Developed by Edward R. Tufte (The Visual Display of Quantitative Information, 1983).
A set of rules for charts and tables (spend ink on data, cut decoration, use small multiples, put words beside numbers) that make a display answer its question.
One revenue chart, cleared one decoration at a time
Eight quarters of revenue from a board deck, drawn with everything the deck had on it. Slide each part that carries no data from 1 (on) to 0 (off) and watch the same numbers get easier to read, and the share of ink spent on data rise. The last switch starts the axis at zero: the truncated axis made the two years' growth look eight times bigger than it was. Illustrative.
- Decorations removed
- 0 of 5
- Data-ink share
- 22%
An example
A finance team's board deck opened with a revenue chart that drew the same question every quarter: how fast are we growing? The chart had heavy gridlines, a gray background inside a border, a boxed legend in the corner, and an axis with ten ticks that started at $4.0 million, so two years of modest growth looked like a rocket.
An analyst wrote the question at the top of a blank page and went through the chart one part at a time, asking of each whether it carried data. The bars stayed. The gridlines went, the legend became values written on the bars, the border and background went, and the axis kept four round ticks. Last, she started the axis at zero, and the growth shrank to its true size: 22 percent over two years. Then she split the company total into four small charts, one per region, on the same scale. At the next meeting nobody asked for a reading; a director pointed straight at the region that had fallen behind.
Why it works
A reader looking at a chart is hunting for the answer to one question. Every mark that carries no data is something to look past on the way, and decoration such as 3D effects can distort the data it frames. Judging each mark by whether it encodes data gives a test anyone can apply, and the chart that survives the test shows its answer at once.
The other rules follow from the same idea. Small multiples, a row of small charts on the same scale, let the eye compare many series without untangling overlapping lines. Words written next to the data save the trip to a legend. Each rule shortens the distance between the reader's question and the numbers that answer it.
The thinking behind it
Edward R. Tufte Edward R. Tufte, a Yale statistician and political scientist, set out his principles for displaying data in The Visual Display of Quantitative Information (1983), which he published himself through Graphics Press. Two of its terms carry the method: the data-ink ratio, the share of a graphic's ink that presents data, which a good display pushes as high as it sensibly can; and chartjunk, his word for decoration that carries no information.
How to run it
- Write the one question. Put the question the display must answer at the top of the page before choosing a chart.
- Pick the form that fits it. Use a line for change over time, bars for comparing amounts, and a table when readers need exact values.
- Draw a rough version. Plot the real numbers plainly first, with no styling.
- Audit every mark. Ask of each element whether it carries data; keep what does, lighten what helps, and cut what does neither.
- Label the data directly. Write series names and key values next to the lines or bars, and drop the legend.
- Split crowded charts into small multiples. Give each series its own small chart, on one shared scale, in a row the eye can scan.
- Test it on someone new. Show it without explanation and ask them to answer the question aloud.
Where it goes wrong
The audit cuts what the reader needed. Stripping axis values or the comparison series leaves a clean chart that answers nothing. Ask whether each cut makes the question harder to answer, and put back what does.
The chart answers a different question. A beautifully minimal chart of the wrong measure still fails. If readers keep asking for something the chart does not show, go back to the question before touching the ink.
The scales lie. A truncated axis or bars of different widths can exaggerate a change several times over. Check that the size of each mark follows the size of the number it stands for.
When to use something else
- Pyramid Principle. When the problem is the order of the argument around the charts, and the reader needs the conclusion first.
- KPI Design. When the trouble is which measures a dashboard shows, before any question of how they are drawn.
- Back of the Napkin. When the idea itself needs a quick sketch to be understood, and no data is being displayed yet.
Where it sits in SOLVE eX
Go to the source
- The Visual Display of Quantitative Information, 1983