Most data scientists give mediocre presentations. But not you.
Identify Your Target
Who is the audience? Your peers, your manager, executives, the entire company? Suppress the urge to show off, and instead make the presentation accessible to the least technical person in the room.
Assume your audience knows nothing about the work. People are busy - they’ll have forgotten last month’s discussion. Introduce the problem properly. Don’t skip straight to methodology. Context is key.
Structure matters. Include an agenda slide. List sections. A beginning, middle, end. Make it easy for your audience to understand where we are and where we’re going.
Regardless of the audience, put your name on the title slide. It’s your work, and you deserve credit.
Keep It Tight
Nobody wants a hundred slides. Add an appendix.
Not sure if a slide should go in the main deck or the appendix? Appendix. Or better yet, delete it. Nobody reads the appendix.
One chart per slide. One metric per chart. Dual axes are the devil. All words matter, so avoid clutter. Eliminate what you don’t need.
Be Credible
Spelling and grammar mistakes are never acceptable. Don’t lose your audience’s trust because of a typo. Ask a friend to review, or use appropriate software.
Formatting should be consistent across slides. Take the time to fix these things. People notice, and it reflects on you and your work.
If your company has a slide template, use it. We’re data scientists, not graphic designers.
Purposeful Presentation
So you did some analysis. Perhaps some excellent analysis. Still, why should anyone care? You need a result, an action, a recommendation, or at the very least a concrete next step. Don’t end your talk with a shoulder shrug.
If your presentation is successful, it will be shared, parsed, copied, discussed, torn down - all in venues outside your control. Make sure each slide stands alone. Not everyone will see the entirety of your grand vision.
The Delivery
Practice makes perfect. So practice. Time yourself. But don’t practice too much or else you’ll sound rehearsed. 3x at most.
Audience questions derail the best of us. Should you allow them at the end? At any time? During section breaks only? Up to you, but specify ground rules up front. It’s your presentation, and you’re in control.
If you take nothing else away from this: SLOW DOWN.
Every data scientist talks too fast. Your audience knows less than you. They need time to absorb. If you’re nervous, you’ll talk too fast. If you’re over-confident, you’ll talk too fast. If you have a time limit, you’ll talk too fast. Slow down, and don’t talk too fast.
You’ll kill it.
Hacking Data Science provides practical career advice for data scientists, aspiring and practicing. You’ve just read an unlocked paid post. Paid subscribers get access to two paid-only posts each week, along with our full archive.