While nonprofit's continue to invest in business intelligence (BI) and analytics tools, they aren't necessarily getting the information they need to improve business decision-making. Data visualizations serve two purposes sense making and communication by transforming complex information into something easier to understand.
However, two people can interpret the same data visualization differently. Notably, data visualizations tend to answer "what" questions, but they don't tend to explain the "why," or provide other contextual information. Data storytelling does exactly that.
"Data storytelling weaves data and visualizations into a narrative tailored to a specific audience in order to convey credibility in the analytical approach, confidence in the results, and a compelling set of insights that is actionable to the audience." said Ryan Fuller, general manager at Microsoft and former CEO and cofounder of enterprise analytics company VoloMetrix, in an interview. "The narrative is the key vehicle to convey insights, and the visualizations are important proof points to back up the narrative."
"One of the biggest mistakes is trying to fit the data to the story, which often results in a jumbled narrative that doesn't arrive at a compelling conclusion," said Francois Ajenstat, VP of product development at BI and analytics solution provider Tableau, in an interview. "Always start with the data, then build your story around it, rather than vice versa."
After speaking with experts in data science and analytics, we've developed the following four tips to help guide your data storytelling.
1. General Storytelling Rules Apply
Effective data storytelling is a lot like storytelling in general. The data story should have a beginning, middle, and an end. It should also include a thesis (or a hypothesis), supporting facts (data), a logical structure, and a compelling presentation. Yet, all too often, those responsible for analyzing data are unable to present it in a way that's meaningful to the audience.
"A common mistake is spending too much time on the technical aspect or methodology and not providing much creativity in pointing out how the data can help the business," said David Liebskind, VP of analytics at consumer financial services company Synchrony Financial, in an interview. "While data visualization tools are effective, the human element to provide context, interpret results, and articulate insights and opportunities is a critical factor to influence key stakeholders and generate dialogue to drive strategic decisions."
"Great storytelling should reveal truths which are hidden and not easy to interpret from just reading or browsing the data or through simply plotting," said Vivian Zhang, founder and CTO of NYC Data Science Academy.
2. Consider the Audience
Sound data analysis starts with a hypothesis, but incorrect assumptions are sometimes made about how analytical results should be presented. One common mistake is to build a one-size-fits-all presentation that doesn't align very well with the needs of any particular audience.
"Knowing your audience is key," said Byrne Hobart, lead Internet analyst at data intelligence company 7Park Data, in an interview. "Too often, the data will make sense to those who put reports together but not to those who might actually read them. A good rule is to have someone outside the organization read it and explain what it means. If they interpret it correctly, you're on a good path."
One reason to tell data stories, rather than using traditional data visualizations, is to ease and expedite the decision-making process.
"Data storytelling is important because everyone is competing for time and attention with executives," said Synchrony Financial's Liebskind. "Therefore, it is essential to understand your audience and synthesize complex data into a meaningful and compelling story that can be [acted] upon in order to drive strategic decisions and guide business strategy."
Like data analytics and data visualizations, data storytelling may lack a connection to business outcomes. When it misses on this point, it may be informative, but not necessarily actionable.
Different people have different opinions about who should be responsible for creating data stories. After all, the best analytical minds aren't necessarily the best storytellers, and the best storytellers aren't necessarily data scientists or business analysts.
Martin Brown, general manager of digital marketing consulting and software development firm FM Outsource, said in an interview that he often has data scientists, business analysts, and marketers collaborating on a story. In doing so, he often finds there are three versions of the same story.
"Ideally, it would be a data scientist with a flair for articulate and emotional evocation. However, I am still looking for this elusive person," said Brown.
Since unicorn data scientists are so rare, some organizations are combining different types of expertise. As a result, team members may include IT staff, data scientists, analysts, marketers, and those with other roles as appropriate.
"[Data storytelling] is definitely an interdisciplinary activity," said eBay's Karu. "Data scientists are needed to extract patterns in the data, visualization experts are needed to convey the message in a compelling easy-to-understand manner, marketing [needs to be included] to understand the needs of and reach the desired target audience, business domain experience is necessary to home in on the right set of questions, and an editorial staff is needed to communicate the surrounding text in a compelling way."
4. Avoid Distractions
Good data stories include enough information to state a case, but not so much information that the audience struggles to understand the point.
"Data stories should address a specific goal and rely only on data and findings that support that goal," said Microsoft's Fuller. "Data storytellers should avoid clouding their story with findings that don't directly address the objective of the analysis. Don't distract your audience -- keep your story clear, simple, and impactful.
One criticism of ineffective data stories is a failure to get to the point fast enough. Far too much time is spent on explaining what went into the analysis, which seems justified, since so much time was spent on it behind the scenes. However, the effort itself and the explanation of it need to be weighted differently.
"Most data science follows an iceberg rule: About 10% of the work gets presented, and the other 90% supports it, so it's critical for data scientists to wrap a narrative around their data," said 7Park Data's Hobart. "Complex charts and graphs that don't provide context aren't helpful. Draw out the most important points and use data to back it up, rather than unloading lots of data onto the reader."