Intelligent Enterprise

I'm going to take you on a short stream-of-consciousness tour through a few of the most common and sometimes downright amusing problems that I see in my work as a data presentation consultant. This article is the second of a five-part series on the fundamentals of effective data presentation. If you produce reports that present data or manage people who do, these articles will offer you practical and clear advice.

To make meaningful judgments about data presentation practices that don't work, we must start with clear principles about the purpose of data presentation and what identifies it when it does work. It really all boils down to one thing: communication. Any presentation of data that you prepare — whether in the form of tables or graphs or in some combination like a dashboard — is only successful to the degree that it communicates to your target audience what you intend for it to communicate. Did your message get through? Was the data understood, accurately and efficiently? It's all about the data. As the renowned expert in the visual display of information, Edward Tufte, so simply yet eloquently put it: "Above all else show the data." Having established communication as the desired outcome, any data presentation practice that doesn't communicate effectively is a problem. Now let's start our tour.

Start with a Clear Message

If we begin our tour at the very inception of the data presentation process — the point when someone first begins to determine what to present and how — the first common mistake is complete ignorance of the message that ought to be communicated. Before you can determine how to effectively present a message you must first know what the message is. It isn't enough to know that the message is about sales. What is it about sales that you've discovered in the data and wish to pass on to others? Have you discovered that sales as a measure of revenue in U.S. dollars have steadily declined in the last three months? Have you found that, even though sales have steadily increased year to date, you are well below your plan for the year? Have you noticed that the only reason that sales are on target for the year is because one particular salesperson in Asia is burning up the market?

Before you decide how to present the data, step back for a moment and think carefully about what you want to say. Actually put it into one or more sentences that are as complete and meaningful as necessary to communicate what your audience needs to understand. This isn't a silly ritual. You can find clues in the words you use to express your message that will direct you to present data in a particular way. For example: "Even though we're five percent ahead of our year-to-date revenue plan, four of our five products have been steadily declining in sales since the beginning of the year. That we're ahead of plan is entirely due to the success of a single product: yet 80 percent of the sales force is dedicated to the four products that are declining." This is interesting information that certainly deserves a response. If you created a graph that displayed overall year-to-date revenue compared to plan, this message would be lost. The fact that the message concerns (1) changing sales through time, (2) a contrast between the four declining products and the one increasing product, and (3) a dominant allocation of sales resources to products that are failing are all important points that determine how the data ought to be presented.

To communicate this message effectively, you still have to know about effective graph design, but you haven't got a chance of doing it right if you don't begin with a clear understanding of your message.

Don't Insist on a Graph

If you can communicate your message clearly, efficiently, and with the desired impact in a simple sentence, that's what you ought to do. If your message requires the precision of a table of numbers and text labels to identify what they are, that's what you ought to use. It's a mistake to force your audience to use visual perception to interpret a graph, struggling to figure out the exact values of the data encoded as bars or lines when the message has nothing to do with the shape of the data. Yes, some people are more impressed with graphs even when they're a poor means of communication, but do you really want to produce bad and just plain stupid work?