7 Tips for Presenting Data Science to CXOs

We’ve all heard that executives want “actionable insights,” but what exactly does that mean, and how do you present your work to the C-Suite?

Here are 7 ways to make a killer data science presentation.

1. Start with the why.

Image for post
Visualization by author, inspired by Simon Sinek's "golden circle"

Most organizations have a North Star Metric. For AirBnB it may be nights booked, for Facebook it might be Daily Active Users, and for Quora it could be the number of questions a user answers.

Some organizations look at a constellation of metrics, including quantity, quality, and efficiency.

Whatever your metrics are, that’s what the C-Suite is focused on, and that’s what your presentation needs to calibrate to.

You can’t walk in the door (or join the Zoom meeting), display your Jupyter Notebook, and talk about the TensorFlow neural network you built.

Before you jump into the solution, start with the problem you’re trying to solve and why it matters. Explain why your analysis, your model, or your insights matter.

From there, step down the ladder, rung-by-rung, moving from the vision into how you solved the problem and what you learned.

2. Describe the problem in-depth.

Image for post
Photo by Jeremy Vessey on Unsplash. Edited by author

Understanding the problem is fundamental to the rest of the presentation.

When describing the problem, make sure to convey its purpose and importance. This stage is where all attendees seek to fully understand the issues and narratives at play.

By doing so, you can create the vision of a desired future and effective ways of creating it, even before delving into your analysis or model.

3. KISS — Keep It Simple Stupid

Image for post
Photo by Joey Nicotra on Unsplash. Edited by author.

While KISS was originally a design principle, it’s a widely useful tool.

Try to simplify your findings and insights as much as possible, and distill everything important down into an overview (say, 1 or 2 slides).

Those 1 or 2 slides, which address why you’re working on this problem, how you solved it, what you learned and what the next steps are, should be easily understood by someone who wasn’t even in the meeting.

4. Explore assumptions.

Image for post
Photo by bruce mars on Unsplash. Edited by author.

We’re all making assumptions, all the time. For instance, you might assume that the data you have is accurate, and that there were no bugs in the data ingestion or data manipulation processes that could impact accuracy.

If you’re dealing with site traffic data, for example, you might also assume that this is data from human beings, although bots make up a significant portion of web traffic.

If you’re building a predictive model, you might assume that how your product or service performed in the past year is reflective of how it will perform in the next, although there’s a pandemic, supply and demand shocks, and overall economic turmoil to take into consideration.

Exploring assumptions will reveal where your model might fall short, while also uncovering potential biases.

5. Explain your model or analysis (in English).

Image for post
Photo by Annie Spratt on Unsplash

Talking about how your “random forest algorithm searches for the best feature among a random subset of features” is technically English, but it’s not plain-speak.

This is similar to point #3, but your model should be explained clearly as well, not just the insights.

For example, your RF model may be better explained as a “classification algorithm that’s resistant to overfitting, which means that it’s able to generalize well to new data. It’s also a really scalable model, and descriptive of the data we used to train it.”

If someone wants to go into more detail, you should have that information prepared in the back of your head (or another set of slides), but you shouldn’t force everyone to sit through it.

6. Make beautiful visualizations.

Image for post
Slide on “Robinhood Growth” based on available estimates. Created by author.

We’re all visual learners, as we process visuals far faster than text. In other words: Use simple, clean graphs over raw data or numbers.

While you don’t need the skills of a designer, understanding some of the basics of design can help you make more memorable, beautiful presentations.

Good design is a lot more than just simplicity, but removing extraneous details from your visualizations is an easy step that can make a huge difference.

7. Write well.

Image for post
Photo by Jeevan Jose on Unsplash

Have you ever sat through a presentation where walls of text were forced upon you? You just think to yourself, “c’mon, I’m not reading that!”

We don’t have the attention span for uninterrupted text.

Good writing comes down to four essential ingredients: Simplicity, clarity, elegance, and evocativeness.

You don’t need to become the next Shakespeare, but you can greatly improve your presentation’s text by removing unnecessary details and striving to write as little as possible, while still getting the meaning across.

Conclusion

CXOs are busy. Calibrate to that reality by giving a purpose-driven, easily-understood presentation with concise writing and beautiful visualizations.

---

Article image by Campaign Creators on Unsplash

Frederik Bussler

Frederik Bussler is the Founder of the Security Token Alliance. As a public speaker, he has presented for audiences including IBM, Nikkei, Slush Tokyo, and the Chinese government, and is featured in outlets including Forbes, Yahoo, Thrive Global, Hacker Noon, European Commission sites, and more. Recently, he represented the Alliance as a V20 delegate.