As data scientists we love to geek out about the latest and greatest approaches to building complex machine learning models. And that’s great; technical nous and a desire to keep improving our modeling skills is WHY we chose this vocation. But we can’t forget that every successful analytics project must ultimately be shared with a wider audience – an audience who aren’t as data savvy as we are. Which is why the ability to craft and tell a compelling story from the analytics is arguably even more important than the analytics themselves.
Most of us have had the experience: you spend some time working with an interesting data set and your analytics show some cool results that you think will have a huge business impact. Your client is keen to get up to speed with what you’ve found and you’re excited to share it with them. But during the meeting the dialog seems mostly to focus on techniques and metrics, and at the end you get the feeling that the client doesn’t seem to have understood the full value of your work. Why? Because the result wasn’t shaped into a cohesive narrative to convey the outcome effectively.
What could have been done differently? By thinking about your result as a story to tell, you frame your efforts in a different way. A story needs a beginning (the business problem to be solved and its context), a middle (the relevant solution approaches), and an end (the insights that can be drawn from the work – what the audience can take action on). Don’t underestimate the time and effort that a good story requires: my rule of thumb is that ⅓ of my efforts on a project will be towards telling the story. But be confident that this time is very well spent. The story is the only part of the project that most people will ever see – make sure they’re wowed by it!
Finding the right story line in your work is often not obvious. Until you’ve spent sufficient time understanding the business question and the context surrounding it, you won’t have the intuition necessary to know what hypotheses are worth testing. And until you do spend some time doing the descriptive analysis and maybe running a few tests or building a basic model, you won’t know where the solution is likely to be. Maybe the story line is beginning to become apparent. But if you find yourself to be so deep into the problem that you can’t see the forest for the trees, try running your thinking by a colleague who isn’t familiar with it – their distance can help them see it more clearly.
Even the best said stories have to go through some rigorous editing. This applies to the story you as a data scientist will be building. So be ready for the analysis that you spent 100 hours building, to not making it to the final cut; Because it may not be necessary to convey the core idea.
Once you’re confident of the story-line, you need to change your working approach so that your further efforts are geared towards telling that story and communicate the results in the most effective manner. From this point forward, before starting any piece of work you should ask yourself the question: “How will this add to the story?” Everything you’re doing should contribute either by making the story clearer or by extending the analysis to complete the story. Note that the story will likely continue to evolve – thinking from the story perspective may point you towards gaps in your work so far.
To wrap up, make sure you give storytelling the time and attention it deserves. Good analytics well told are much more powerful than great analytics poorly communicated. Make storytelling with data one of your key differentiating skills!
And finally, if you can appreciate the picture below, it should be easy enough to understand why you need to take the effort to “plate your story.”