How to read a chart without being misled

A chart is an argument made of shapes. Before accepting the argument, ask what the shape leaves out. The questions below take about thirty seconds and catch most of the ways a chart can mislead without any number being wrong.

1. Check the axis

• Do bars start at zero? Bar length encodes the value, so a truncated bar axis exaggerates every difference. A chart showing 51% against 49% with an axis from 48 to 52 can look like a chasm.
• Is the unit stated? “Revenue” without a currency, “growth” without “per cent per month”, “users” without active or registered — the number can be right and the meaning missing.
• Are the tick intervals even? A broken or uneven axis changes the slope and the comparison.

2. Check the time window

The same data can tell opposite stories depending on where the window starts. A “record high” after three bad months is not the same as a record high over ten years. Ask: why does the chart start and end here? What would it look like with one more year on either side?

3. Check the denominator

“Cases doubled” means something different when the base is 10 and when it is 100,000. Percentages without totals hide the size of the thing being described. If a chart shows shares, find the counts; if it shows counts, ask what population they came from.

4. Check what the average hides

Two groups can have the same average and completely different distributions — one tight, one split into two peaks. An average is a summary, not a picture. When the question is about people, orders or outcomes, look for the distribution, not only the mean.

5. Check for area and angle distortion

• Bubbles: area grows faster than radius. A bubble that is twice as wide is four times the area, so size differences look bigger than they are.
• Pies: beyond four or five slices, angles are impossible to compare; similar slices look identical.
• 3D effects: perspective changes lengths and areas in ways that no reader can undo.

6. Check the second axis

A combo chart with two y-axes can make any two series look correlated — the scaling of each axis is a choice. If the chart shows two axes, read both scales and ask whether the relationship would survive a different scaling. If it would not, the relationship is in the chart, not in the data.

7. Check the comparison

Is there a baseline? Before-and-after charts often omit the “before”. Is there a control group? A campaign chart that shows sales rising after launch does not show what sales would have done without the campaign. A chart of the winners usually leaves out everyone who did not win.

8. Correlation is not cause

Two measures can move together because one causes the other, because a third factor drives both, or by coincidence. A scatter plot can show association; it cannot show which explanation is true. The chart should say “moves with”, not “drives”.

A thirty-second checklist

1. Where does the axis start, and what are the units?
2. Why does the time window start and end where it does?
3. What is the denominator behind each percentage?
4. Is the average hiding a split distribution?
5. Are sizes shown by length, area or angle — and is it honest?
6. Are there two axes, and is the relationship real?
7. What comparison or control is missing?

When you are the one making the chart

The same questions work in reverse. Choose the axis deliberately, state the units, show the denominator, and if you highlight a comparison, show what it is being compared with. A chart that survives these questions is also the chart that survives a review.

The chart-choice guide covers picking the right type; the colour guide covers making the series readable. If you want to apply the checklist to your own data, paste it into a chart maker and look at the result with the questions above in hand.