Preparing to apply for our immersive bootcamp? View Resources

3 Ways to Improve Your Data Science Communication Skills

By Lara Kattan • August 10, 2019

This post was written by Lara Kattan, Metis Sr. Data Scientist.

Data scientists are often described as hybrids: part statistician, part computer scientist; part analyst, part strategist. But while we focus on the myriad technical skills that a data scientist should possess, we often overlook one of the foundational skills (without which the whole edifice falls apart): communication skills. Being able to build models is all well and good, but if you can’t succinctly express the assumptions, takeaways, and next steps to the rest of the company, then your model isn’t worth the bits it’s made up of. 

I’d like to take a minute to convince you that communications skills are essential even for a technical position like data scientist – but more importantly, I want to leave you with actionable strategies you can start putting into place today to improve your own, or your team’s, communication.  

A talented data scientist working in isolation doesn’t help anybody. Communication is what turns potential insights into actions that benefit the business. And that’s why it’s essential for those in technical roles to develop their ability to communicate potential and risks to all decision-makers. 

If you can’t trust your data scientists to talk to your C-suite, you have a problem. You can invest in translators or team members who specialize in turning data science into business-speak, but this may not be the right choice at every scale. 

But worry not! No matter if you’re a seasoned presenter, or if the idea of public speaking has you breaking out in hives, there are proven strategies to make you more comfortable and effective. You can apply these strategies to job interviews, presentations, conference talks, and to improve your team’s outward communication. There’s nowhere that great communication is out of place. So let’s get to it! 

  1. Avoid presentations of how you "data science-d"  

    I teach a data science bootcamp that requires regular presentations, and this is the advice my students are most often vexed by, especially in their first presentations.

    If you don’t know how to structure your presentation or explanation of a project, you might be tempted to talk about how you data science-d. This might look like: I got data from such source; added it to this database; downloaded it into a Pandas dataframe; ran this scikit-learn model on it; and here are my model coefficients and test statistics. No one, not even your data science instructor, wants to hear such a presentation.

    The parts of your project that took the longest time to figure out or problem-solve might not actually be the right ones to spend the bulk (or any) of your presentation on. A phrase we borrow from writing is to “kill your darlings.” In most cases, you’ll want to keep your presentation about why you chose a particular model and how the business can use it – not the technical differences between L1 and L2 regularization.

    An effective and simple way – especially when you’re getting started – to avoid the how-I-data-scienced trap is to start with the model outputs and conclusions. If you begin with the output, then it’s harder to simply recount your model building exercise. As much as we may enjoy the intellectual exercise of model building, that’s not usually why we’ve been hired as data scientists. You’ll build trust throughout the organization, and be tapped again for new assignments if you’re able to translate the tedious work you did into pithy but rigorous next steps for the business. 

  2. Calm your nerves with deliberate practice and breathing exercises   

    You wrote and practiced (you did practice, right?) an excellent decision-driven presentation. But the time to speak is now, and you’re frozen. What to do? First, know that nerves happen to even the most seasoned speakers and anxiety alone won’t derail your presentation. Second, deliberate practice beforehand is the key to executing well things that make you nervous, because you’ll be able to ease into autopilot a bit.

    By deliberate practice I don’t mean sitting at your desk, scrolling through your PowerPoint and imagining your presentation in your head. I mean actually talking, out loud, and in conditions as close to the real thing as you can manufacture. The more you do the scary thing and survive, the more you’re conditioning yourself to realize how not scary it is. And once the time comes for the actual presentation, you can rely on that practice to take you through a good bit of anxiety.

    Breathing exercises may seem silly and embarrassing, but they work. Your body is smart, and it’s evolved to react in certain ways to protect you when there are threats. But your body is also a little dumb because it can’t differentiate having to run from a lion from having to deliver your model to the risk committee.

    So when you get nervous and your body thinks there’s a threat, your autonomic nervous system kicks in and changes your breathing, amongst other things. Now we can sort of trick it and get it to work for us by slowing down and regulating our breathing so it thinks the threat has gone away (activate the parasympathetic nervous system, and so blunt the sympathetic nervous system). My favorite breathing exercise is to breathe in for five counts, hold it for one count, then breathe out for five counts. Repeat this a dozen times.

  3. Add vocal variety and emotion to engage your audience 

    If you’re not used to practicing vocal variety, you may feel a little silly at first. But as I tell my students, and as my speech and debate coaches told me for years in school, is that the effect is diminished the further away you are from the person speaking. This means that whatever feels excessively emotional and overdone in your own head translates as only minorly affected to your audience; and the bigger the audience and farther away they’re sitting, the more your tone is muted.

    I generally tell my students to speak at a volume and with enough inflection that they feel inside their heads that they’re being a little ridiculous. This is about the amount of ridiculous you should feel for your audience to find you sufficiently engaging.

    Volume is, of course, one of the more important characteristics you can change. If people have to strain to hear you, they’re going to stop trying at a certain point. This may not be fair if you’re naturally a soft-spoken person, I get that, but you don’t need to change your natural personality in order to project a little more.

    Experiment with vocal variety by talking a little faster sometimes, then a little slower. Remember in the last point we talked about deliberate practice. Work vocal variety into your deliberate practice; go through your script and find where it makes sense to take a little pause, then take it. As you work on it, you’ll become more comfortable with holding silence for a second or two in service of a point.

    If this doesn’t seem natural to you, record yourself as you practice. Honestly, you don’t even need to listen back to it for the recording to be useful; sometimes just knowing that you’re being recorded can be good practice for speaking to an audience. And if you can stomach listening to your recorded self, even better! Play it back, but be kind to yourself. Make note of things you’d like to work on – and try again. Deliberate practice! 

You probably already have strong intuitions about what you like and don’t like to see when others present. The quickest way to improve your own presentations is to note what you like and try to do more of it, note what you don’t like and try to do less of it. Once you start doing this accounting exercise, you may even notice that things you hated about your own presentation style, like a verbal tic, aren't as off-putting as you had imagined. This is true; notice how much more forgiving you are when others present than when you yourself are up there, and then give yourself a break!

___

Learn more about Lara and the rest of the Metis team here


Similar Posts

data science
Does the Data Speak for Itself?

By Alex Nathan • November 25, 2019

In this post, Alex Nathan, Co-Founder of Aventrix Analytics and part-time Metis Corporate Training Instructor, asks the question: Does the Data Speak for Itself? As it turns out, the answer is no, because data alone is typically not enough to arrive at any conclusions. The following equation explains the source of discrepancy: DATA + ASSUMPTIONS = CONCLUSIONS. Read on to find out how data scientists can address this issue.

data science
Made at Metis: Clustering NBA Playstyles Using Machine Learning; Automatic Pricing on Etsy

By Emily Wilson • November 04, 2019

This post features two final projects from recent graduates of our data science bootcamp. Take a look at what's possible to create in just 12 weeks, including these projects on Clustering NBA Playstyles Using Machine Learning and Automatic Pricing for Etsy Sellers.

data science
Made at Metis: Classifying Car Images; Navigating Media Bubbles with Data Science

By Emily Wilson • December 04, 2019

Read about two projects from recent graduates of our data science bootcamp to see examples of what's possible to create in just 12 weeks.