Storytelling in Analytics, Part 2

Andy Fox Engagement Manager
Read Time: 4 minutes apprx.
analytics color D3 data visualization story telling Tableau

In this 2-part blog series, we reflect on the Opex Analytics Academy Session, Best Practices in Storytelling and Visualization. Part 1 of the series recapped our first 2 themes in effective analytics storytelling–Keep It Simple and Make It Personal–through frameworks like SCQA. This post dives into tips for the final theme, Deliberate on Visualization.

Great visualizations have the power to excite an organization. The challenge: even within the data science community, experts in visualization are viewed as the Data Creatives, artists that were born with the eye for color, shape, pattern, design, and other characteristics that bring data to life. Further, when evaluating visualizations, feedback often sounds like a matter of taste (people have some strong opinions on what makes a nice color palette). Yet, there is hope for those of us who can’t draw more than stick figures; visualization has been studied and can be taught with a few simple tips and tricks.

Two leading, albeit contradictory, viewpoints towards visualization illustrate many best practices in analytics storytelling:

    1. The purpose of visualization is to provide Immediate Insight i


  1. The purpose of visualization is to promote Organizational Interaction ii

In the first, the analyst seeks the perfect visual to communicate exactly the insight they are attempting to show. This viewpoint lends itself well to understanding the design principles that lead to human perception. In the second, the analyst does not spend as much time on one conclusion, but instead builds an interactive portal that promotes meaningful discussion. This is often enabled by business intelligence technology.

Immediate Insight

Edward Tufte, one of the modern-day pioneers of data visualization, has a great philosophy for communicating immediate insight: “Give the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space”. This requires the visualization creator to know and simplify what insight they want to convey; anything not leading to this message is, as Tufte calls it, “chart junk”.

In the example below, we attempt to identify a common problem: missing demand data. During what times of day and for what products are we missing critical data?

As a default, we might show the demand levels by hour in a bar chart. To differentiate products, we might add color. This gives a lot of information of our demand levels, but does knowing every value really answer our question? Also, the color adds confusion; most cultures assume green is a positive KPI and red a negative one; is product 5 any better than product 11?

Next, we will greatly simplify the visualization to only show a yes or no (using green and red) if demand data exists for the product during the hour: no more and no less than the question being asked. By actually taking away information, we get better insight!

Tufte and other visualization experts have many more tips and tricks that can help to edit your visualization to show and see the point more clearly.

Organizational Interaction

In this line of thought, visualization is all about the conversation and interaction it generates. This is the scenario that you may have experienced or requested yourself — “I want to be able to get hands-on with the visual. Double-click. Drill-down. Change the assumptions. What happens when…”. Technology is vastly helping to meet this visualization need, from commercial software like Tableau and Microsoft Power BI to applications built with open-source tools like R and D3.js.

Ever wanted to visualize how products flow through a system? Try a Sankey Diagram in D3’s Gallery.

Tired of seeing or creating the same basic dashboards for reporting business metrics? Check out the Tableau Public Gallery for design ideas and templates.

Finally, the open-source community is constantly contributing new ideas and new platforms for visualization. Articles like this one showing Uber and Lyft usage in San Francisco typically come with a link to the code base for the powerful visualizations the data scientist creates. With resources like these, the ability for a visualization to become highly customized for organizational interaction is growing ever more accessible.

The two approaches towards visualization perhaps each have their own use cases. Regardless, the unifying theme is that visuals are the one of the best tools an analyst can use to convert numbers to the appropriate interpretation and action. Rather than viewing it as an afterthought, thoughtful visualization — being deliberate — supports telling a strong analytics story.