By Michael Berk, Data Scientist at Tubi
Communicating is one of the most challenging aspects of a data science job. Here are my notes…
What the Internet Thinks
There’s an old research-based adage that 93% of communication is nonverbal. 55% of your communication is body language, 38% is voice tone, and 7% is spoken word.
Figure 1: communication breakdown. Image by author.
What happens to body language communication over a video call? Well it usually just disappears.
You should try to recoup that 55% and place more emphasis on tone of voice and spoken word.
Some useful for tips for bridging the body language gap include exaggerating vocal inflections and changing your volume. Increasing your facial expressions and hand gestures can also improve the reception of your presentations. It might seem unnatural, but if you record your talks and rewatch them, you’ll be surprised at how normal and charismatic these changes appear.
Another interesting idea is the content, design, and delivery framework — src.
Figure 2: Content, Design, Delivery framework. Image by author.
The latter two, design and delivery, just refer to minimalist slide design and simple phrasing of complex topics, respectively. However, the content portion was really interesting.
In short, the article postulated that your listeners will only take away one sentence from your presentation, so make it count. To do this, you need to understand their technical level, expectations, and prior knowledge of the project. If you tailor your talk to your audience, you can make that one sentence count.
One really easy trick is to incorporate visuals from other teams in the organization. For instance, show a UX research video that supports your claim. By leveraging prior work, you save time, create a compelling presentation, and build relationships within the company.
What I Think
Despite those awesome ideas, the vast majority of information on the internet is common sense. In this section, we’re going to focus on strategies that are less obvious. Let’s dive in…
1 — Get Inside Your Audience’s Head
For meetings with more than ~10 people, it’s safe to assume that one of them doesn’t want to be there. They are busy people with busy lives, just like you.
So, try to understand what makes your audience excited. Excitement is what gets stuff done. It makes your work visible and thereby impactful.
While roles differ can greatly between organizations, most data scientist have some freedom to select and develop their own projects. If the success of your work involves stakeholder buy-in, you must get them excited about the project. Period.
There unfortunately isn’t a clear shortcut to doing this, but here are some things that worked for me:
- Socialize ideas early and often. By getting feedback from teammates and stakeholders throughout the project, you build excitement and produce a more valuable final project. Trust me, it’s worth the effort.
- Understand what your boss’s boss cares about. If you know what will make upper-level management excited, you know what will make lower-level management excited. Schedule a quick one-on-one and ask good questions.
By connecting your work to exciting ideas, you can dramatically increase the value of your presentation.
2 — End Meetings Early
Unnecessary information is harmful. You’d think that showing your steps and assumptions would be beneficial to your listener. Well, in most cases it’s not.
By including information that is unnecessary to their comprehension, you are…
- Increasing cognitive strain. This decreases attention span, reduces trust, and often leads to multitasking during your presentation.
- Subconsciously biasing your listener. Everyone has a confirmation bias which influences them to affirm their beliefs. If you give the listener extra information, they are more likely to latch on to attractive ideas and miss key conclusions.
- Wasting time and money. Time is money. Try this exercise — add up the hourly salary of everyone in the call. It’s expensive to present unnecessary information.
A simple yet effective proxy is to try to end meetings early. One of my teammates has become beloved by stakeholders and data scientists alike because he always ends meetings early. Now, there are obviously exceptions to this rule, but those exceptions are much rarer than you’d think.
If you aim to end meetings early, you are forced to be structured, concise, and relevant.
3 — Some 80/20 Tips
The above two sections can be labor intensive, so here are some simple tips that will hopefully get you 80% of the results in 20% of the time.
- Lean in to your communication style. If you are a formal speaker, speak formally. If you tell jokes, tell jokes. By leveraging how you naturally talk, you become more confident and relatable, and thereby effective.
- Start and end with the conclusions. By starting a presentation with the conclusion, you reduce cognitive load and allow listeners to think more deeply about your ideas. By restating the conclusion at the end, you facilitate mental chunking which promotes recall and helps them leverage the new information.
- Say “um” and “like” less. Repeatedly saying these words sadly make you sound dumber. Sorry. Here’s some base code for a very poorly-made but free “um/like” detector.
- For technical concepts, treat it like a black box. Most listeners don’t care how the method works. They care what it does. So all you have to do is explain the inputs and the outputs.
Thanks for reading! All of the resources linked above have been really influential for me and my career. Please share your own.
Original. Reposted with permission.
- How to Find Weaknesses in your Machine Learning Models
- Don’t Waste Time Building Your Data Science Network
- 5 Things That Set a Data Scientist Apart From Other Professions