Originally posted on Medium in 2022 July
After I finished writing my thesis, I had one last task before completing my bachelor’s degree — giving a 10-minute long presentation about my results. I was looking forward to this task, as I love speaking about topics related to data and machine learning. People were praising my work and the results, so I thought: what can go wrong? Well, as it turned out, quite a lot.
Preparation and presentation
Before I started putting together my slides, I did a little research on how to give a great presentation. One common tip is to always know your audience. Mine were university professors, most of whom have little to no connection to data science. I believed this makes my job a lot harder, as developers like to talk about how they achieved their results, and I’m no exception. A simple question kept bugging me:
“Should I treat this as a technical summary of the project or a speech where I want to promote a product?”
Without simplifying my work too much, my general goal was to showcase the outline of the architecture I used and demonstrate my results in a way so that non-data-scientists can relate to it too. At that time I believed this is the correct answer as most IT professors are intrigued by the technical details.
Then the ominous day came. There I was, standing in the room, waiting for the committee to allow me to begin my presentation. I got off to a great start, but as it went on, I felt like I was losing the committee slowly but gradually. In the beginning, only occasionally were they distracted, but as I kept speaking, it got worse and worse. Seeing this I got frustrated, but I already had my slides on my pen drive, and there was no way to edit them. Soon I finished my presentation with some blank looks. I was not satisfied, so I asked myself: is there something I could’ve done differently?
A week later, I ran into a video from Bill Franks promoting his new book, Winning The Room, where he shared some tips on how to deliver a killing presentation. I was curious whether it’ll be another list of obvious tips and tricks, or will it help me pinpoint my mistakes. It was definitely the latter.
Below I’ll describe the ideas from the video that resonated with me the most, and how they could’ve bolstered my talk had I listened to it earlier.
Takeaways
1. Focus on how to USE your results, not how you got there
In order to make it easier for the audience to understand what you achieved and why your result is important, it’s best to tell them how you would use it. For example, a lot of people have a hard time understanding what a shot of 0.5xG means in soccer. But if you say that an average player has a 50% chance of scoring, suddenly it starts to make sense. Create two chances like that for the player, and he will likely score a goal.
2. Charts and graphs are like jokes: if you have to explain them, you failed already
This is probably too obvious for data folks working close to business stakeholders, but for me, as a student, sadly it wasn’t. Visualizations are best if they can be understood without the presenter saying a single word about them. Let people digest the material themselves, but aid them by following visualization best practices. Colors can help you a lot here.
3. Read the room and adapt (have a plan B)
In case things go awry, you should always have a backup strategy. You might think that there is no reason why your presentation wouldn’t go as well as you expect (just like me, mind you), but you have no idea what your audience will feel, how will they interpret your words or whether your pacing is right for them or not. It’s best to assume something will break down.
If I had to go at it again, I would ask myself the following questions:
- What measure can I take if I see that my pacing is off?
- Which slides can you omit if you are running out of time?
- How can you summarize your presentation in fewer minutes (10, 25, or 50% of your budgeted time)?
4. Close with a “wow” tied to a larger audience
We’ve all heard presentations where we felt like it was unique. It gave us something different, something we haven’t experienced before, something we will try to hold in our memories for a long time. Oftentimes the key to these talks was the closing part: they made us feel as if something extraordinary has happened.
If you have a presentation soon, imitate them by zooming out of your topic and explaining to them what the results could mean not just for you or your audience, but for anyone. It’s too easy to forget about this. It comes as second nature for you already because you already know the ins and outs of your results. But your audience doesn’t, so help them out.
5. The goal is to make the AUDIENCE comfortable during the presentation, not you (the presenter)
I have no idea whether this idea holds in practice or not, but my experience supports it anyways. The question to ask here is: how can you make your audience comfortable, regardless of the theme of your presentation? I’d advise you to hold their hand while you speak. Make sure not to leave anyone behind, and methodically help them in the difficult areas of your talk. Making sure these are well met, your audience will feel competent and confident, regardless of how well-educated they were on the topic beforehand.
Conclusion
Even though the experience of presenting my thesis wasn’t a happy one, I learned two important lessons along the way.
- Giving a speech on a data science topic for people without domain knowledge is hard, especially if you have limited time to explain the main characteristics of the field.
- It’s tougher to present to non-technical people as someone who is heavily involved in the development, as we are moved by different things: nicely engineered solutions versus business impact.
Thanks to Bill’s video, I managed to improve my understanding of what makes a good talk and I hope you got something useful out of this article as well. Regardless, best of luck to you with your next presentation! 😉
Note: If you’re interested in what my thesis was about and how to find the perfect topic as an IT student, I recommend you to read my post on it.