What is storytelling in Data Science? : Interview with Sergio Ballesteros Solanas

 Can you make a brief introduction of your professional background? 

I’ve been working in TomTom for one year as a data scientist.

First, started in the MyDrive team. They have a platform called Road Trips which allows you to create personalized routes which can be your commuting trip to work or your next holidays. I did my Master Thesis on how to use machine learning to improve the recommendations to the consumers.

Then, I moved to the navigation department. That's where I’m working now. I’m involved in decisions making and providing insights based on data. My work is provide advice on  decisions that have high risks.

Before that, I had several jobs. I was a mathematics, physics, and chemistry teacher for high-school and university students.

Storytelling was an important part of the job to keep students engaged. It was keeping them away from distraction.

After that, I worked as a professional fitness coach to help people to reach their fitness goals.

Whether they wanted to lose weight, being faster at running, it was important to show them their progress. For that reason, I was tracking their performance using data and analytics.

 


 What is storytelling in Data Science? 

Storytelling in a broad sense is a tool that helps people to communicate better. It gives a better understanding of your work and makes people remember your talks longer. It allows them to understand the insights and their impact.

This is because we are living in a world overloaded with data!

We need to find a way to connect all these points with a series of logical events.

It has two main benefits:

  • It retains attention: in any given day, decision makes might see many presentations from many people, so with story telling you can avoid that they are distracted and they lose the interest in the first stages of your talk.
  • It gives allows them better understanding: By nature, people have a better understanding of connected events. When you tell a story, they understand better your motivation and your action point or what you’re trying to explain.

 


 How would you present technical results to a non-technical person? 

I come from a background of physics. In science, storytelling is a hot topic. You need to explain very complex stuff to people who have never heard of it.

It is important to focus on the what and the why but not so much on the how.

This means that you need to clearly show what is the context of the analysis and why it is important, and what are the data input for you analysis, what variables you decided to study and which ones to leave out. A priori, you can leave your complex advanced analytics methods as black boxes which can be explained at a high level upon request.

The message that you want to send has to be clear, and you should supplement it with intuitive visualizations.

For example, we work a lot with trace data (so called GPS coordinates). When we want to understand the behavior of our drivers, instead of describing it only with words and graphs, I often use the TomTom Maps API’s which contains some rich tools that allow correlating the trace data with real world events.

 


 Where do you find your story inspiration? 

I always try to start a presentation giving the motivation of why I’m personally interested by doing this analysis.

It is not only doing analysis and showing the results. In the process, you can always identify yourself in the problem you are trying to solve.

In TomTom we have a lot of projects available and often we can choose which one we want to work on. Of course, you have to choose the one that has the most business value but you can always relate to why you chose.

When I work on a new feature that uses machine learning, I imagine myself on how this feature would have made my life a lot easier.

That is the initial inspiration. Many other times, if instead of creating a feature you are doing an analysis for data driven decision making, the story to tell is the story contained by the data itself, and the only thing that you need to do is to show how you tried different hypothesis before coming to the big, final insight.

 


How do you organize yourself when you make a presentation on a project that you are working on?

 

The first question that I always ask myself is: Who am I presenting to?

Are there technical people, are there Data Scientist or are there decision makers?. When I speak to technical people they might be more interested in the methods that I used. Then, I can go a little bit deeper into the math behind. If there are not technical people, I follow the guidelines described previously, but still there are different way in which you can focus your talk:

  • Are you presenting to a manager?
  • Are you presenting to a product owner?
  • Are you presenting to the CTO?

Depending on who I am presenting to, I decide to include certain topics and exclude or do less emphasis on others. For example, for a product owner, I would focus more on the actual business value that you the decisions that I suggest will bring.

That it concerning the content, but structure of how to show this content always follows the classical narrative of a story.

I always begin with the context of the analysis, this is what is the current situation of the project, or the company. Afterwards, I present the main challenge or problem to solve, and I provide several hypothesis that I tested but turned out to be false. This is important because many times they even jump to ask more details about that. When the audience is engaged I finally I show which one was the final hypothesis that turned out to be true was, and what are my recommendations and action points to be considered.

Using this path, you show to the decision makers that you considered several options and they will understand that you are providing the best possible advice.

 


In your opinion, to what extent storytelling is a skill that data scientist should learn?

This is a very important skill. You can be a great coder, good at math, or perform great analyses.

But if you don’t know how to present it in a way that people are going to engage, then people will forget your report.

That is pretty much how everyone, including myself started in this field. Then, I started structuring my presentations in such a way that people will be able to connect all the dots by narrative.

I realized that people were more engaged. They even approached me when the presentation was over to ask me questions. And in fact, people liked some of my presentations so much that they talked about them during the following days and new opportunities where opened in similar projects. This indeed leads to a greater impact of my work.

 


 How did you learn storytelling? 

It came naturally for me. When I was teaching physics to students, I didn’t want to just show them the formulas and explain the theory, I went one step beyond and explained what the historical context was when those formulas were discovered by the scientists. I realized that telling short stories would even keep engaged the students who were little interested in science.

Unfortunately, this is not something usually taught in many Universities. For example during my Master’s Degree in Italy, the courses focused a lot on math and software engineering - not so much on communication.

 


 What advice would you give to a Junior Data Scientist who would like to learn storytelling? 

You don’t need to jump right away into Data Science storytelling. You can start using storytelling in general.

If you are able to approach any of your friends, family or someone you just met at a conference, and tell a story that makes stay still and listen quietly to you, then you have 90% of the job done. The remaining 10% is applying the same method to the data analysis that you have done, and combine it with impactful images and plots.

 

Follow us on Linkedin and read our daily articles articles: https://www.linkedin.com/company/digital-source/