know much more than that. As long as you, and the other experts, validate that the models are working properly and with an appropriate level of accuracy, just focus the sponsor on how to interpret and use the results in the simple fashion outlined above. The same concept applies broadly whether predicting response, death, equipment maintenance issues, or myriad other things.

Enabling Model Usage, Not Educating About Model Creation

The takeaway here is that you should sharply, if not exclusively, focus on how to make use of analytical models in a practical, business context. In other words, focus on how to use the results rather than how the results were obtained. Even most teenage school children can understand the concept of a probability between 0 and 1. Most teens can also understand how to trade off the benefits of a response with the costs of reaching out to a customer since it is a simple subtraction. As complex as the analytics to get to those probabilities might be, putting the results to use is truly quite simple.

The secret to successfully delivering the right amount of analytical detail, then, is to deliver very little analytical detail at all! Put focus purely on the ease of decision making based upon the analytics and you can get business sponsors on board very quickly. Luckily, as the examples in this blog illustrate, how to make use of analytical results within a decision-making context is usually surprisingly straight forward and nonthreatening to even nontechnical business sponsors.

Originally published by the International Institute for Analytics