Storytelling with Data

In my previous role as an HR Analyst I found that it was easy to get lost in the data and come away with a lot of visuals and very little insight. Often, when there was insight, it got lost in the noise of the surrounding data. It is for this reason that data visualization and effective storytelling are essential skills for data anlysts.

Two essential readings on these topics are Storytelling with Data by Cole Nussbaumer Knaflic and The Visual Display of Quantitative Information by Edward Tufte. Where Tufte offered a theoretical and historical approach to data visualization, Knaflic was able to give a holistic approach to the story you want to tell using modern tools.

Here a few key rules and takeaways from my own experience and learnings from these books.

Exploratory vs Explanatory

Don’t let your exploratory analysis fill in as your explanatory analysis.

Exploratory analysis is the process of prodding and reshaping the data to find insight. When I would get access to a new survey or database I would pull it into Power BI and start making graphs. I would try to answer questions I had and find aways to visualize the different dimensions. While this did lead to insight, it also led to a lot of dead ends.

The mistake I made was to then use those same visuals and dashboards to convey the story to others and the insight would inevitably get lost in the noise.

Knaflic compares this to searching for pearls in oysters and instead of just showing the pearls you show the hundreds of empty shells.

The key here is to identify insight and then to start with a fresh canvas and find the best way to convey that insight to your audience.

Outline Your Story

As with any story, it is important to understand your audience and medium then to introduce the plot, conflict, and a call to action.

In evaluating audience Knaflic encourages looking first at overall context. Who are speaking to? What decision do they need to make? Why does what your sharing matter to them in particular?

It is also important to understand the medium that you have to communicate with and how you can uniquely leverage that to convey your message. Is this a white paper, a presentation or a one-on-one conversation? Do you need to grab peoples attention or do you have it already? Each of these should effect the way that you construct the story and your visuals.

Setting the plot should be targeted directly at your audience. What information do they need to understand the problem? What information do they already have that you don’t need to repeat? The beginning of your story needs to give just the right amount of information so they don’t get distracted from the context of the problem you are about to present.

Now that the table is set, it is time to get to the conflict. What is the problem that needs to be solved? This is where you share the couple of shells that yielded pearls; if you show all the empty shells you will loose their attention. Focus spefically on the problems that yeilded insights associated with the action you need them to take.

Finally, call them to action. Don’t have your message end with, “Isn’t that interesting?” Create room for discussion about what is to be done and guide them to one or two things that can be done next to address the problem.

All to often I have shown my data, received some ooohs and aaahs, and then have every one walk away and not doing anything different because of what we saw. This can be fixed by showing the data that leads to action and then making sure the next couple of steps are outlined before you leave.

Three-minute Test

One tool that Knaflic gives, which I found especially useful, is the 3-minute story and it really is as simple as it sounds. Once you have outlined the story the goal is to make sure you have have it so concise that you can tell it in 3 minutes or less. This forces you to get rid of the non-essential and ensure that the the argument flows. Get each element (plot introduction, conflict and call to action) down to their essential elements then start building your final product.

Refine Your Visuals

Once you are have the right insights and the flow of your story it is time to select your visuals. Tufte gives two ratios to use in evaluating your visuals. The first is the data-ink and the second is data density ratio.

The data-ink ration is the amount of data ink divided by total ink used. The goal is to maximize the percentage of ink in the graphic that is being used to convey data. There are a number of tricks that you can use to improve this ratio, such as:

• Multifunctioning graphical elments — use the x and y axis labels to call attention to the points or use the actual numbers in place of dots in a scatterplot
• Get rid of standard chart decoration that does not contribute — this includes grids, 3D (always get rid of 3D!), borders, etc.
• Redundant chart elements — if a line gives the same data that another line gives get rid of it.

The second ratio to evaluate is the data density ratio which looks at the number of entries in a matrix divided by area of graphic. The goal here is to not use a lot of space for very little data. This is one reason that pie charts are always a bad idea — they either give one or two percentages that could be better conveyed in text or are not readable because they do not convey dense data well.

Pull It All Together

Now it is time to pull it all together and ensure that your story flows, your visuals convey only the essential information and your audience leaves knowing what they need to do next.

I highly recommend both these books to get a plethora of great examples of what each of these tips looks like. Both are visual, easy to read and will demonstrate in much better detail a practical guide to implementing these steps.