A chart is an argument
A chart is a way of making a point with data. Every chart involves choices: what to include, what scale to use, which period to show and how to label it. Many of those choices are made carelessly rather than dishonestly, but some are made to push a conclusion. Either way, the picture can leave an impression that the numbers themselves do not support.
Reading a chart well means looking at the choices as well as the shape.
Truncated axes
On a bar chart, the length of each bar is meant to represent the size of the value, which only works if the axis starts at zero. When the axis starts higher, small differences look large.
Here is a hypothetical example. Suppose two products have satisfaction scores of 98 and 100 out of 100. The real difference is 2 points, about 2 percent of 98. If the vertical axis starts at 97, the first bar is 1 unit tall (98 minus 97) and the second is 3 units tall (100 minus 97). The second bar looks three times as tall, even though the underlying values differ only slightly.
Line charts are more flexible, since a line shows change rather than total size, and a non-zero starting point can be reasonable when the changes are small compared with the values. Even then, check the numbers on the axis, because a narrow range can make ordinary fluctuations look dramatic.
Cherry-picked ranges
Choosing the start and end dates can change a trend. A series that rises over ten years may show a sharp fall if you only look at the last three months. Similarly, starting just after an unusually high or low point exaggerates the change that follows.
Ask whether the period shown is the natural one for the question or one that happens to support a point. If a chart shows a short window, look for a longer one, and consider whether seasonal patterns could explain what you see.
Dual axes and mismatched scales
Charts that put two different measures on left and right axes can make unrelated things look closely linked, because the scales can be adjusted until the lines overlap. If two lines appear to track each other, check what each axis represents and where each starts. Even a close visual match is a correlation at most, and the points made in correlation causation apply fully.
Missing context
A chart may show a count when a rate would be fairer. A town with more people will usually have more of almost anything, so comparing raw numbers between large and small places can mislead. A percentage might hide a tiny base, such as “up 100 percent” when the number went from 1 to 2. Check what the unit is, what is being compared and whether the figure is adjusted for population, inflation or time.
Missing labels are another problem. If an axis has no title, unit or source, you cannot be sure what is being measured. Reputable charts usually name the source of the data.
Area, size and 3D effects
When a chart uses icons or circles, people tend to judge by area, but creators sometimes scale height and width together. If a symbol is made twice as tall and twice as wide to show a doubling, its area is actually four times larger. For example, doubling both dimensions of a 1 by 1 square gives a 2 by 2 square, area 4 compared with 1. Three-dimensional effects, shadows and tilted perspectives can similarly distort the apparent size of a slice or bar.
Pie charts are also easy to misread when there are many slices, or when the slices do not add up to a whole.
Percentages and proportions
Charts of percentages need a clear base. “Support increased by 10 percent” can mean an increase of 10 percentage points (from 40 percent to 50 percent) or a relative increase of 10 percent (from 40 percent to 44 percent). These are different. If the chart does not make clear which is meant, treat the claim with caution. Practicing with percentage calculations makes this much easier to spot.
Smoothing and selective data
Some charts smooth lines or average data in ways that hide variation. Others leave out outliers or categories without saying so. Look for notes about how data were processed, and for gaps in the series where data are missing.
A checklist for reading a chart
- Read the title, then the axis labels and units.
- Check whether the axis starts at zero, and whether that matters for this type of chart.
- Note the time period and ask what happens just before and after.
- See whether numbers are counts, rates or percentages, and what the base is.
- Look for the source of the data and the date.
- Ask what is not shown that would help you judge.
- Find the numbers themselves if you can, and compare them to the picture.
Learning the principles of graph reading helps here, since you can read values directly instead of relying on the impression. If you create charts yourself, the same standards apply in reverse, as covered in basic charts: label clearly, state the source, choose scales that do not exaggerate and show enough context for readers to judge.
Keeping perspective
Most misleading charts are not clever deceptions. They are often rushed or designed to look dramatic. Even so, reading charts with some skepticism is a sensible default. When a visual seems to prove something striking, find the underlying data and see whether the story still holds.