On Business Communication and Data Science Presentations

I’ve been fortunate enough early in my career to have a tremendous amount of exposure to the applications of data science, and I can confidently say that there is a data driven solution to every business problem if you’re innovative enough.

During my time at Metis I directed ten iterations of the data science bootcamp, graduating 218 students. I’ve seen over a thousand data science presentations, covering topics including predicting overdose mortality, optimizing an NBA career, predicting court decisions using machine learning, and even music composition using machine learning, to name a few.

After collecting approximately one thousand data points, here are some best practices for crafting your data science presentation.

Streamline your introduction.

Your introduction shouldn’t be the longest part of your presentation. Focus on presenting the basic context of your problem and highlighting the business use case, especially anything that has a financial impact. If you can highlight specific figures, such as potential savings as a product of your solution, be sure to do so. Including an overview of your presentation or the workflow you used to create your model will help give context for what you’re about to deliver.

Don’t: spend two minutes trying to convince me that you’ve found a problem worth solving.
Do: focus on big picture items, especially any financial impacts of the problem or solution.

Visuals need to serve a purpose.

Visualizations are one of the most important parts of a data science presentation. I’m a big fan of making things pretty, but not at the cost of interpretability. Everything about a visualization matters: the title, axes labels, the colors you use, and how suitable the visualization is for the data.

Don’t: make things colorful without a purpose, choose a slide deck template that doesn’t match the presentation or one with a lot of unusable space.
Do: prioritize visuals in your slide deck. A good rule of thumb is that each visualization should have its own slide and take up most of it. If you have multiple visuals you’d like to compare side-by-side, introduce each one on its own slide before showing the slide with both.

Your slide deck should be a vehicle for your visualizations.

My golden rule for any presentation is that if you’re going to have something on a slide, it should never be a full sentence. Large amounts of text on a slide distract from the presentation because if I’m reading your slides, I’m not listening to what you’re saying.

Don’t: rely on your slide deck to tell the whole story. Avoid the use of full sentences, writing things out that you will be saying out loud, fonts or colors that are hard to see, and negative space on slides.
Do: use short sentences or phrases and the important numbers (statistics, model results, financial impact, etc.) to cue yourself on what to say next. Pick a slide deck that allows you to highlight your visualizations and any model results.

Rehearse.

This one seems self-explanatory, but you’d be surprised at how many seasoned professionals will show up without having rehearsed. Even if you’re someone to whom speaking naturally and off-the-cuff comes easily, it’s still important to practice. An audience will notice if you aren’t fully familiar with the slides or if you’re scrambling to find a way to explain your findings. Remember that a good presentation isn’t just about showcasing your findings, but about communicating your solution in the best possible way.

Don’t: wait until the day of to practice what you’re going to say. On the day of the presentation, be careful not to say anything off-the-cuff that can be disproved.
Do: practice with a timer, in front of a friend if you can or a mirror if you can’t. Add a cue for yourself at the places you normally get stuck — tying a sentence to a visual aid is a great way to help yourself during presentations.

Own your narrative.

Guide me through your presentation — guide my thinking. The same way a magician uses subtle techniques to guide your attention away from the hand doing the trick and towards something specific, guide my attention towards your goal.

Don’t: sow seeds of doubt or make unsubstantiated claims — if you say something and follow it with self-deprecating remarks, you’ve tanked your credibility with your audience. Avoid using absolutes or similar phrases that will nudge your audience to disprove them.
Do: be mindful about the order in which you present concepts, and structure your presentation with a clear beginning, middle, and end. Remember the serial position effect — things mentioned at the start of a presentation are best remembered, followed by the last things you say, with the things in the middle being the least remembered.

Pay attention to the little things.

Know definitions before you use them (i.e. don’t refer to R2 without being able to explain what that score means). Be mindful of your word choice and the aspects of your model that you highlight. Beyond what you say, make sure that everything on your slide deck is in good order. All your graphs should be labeled, everything should be spelled correctly and formatted properly. These small details are like the stage hands in a play — if all goes well nobody will notice them, but one step out of place and that’s all they’ll remember.

Don’t: use your first draft at a slide deck as your final presentation.
Do: Proofread for errors in spelling, grammar, formatting, etc.

Know your audience.

Tailor your presentation to your specific audience. What you highlight will be different when you’re talking to a team of senior data scientists and when you’re speaking to a board of directors.

Don’t: deliver an extremely technical talk to a non-technical audience. Similarly, don’t give a broad overview without technical depth to a technical one.
Do: research who you’ll be speaking with ahead of time. If it’s an onsite engagement, try to arrive early and test drive the presentation room.

Remember the money.

At the end of the day we live within a capitalist system, so you need to remember the money when crafting your presentation. The easiest way to get buy-in from shareholders is to emphasize the financial impact of the problem or solution you’re explaining. Remember that social good is something you can highlight as well, so if your problem affects stakeholders factor that in.

Don’t: forget about profits when creating your presentation.
Do: showcase any financial implications in your slide deck, and work to maximize the business use case of the solution or problem you’re presenting about.

Trust yourself.

Imposter syndrome is real. Cohort after cohort I’ve seen accomplished, intelligent students — from those with nontechnical undergraduate degrees to PhDs in computer science — suffer from feelings of inadequacy.

Don’t: doubt your abilities once you’ve proven yourself.
Do: find ways to stay motivated and showcase your growth to yourself as you develop yourself professionally. Go back and look at your accomplishments regularly, and remember to show compassion to yourself. Being a person is hard.

Use your landing slide.

At the end of your presentation, your audience will need to look at something. Leaning on the serial position effect again, make sure you end on what you want your audience to remember. If you’re actively seeking new employment, showcase your contact information. If you’re presenting a new chatbot to the marketing team of your company, showcase the results in a way that matters to them (in this example, this might be how many hours using a chatbot can save their team).

Don’t: end your presentation on your last data slide.
Do: create a landing slide that features your name, any relevant contact information (LinkedIn URL and github username are good ones).

As I wrap up my time at Metis, I hope my experiences can help the next time you need to create a presentation, data oriented or otherwise.