Join Doug Rose for an in-depth discussion in this video Avoid pitfalls, part of Learning Data Science: Tell Stories With Data.
- [Narrator] Many teams are more comfortable just presenting data as a raw set of numbers. They feel that data is the facts and that a story is less professional. They believe that the data speaks for itself and that the sheer power of the numbers will compel the audience to act. This is especially true if your culture is focused on objectives and compliance. In these organizations you don't need to tell a story about how your project is on budget. You don't have a story about the number of milestones you've completed.
Remember when you're working in data science you're not trying to communicate status reports instead you're trying to discover something new. The science in data science is about exploring the data. Data is complex and needs to be interpreted. Your audience will look to you to show them more than just well designed reports. They'll want the meaning behind the data. Think about any time that you've come in contact with complex data. Maybe you checked out the weather information, maybe you wanted to see how a candidate was faring in an upcoming election.
Each of these are a complex data problem. That's why weather reports are not always accurate, that's also why political polls are not always right about an election. The majority of people won't go deep into your reports and analysis. Instead they'll want to be told a story. They want to hear what you think about the data. Giving them too much data is not only unhelpful it might even be overwhelming. Imagine you're watching a political show and the commentator puts up four complex charts.
Then they say, "Well as you can see "the data speaks for itself." Most people just change the channel. It's the same with data science. If your story is just data visualizations then your audience will quickly dismiss your presentation. A good data story will use visualizations as a garnish. The story behind the data is the meal. There are a few things to watch out for to make sure you don't over rely on your data visualizations. The first thing you can do is limit the data in your presentation.
If you're using slides then how many do you have? If it's an hour presentation and you have 30 visualizations then you're not telling a story. Instead you're probably just showing everyone your data. The second thing you can watch out for is how much time are you spending preparing your visualizations? It's terrific to you make sure that your charts are clear, just remember that the charts are one of the first things that your audience will forget. If you want to have maximum impact then you should focus on the things your audience will remember. Your audience is more likely to remember a short interesting story.
If your organization has a very conservative management culture then it might be difficult to tell stories. Often in these organizations it's politically safer to just present your visualizations and then leave it to the managers to interpret the data. You might be tempted to just portray yourself as an impartial presenter. The problem with this approach is that if you're on a data science team you're still responsible for the outcome, so you'll be on the hook for whoever interprets your data. In these situations it's usually better to use a story to express your opinion.
That way at least you'll have some control over whatever happens to your results. Finally it's often very difficult for new teams to accept that you can create a story from data. Some data just looks like lifeless columns of numbers. It's a real challenge for those teams to look at those digits and reverse engineer the humans that created them. Frankly it's one of the biggest challenges in being on a data science team. The best way to avoid this is to try and humanize your reports.
Don't call a report, "Upcoming Consumer Trends", instead call it something like, "What People Are Buying". These little steps can help make it easier to think about your data as real events. Like any skill data storytelling takes time to improve. Start thinking about the key characteristics of a story such as plot and conflict, then work to present your data in an interesting way. Over time your stories will become more interesting, you might even make stronger conclusions or braver interpretations.
Try to remember to have fun with your stories. It will improve your delivery and make you a more interesting storyteller.
- Structuring a data science story
- Defining plot, conflict, and details
- Going beyond reporting
- Knowing your audience
- Working with data
- Introducing visuals
- Eliminating distractions
- Incorporating metaphors
- Motivating the audience
- Avoiding pitfalls