Present the Data Story Persuasively to Make the Point
Concise, clear storytelling by data analysts is essential when critical decisions depend on their analyses
As analytics and advanced analytics become more mainstream, being able to tell a story that utilizes analytics becomes an important skill. A story—a narrative describing the pursuit of a goal—can inspire people. A data analytics story—a narrative that includes analysis—can move beyond recounting facts to weave together pieces of analysis that make an impact and spur people to action.
Most people think telling the data analytics story requires getting up in front of an audience and presenting their findings. Analysts often talk about two different kinds of data stories. The first is the traditional data story that is basically a one-time occurrence with the classic beginning, middle, and end. The analysis often forms the middle part of the narrative arc, and the resolution—which is often a call to action—forms the end. This mode of storytelling is valuable.
Telling an accurate story
Then there are the updated living stories that are dynamic and change through time. For instance, the manager of a cosmetics company might track the weekly fluctuation of sales of various products across regions or stores. This tracking may produce a recurring story with data that can be illustrated in a dashboard or on a storyboard that others can access. As new data arrives, the view is updated. This analysis may be shared with others who comment and build an iterative and often interactive story. This form of storytelling is sometimes referred to as modern storytelling. Regardless of whether the story is traditional or modern, there are important principles to consider: pointed, persuasive, and presentable, or the three Ps.
There must be a point to a story. Otherwise, why tell it? The point of the story is often the business problem an organization is trying to solve with the analysis. After—or even during, in the case of a modern story—the analysis takes place, getting the key idea of the analysis across is vital. To tell the story, properly framing the tale is important. What is the point? Why should the audience care?
These questions borrow from some ideas in journalism and advertising. For instance, in a churn analysis the point isn’t that a customer churn analysis was performed. The point is that customer churn is, for example, up 25 percent from last year—from 1,000 churners to 1,250 churners—along with the reason for that increase and how the organization should react. This story can be told as a one-time story or even as an iterative story in which people are collaborating to uncover the cause and determine what to do about it.
The points in the story need to be compelling. Each individual crafts the story. However, while being persuasive is important, not being vague or misleading the audience with the analysis or visuals is equally important. Three key points—ambiguity, misinformation, and the right data—can be considered for the persuasive principle.
Think about ambiguity. Data presented in a data story that is not presented in context can be very frustrating for the audience, and often raises more questions than it answers. In the previous customer churn example, the number of churners was provided rather than simply stating that churn was up 25 percent from the previous year. The audience for the analysis doesn’t know whether 25 percent is good or bad if it’s simply handed a number with no context. Sure, it sounds bad in this case, but what about the case in which an organization says its merchandise sales are up 200 percent, yet it sold only USD5,000 the previous year? Context is essential.
Consider another example in which changes in crime during an 11-year period are tracked (see figure). What can be made of this chart? It raises a number of questions because the chart is ambiguous. It illustrates percentage decreases in certain crimes between 2000 and 2010. At first glance, the observer may think, “Great, crime is decreasing.” But after taking a deeper look, the following questions may quickly come to mind:
- Is the chart showing approximately 10,000 robberies in 2000 or 100,000 robberies?
- Is it representing a cumulative percentage drop year over year?
- What is the source of the data the chart represents?
The ambiguity revealed in these questions can be very misleading to the audience.
Data for specific crimes committed over an 11-year period that portrays ambiguity
Avoid being misleading. Although a goal may be to persuade an audience one way, storytellers should ensure they maintain integrity—which means that if analysis results end up being something other than what was intended, then so be it. The visuals should not be purposefully used to mislead the audience. The following techniques are commonly used for misleading:
- Cropping the axis: Truncating the y-axis is an easy way to make something look more important than it actually is. For instance, if profits rose from 3.2 to 3.5 percent of USD100,000 over the previous year, the increase can be represented to look more significant than it is by expanding the y-axis scale from 3 to 3.5 percent rather than from 0 to 3.5 percent.
- Omitting data: Sometimes people leave out data to make a point. For instance, the storyteller can make something appear to be an ongoing problem that really started only a few years ago by not showing previous years on the graph.
- Showing cumulative values: Storytellers often obscure what is actually happening by running cumulative totals and showing them on a yearly plot instead of showing the actual year-over-year values. As shown in the crime changes chart, the robbery rate dropped 10 percent/year. However, overall, robberies actually dropped 65 percent from 2000, not 100 percent as the chart leads the audience to believe. Likewise, the same distortion shows up for the murder rate, and it’s a bit more confusing because there is a rise in the murder rate shown for 2006.
- Select the right data: Not all of an analysis needs to be factored into the story. However, the storyteller must include the parts of the analysis that highlight the key points of the story.
Graphics provided with the story should help illustrate the key points, not detract from them. Consider the following recommendations when developing visuals*:
- Avoid clutter. Edward Tufte, statistician; professor emeritus of political science, statistics, and computer science at Yale University; and a luminary in the field of data visualization, once said that the ink-to-data ratio should be low on a chart. Avoid distracting the audience with irrelevant data. Highlight the significant themes, but also consider removing backgrounds, borders, and special effects such as 3-D representation. Keeping things simple generally works well.
- Highlight significant findings. In parallel with avoiding clutter, clearly state the point of the graphic. For example, consider using text or a circle to highlight or emphasize a particular piece of information.
- Cite the source. Presenting a chart without referencing the source can be quite irritating to audiences. Always give credit where credit is due. If the visual raises questions, then the audience can consult the source. In addition, knowing how recent the chart data is can be helpful to audiences.
- Avoid extraneous visuals. Although pie charts can be used to illustrate data if they are done right, they can end up being extraneous when they can’t stand on their own. For example, some pie charts simply have numbers on pie slices without any context provided. Also, bar charts can serve little purpose, particularly those showing bars that are all essentially the same length. Although storytellers may want to add visuals to the story, they may not necessarily be useful.
Following guideline principles
Data analysts can keep these three principles in mind when formulating their stories to present analysis findings to the audiences that need to hear and digest the results before making important business decisions or strategic moves. As mentioned previously, simplicity and clarity should guide the way for shaping the story to ensure it has a point, is persuasive by avoiding ambiguity and misleading information, and is presentable by following specific guidelines for developing visuals.
Please share any thoughts or questions in the comments.
* The Data Warehousing Institute (TDWI) has developed classes around these storytelling principles and published some checklists that can be used for shaping storytelling to present analysis findings. In addition, the following links provide brief synopses of sessions available at the upcoming TDWI Las Vegas event February 22–27, 2015 that offer details on aspects of storytelling.
TDWI Data Visualization Fundamentals, presented by Dave Wells, BI consultant, mentor, and teacher, February 22, 2015.
Overcoming Information Overload with Best Practices in Data Visualization, presented by Stephen Brobst and Andrew Cardno, February 23, 2015.
Data Storytelling: The New Horizon in Business Analytics, presented by Ted Cuzzillo and Dave Wells, February 25, 2015.
BI Delivery Formula: Planning, Scoping, and Storyboarding, presented by Mico Yuk, February 26, 2015.