So you've done an analysis that's developed a powerful insight. Now what? If you can't communicate your results to the appropriate audience, whether it be line-of-business workers or executives, the analysis isn't worth the code it ran on.
Simply giving someone a number from an analysis rarely accomplishes anything. People want to know why. Data science storytelling aims to give your audience that concrete, numerical answer and communicate it in a compelling way, turning complex analyses into actionable insights.
"Data storytelling has spelled the difference between getting project stakeholders' buy-in or not," said Denise Tan, data storytelling analyst at StoryIQ, a data analytics consultancy.
Getting a project's initial approval requires presenting data to support its merits as well as show its risks. With data science storytelling, the focus shifts from passively sharing information to proactively demonstrating how a stakeholder's objectives will be met, how their fears will be addressed and how even a small project will fit into organizational goals.
Tan said the power of data science storytelling comes from communicating data in a way that makes it understandable and relatable and connects the data to the human experience. Data storytelling has emerged as a critical skill in data science projects.
Complementing other roles
Andrea Godfrey Flynn, associate professor of marketing and the academic director for the Master of Science in Business Analytics program at the University of San Diego School of Business, said that data translators who specialize in storytelling have evolved to become one of the key roles to complement data engineers, data scientists and data visualization experts. Data translators serve as a link between these other team members and the business managers or clients who will put the analysis to use to solve problems.
These individuals need excellent communication skills, since they need to be able to speak both the technical language of the data scientists, data engineers and data visualization specialists and the managerial language of the businesspeople who have pressing questions to address.
"Using effective storytelling to translate the analytics into meaningful insights for the business users is critical," Flynn said. Otherwise the two groups can experience a lot of frustration trying to understand each other's work and objectives, she explained.
Ian Rowlands, vice president of product marketing at ASG technology, an enterprise information management tools provider, said the three key components to good data storytelling include data science, visualization and narrative.
Data science draws together discovering useful data, preparing it for analysis and building models to understand meaningful relationships. Visualization represents data in ways that users can easily comprehend. "But visualization doesn't complete the story," Rowlands said. Data storytellers need to explain "what the data means to us."
Narrative provides the context that is key to understanding the value of data. "It might be the most difficult part of data storytelling," Rowlands said. Good narrative requires establishing credibility, reasoning convincingly and connecting emotionally with the audience.
"The single most effective way to become a better storyteller is to tell stories," Rowlands said. A story has structure that includes setting the scene, telling the tale and drawing a conclusion. Its key to get to a conclusion or recommendation. It's also important to recognize the risks and alternatives.
You should also plan for objections so you can meet them in the story itself without getting dragged off track. A good practice is to rehearse with a knowledgeable audience that can challenge you and tease out questions you may have to deal with during the actual presentation.
Concisely articulate the story
"In any scenario, it is essential to know whom you will be presenting to and what they care about," said Flynn. This is especially important when communicating with data because you want to tailor your story to highlight the elements of the analysis that are most important to your audience. Will they want to know about the technical details of the analysis or the key findings? Will it be helpful for them to have a clear understanding of the data behind the analysis, or should more time be spent linking the findings to a call to action?
Once this is sorted out, think about how to concisely articulate the central idea of the story that is important for your audience.
"If the audience doesn't understand the central idea and why it is important to them from the beginning, they will check out," Flynn said. When you are crystal clear about the central point of your story, it is easier to weave that idea through the more detailed story you share with your audience.
You can improve these data science storytelling skills by practicing and seeking feedback from others on ways you can improve. Taking an improv comedy class can also help you practice being able to adapt your story on the fly.
Connect the question to the visualization
Kislaya Prasad, research professor in the online Master of Science in Business Analytics program at the University of Maryland Robert H. Smith School of Business, suggests that you should learn how to quickly articulate a question that will act as a hook to grab your intended audience. This should be connected to a clear and focused arc to the narrative. One thing needs to follow another in a related manner and be connected to a sharp conclusion.
It is well worth it to go back to the timeless classics like fairy tales and children's stories to master this. Prasad said, "Anyone who has sat with a child to read a story knows that our minds are attuned to stories and narratives."
Kislaya PrasadResearch professor, University of Maryland Robert H. Smith School of Business
In any data analysis project, the data has a story to tell. "The data exploration and discovery are where we try and figure out the story, and in the presentation of results we tell the story to others," he said. Mastering data visualization techniques can help. It's important to study good visualization and ask yourself what is bad about visualizations that are not great.
Cultivate emotional appeal
Things like emotional appeal, drawing your audience in and engaging them, and personal appeals apply to data science the same way they do to any other type of storytelling, said Angel Durr, founder of DataReady, a data science training program. TED talks are excellent examples of this in action. "I love how quickly the speakers draw you in and keep you engaged in a short amount of time yet still relay an important idea that people connect with on a personal level," Durr said.
She recommends people practice storytelling in general to get better at data storytelling. A couple of good places to start are by volunteering at a local elementary school or reading to the elderly. "If you can read a story to an audience like this and keep them engaged and interested, you are not only helping your local community but also developing the types of skills you will need to articulate complex data ideas to an audience with limited data knowledge," Durr said.
USD's Flynn said her favorite example of data science storytelling was Hans Rosling's Ted Talk on The Best Stats You've Ever Seen. He uses a sophisticated, interactive data visualization based on over 50 years of UN demographic data to illustrate how life span and family size have evolved in countries around the world. "At face value, this could be a very technical and dry presentation, but he uses his talented storytelling skills to powerfully illustrate how our preconceived ideas about the world bias our understanding of different countries and regions," she said.