Data science is a discipline in which art meets science. Quantitative best practices are only part of your project’s analytical framework. At the end of the day, numbers need to tell stories that appeal to a wide audience.
“When telling stories through the data, you need to be able to place yourself in the shoes of the listener,” said Matti Aksela, vice president of business analytics at Comptel, a company that automates customer interactions. “In many cases, that is someone without your own knowledge of the statics, models and underlying assumptions.”
Keeping a balance among stories, data integrity and results is a mission-critical task for businesses and department groups of all sizes. Whether you’re running a marketing team, small business, or engineering team, you need to make sure that you tell a quantitative story that connects with a range of audiences.
“Think about a really classy restaurant,” said Aksela. “Will they serve their food just thrown onto the plate?”
Indeed, presentation is key. When telling a story through data, it’s important to have an idea of where you’re headed in terms of your messaging, flow and most valuable takeaways, say experts. This can be especially challenging for SMBs, with limited time and staff, but the following best practices will serve as a guide:
1. Jump into your research with a blank slate
As tempting as it is to seek out “shocking findings” in your data, remember that you are, first and foremost, a scientist. Data-driven stories will unfold through objective analysis. If you’re fishing for information, you may miss out on a story that hasn’t yet been told–something that an open mind can help you uncover.
2. “Read” between the data lines
If you’re building a regression model, test the impact of adding one or two additional variables. Come up with a hypothesis that you hadn’t yet considered. Test it. Measure it as part of your analytical framework.
Some of the most engaging stories are ones that aren’t obvious. Keep your eyes peeled for a new spin on an old topic. Instead of telling the same story over and over, seek out nuances in your data.
3. Incorporate multiple storytelling mediums
People process information differently. Some prefer visual tools like infographics. Others thrive on podcasts and lectures. Hands-on demonstrations are also valuable for teaching core concepts. Make sure that your messaging appeals to a range of learning types.
4. Validate your findings
This concept sounds more complicated than it actually is. In a nutshell, “validation” means testing your numbers against common sense. Can you reasonably explain the trends that you’re finding? Can you come up with an alternative explanation? Talk through your logic, trying to anticipate any and all possible counter arguments.
5. Look for confounders
Confounders are “hidden” variables that affect both the causes and effects in your analysis. For example, you may think that a hard drive failure caused a system failure when, in reality, the two were not related at all. There could have been a virus (the confounder) that caused both to fail independently of one another. Don’t assume that correlation is causation, and always look for alternate outcomes in your results.
6. Put faces to the numbers
People are the heart of great storytelling. Include real-life narratives to make your findings as tangible and human-interest-driven as possible. Your job as a storyteller is to make people care.