3 Tweaks to Meeker’s Slides

People are usually surprised to hear that I don’t have a problem with Mary Meeker’s huge, yet very unattractive, annual deck of slides.

People come out of the woodwork each time it’s released asking me to make them over. Last year, one Paris designer did just that. But dolling up her data took away some of the meaning. For me, the usefulness of how she collected and ordered the information trumps the need to finesse the design of her charts.

(Read my analysis of her unique but effective "Meeker Method" here.)

Instead of redesigning Meeker's slides, we chose instead to focus on one data visualization problem child: pie charts. Meeker's deck doesn’t have many pie charts, but all of them warranted commentary.

Most information conveyed in a pie chart can be clearer if a bar chart is chosen instead. It’s nearly impossible for the human brain to visually translate the area of a pie chart into accurate data. The length of the arc is the data. Data visualization experts Edward Tufte and Stephen Few feel pie charts should be avoided, always. I’m not as rigorous about it because there are a few times when a pie chart is actually better.

We did not redesign any of her slides, but instead are examining the effectiveness of pie charts.

If a slice of the pie is larger than 50% of the whole, it may be more effective than a bar chart for the reader to see the scale in the proportion.

Slide 180 is trying to convey: the scale of growth in China for on-demand transportation.

Problem: 
There are two problems with this chart. First, the pies on this slide have been scaled from a tiny size to a large size. I assume that was to show that China has grown from small to large but the scale of the circles isn’t accurate data. Secondly, if the point she’s making is growth in China, you can’t even see how tiny it was in Q1:13 so you can visually compare the growth to the current quarter.

Let's look at the data: 
If you take the numbers for China (the red pie slices) that are annotated above each pie (25MM, 750MM, 1.7B and 6.3B) and plot them in a bar chart, you can see the inaccuracy in their sizing.

If the pies had been scaled to an accurate size, the first pie chart on the left would be the size of almost a pixel, and the pie chart on the right would have been much larger. So, the way the pies were scales isn’t accurate data, it is non-data.

More interestingly, if you show the first pie chart and the last pie chart side-by-side at the same size, you can actually see how much the percent of growth China has had. This information was visually lost because the first chart was plotted so tiny.

For kicks, we re-colored the non-China information into shades of gray here since the insight is about a shift in proportions, which is the best use of a pie.

Slide 172 is trying to convey:  On the left, we can see clearly that 31% of WeChat users make purchases in-app. On the right, the message is not so clear.

Problem: 
The chart on the left fulfills the promise of the title. But with the donut chart on the right, it’s hard to visually distinguish who is ahead and who is losing there because the shapes are nearly equal in size.

Let's look at the data:
We took the data in the donut and plotted it in a bar chart.

When you compare the donut slices as a side-by-side bar chart, you can see that the slices aren’t as equal as they seemed. Just like pie charts, donut charts make it difficult to distinguish the nuances of data when processed as area. If there isn’t much variance in the size of the pie slices, you need to use a bar chart so people can process the difference (unless the point you’re trying to make is that each piece is close to equal).

When the slices are similar in size, it’s a diagram, not data. By the way, Duarte Inc. has posted 4,000 diagrams for you to peruse through — many of which you can substitute for donut-type graphics. 

Slide 22 is trying to convey: the amount of growth since 1985 and the transition of that growth to new markets.

Problem: 
The scale of the circles conveys that the market was smaller in 1985 than 2015, but could it be more effective if you could see what's actually in the left pie chart? Some color coding could make it quicker to process. 

Let's look at the data:
The best use for a pie chart is to show large differences in proportions and this set of charts does just that. But, the original charts use so many colors, it’s hard to see what proportions are being compared. Here’s the same chart but with the main points more effectively color-coded: 

First, we made the pies the same size and rely on the label to speak for itself. Shades of the same color will group like-data together. Since we’re comparing “North America, Europe, and Japan” with “China and Emerging” by reducing those groupings into only two sets of colors, you can see that it makes the growth in the China market much more self-evident.

Conclusion:
Some of the charts for this report are generated by Meeker’s team and others are collected from various sources and used in the visual form provided by the source. Either way, whenever tempted to use a pie chart, also try plotting it as a bar chart. Experiment with color to make sure the insights into the data is most self-evident. There were about 10 charts in her deck that took me over 3 minutes to understand what the chart meant and what the take-away was. That’s too long for any readership.

Nancy Duarte is the CEO of Duarte, Inc. ([email protected]) Nancy is the co-author of the recently released Illuminate: Ignite Change Through Speeches, Stories, Ceremonies, and Symbols and the author of Resonate, Slide:ology, and the HBR Guide to Persuasive Presentations. Follow Nancy on Twitter at @nancyduarte, and purchase Illuminate here!