An executive will rarely come to you with a clear, straightforward deliverable that they are responsible for. Most of the time, they will approach you with an abstract business problem, general guidelines, and an ultimate goal. It is up to the data scientist—yes, you—to help them define the problem and choose relevant data sources. It is critical that communication with upper-level management begins from the very first interaction, rather than the final output. 

This point is lost on many analysts, especially us data scientists. Data scientists come from a diverse range of backgrounds and fields, from computer science to physics, and so there may be many approaches to the same problem. As for myself, I am an econometrician by training and now work at an AI-powered healthcare guidance startup. In my experience, I have found that a more formulated approach can help when communicating with upper-level management. Let’s consider the following scenario.

Your manager stops by to tell you there is a meeting with the VP of Marketing tomorrow to discuss one of their year-end goals. At the meeting, the VP of Marketing tells you they would like to build strategic partnerships outside of the industry to get a larger consumer base interested in their products. What do you do after hearing this idea?

First, it’s essential to tailor your inquiries to focus on the deliverable, which will help you bring the most value for the executive. You might ask the question, “Is your goal to earn brand recognition for a large number of consumers, or to reach the individuals who are most likely to purchase?” Knowing this will tell you whether to target pure volume or a specific subset of consumers.

Subsequently, you’d ask, “What is the data source?” as well as “Do you know of an internal data source with this information? If so, are there any pitfalls in the data?” This information can help you in later discussions. Make sure not to get lost in the weeds or ask for too much information (such as which table the data is in). However, if there is a critical piece of information you need to complete the project, it is always best to address it.

Now that you have your mission, you are well equipped to solve the next steps with your data science tool belt. However, before you jump into setting up your deep learning, you should first consider the problem and your audience. If you are producing output for a current or former data scientist, they may prefer the most cutting-edge model. In situations like the one with the VP of Marketing, this will not be the case. They are more likely versed in management principles than data analytics. That means a simpler model may be better. In fact, one of the reasons a logit is still popular is because log-odds are interpretable and easy to explain. (Note: I have also seen this model misused because the underpinning data violates a logistic regression assumption.)

This leads to a point that is hard to admit but incredibly important to understand: most executives do not care about what model you use. They like buzzwords such as artificial intelligence, machine learning, and deep learning, but probably do not care whether you use logistic regression, clustering, neural net, etc. What they care most about are the results, and even more importantly, your confidence in those results. This isn’t just the size of your model error—it’s everything from start to finish. Did you understand all shortcomings of the data, do you have sufficient sample size, did you specify the model correctly, how did you validate your results, and did you interpret the results correctly? The exact answers to each of these points may not be relevant, but your tone and results that you present are. Decisions at this organizational level carry the most weight, as they have ripple effects on sales, budget, human capital, etc. 

The final piece is the crème de la crème: presenting your results. After all that hard work, this is your chance to shine. What is the best format—white paper, PowerPoint, model output (just kidding), tables, or charts? The “right” answer is highly subjective and should be tailored to your executive’s preferences. You should ask around, use your judgment, and ask your immediate manager for help if you are comfortable with their judgment.

Throughout my tenure as a data scientist, I have picked up some additional pieces of advice that may help you as well. I had a conversation with a well-respected brand leader I used to work under. When I started working in her division, she told me, “If you want me to get something out of an email, put it in the first three lines—after that I will probably stop reading.” It’s blunt but honest advice. For me this was a lightbulb moment; it made me realize the importance of keeping things short and sweet. Later on, I learned that the most effective mode of communication for me is an “executive pitch.” It is similar to a sales sheet: a one-page, double-sided document that is a combination of infographics and text.

This format does two things well: it forces you to cut information down to the critical necessities and allows you to focus the reader’s attention on key results through infographics and visuals. Bullets alongside the infographics are a good way to incorporate crucial text points, and it’s good to begin with an executive summary to lay out the problem and key outcome. The key difficulties with this format are the graphics and layout. If you have access to in-house graphic artists at your company, they are a great resource, and I would highly recommend reaching out. They are masters of the visual arts and experts at focusing a reader's attention.

Based on my observations, these tips should put you on a solid path towards effective communication with executives and help you extract maximum value from your technical expertise.

What has your experience been?

Scott Murdoch
Scott Murdoch

Scott Murdoch, Ph.D. is the Director of Data Science at, a healthcare guidance platform that uses AI to reduce healthcare costs. Scott has worked in a wide range of topics in the healthcare data arena, from analyzing billions of medical claims to understanding what drives the consumer experience. His focus is the ingratiation of economics and artificial intelligence.

Related Content