Some people think that Data Analysts must be good communicators, but for Machine-learning Engineers it isn't so important. This is not true! Regardless of your role and level of seniority, as soon as you work with other people and not robots (99% of cases) not knowing how to communicate will make you a person hard to work with and sooner or later slow down your career.
Verbal communication, e.g presenting your results at a working meeting or conference, is very challenging. Below are 5 things, which I find helpful to get the most out of presentations I do.
1. Know and respect your audience
When your colleague is asking you "How is work?" your answer will be different compare to answer for the same question from your grandmother. The same should be true when you communicate your results. Always try to find out who is going to listen to your presentation and adjust the topic and a way of presenting it accordingly.
Keep in mind that people had to make time in their busy schedule to come and listen to you. Just preparing a talk, but not preparing it for THEM is essentially showing disrespect to your audience and their time.
2. Be clear about your goals
Even if your presentation is a status update, there is always a goal you would like to achieve.
Examples:
- Explaining why fixing the bug took so much effort
- Justifying a new metric for ML model evaluation
Everything, which is not related to the goal, should NOT be part of this presentation.
3. Tell a story
Our mind is always looking for the best usage of its time. This is why focusing attention is such a difficult task. Your audience is constantly switching between listening to you and thinking all kinds of thoughts. A story is like a line, which your audience minds can follow and naturally get less distracted.
Nobody expects Agatha Christie level of stories in your presentations. Often making sure that there is an introduction explaining the main purpose of your talk and a connection between all slides is already good.
4. Less is more
Slides with a lot of of information, e.g. long sentences and/or overloaded graphics and/or tables with many numbers, are overwhelming. Quite often this slide will become the point where the audience will switch to their own thoughts. To avoid it remove everything which is not crucial for your message from the slide: distillate sentences to few words, leave table only if there is no time to create a graphic out of it, remove not needed lines, grids and other visual noise from your plots.
5. Rehearsal
If there are just 5 people, who will listen to your 30-min talk, it's already 150 min of other humans' time. If your presentation is bad, it's 150 min of wasted time. Rehearsal is the best way to minimise the risk.
Set-up a timer and go through the slides loudly saying what you will say at the actual presentation.
Keep in mind the following points:
- Will you manage to deliver your message in the allocated time?
- Does the slide layout help you to remember the order of things you wanted to discuss?
Ask your friends or colleagues to participate in a rehearsal and give you honest feedback.
To learn more about story-telling and visual presentation of results I highly recommend book "Storytelling with Data" by Cole Nussbaumer Knaflic.
And here are few talks, I find very inspiring:
Comments