Anatomy of a Bad Chart

We often learn a lot more from bad examples than from the good ones. So, here is a really bad one (you can find many examples like this in corporate annual reports…). So, what’s wrong here?

What do they want to say?

If you don’t know what to say, you probably should keep your mouth shut. Are we supposed to compare proportions? Are we supposed to compare 2005 to 2006? We are clueless, so is the author. This is what happens when you use charts to illustrate some numbers. A chart is an argument, an answer to a specific question. Don’t start answering without knowing the question, unless you are a professional politician.

Wrong chart

And since you don’t have a question, you can’t frame your answer. You start babbling incoherently, hopping that your amazing graphic design skills can save you. They don’t (and, by the way, templates and canned effects are not substitutes for real skills). So, what is the question about? Trends? Relationships? Proportions? There is always the right chart.

Wrong Title

While there is nothing inherently wrong with descriptive titles (“Sales by Product”), I do prefer the “as-you-can-see” title. If sales are growing, why don’t you say it in your chart title? “Sales are growing” (as you can see). And you can use titles to connect charts and get  a consistent message.

Don’t mess with time

I would call this “misplaced recency”. Everyone knows that, at least in the Western societies, time flows from left to right. If I see 38% on the left and 40% on the right I’ll assume that the slice is growing. But apparently its the other way around. Conventions work. Stick to them if you don’t have a good reason to break them.

DON’T EMPHASIZE EVERYTHING

If you label every single data point in a line chart; if you use primary colors only; if you explode every single slice in a pie chart, then you are yelling at the reader, while you render the chart unreadable.

Too much data

Exercise your editorial judgment when making a chart. You are the expert. You must know what makes the cut and what can safely be removed or grouped.

What is what?

When color-coding and labeling your chart, make sure that each series is clearly identified. That’s not the case here. And only use a legend if you must to. Label the data points in a pie chart and the label the series in a line chart.

Alphabetical sorting? Never!

I wrote about alphabetical sorting recently. Don’t ever do it. The sorting key must come from the data itself. When you sort the data using an external key you are hiding the patterns. A chart is not a table.

Remove decimal places

If you are a precision freak I am sure you hate charts. If you are not, then remove decimal places from axis and data point labels. Remove the percentage sign too (place it near the axis or in a sub-title). Do the same with large numbers (1 instead of 1,000,000).

They look professional, don’t they?

Professional-looking charts? The wow factor? Memorable charts? Don’t make me laugh. Yes, we are attracted to shiny things, but, well, not all that glitters is gold. From the Wikipedia:

Panning for gold often results in finding pyrite, nicknamed fool’s gold, which reflects substantially more light than authentic gold does. Gold in its raw form appears dull and does not glitter.

I rest my case, and I think I’ll copy this paragraph to the Data Visualization article.

This is not about pies…

Tell me the true: you though I was going to write yet another blog post about pie charts and how evil they are, right? Sorry to disappoint you… But I admit: it’s easier to make all these mistakes in one or two pie charts.