Misleading visuals and how to spot them

Why this matters

A chart can be built from accurate data and still mislead. A few design choices, often made by default settings rather than intent, can turn a small change into a dramatic one, invent a relationship, or hide the context that would change the reader's conclusion. Analysts need to avoid these in their own charts and to spot them in the charts they are shown.

The exam includes recognising misleading visuals as part of creating effective ones. This lesson covers truncated axes, dual axes, 3-D and area effects, cherry-picked time windows, and the habit of checking a chart against its numbers.

The lesson

Truncated and broken axes

A bar chart shows values by bar length, so its value axis should start at zero. Starting it anywhere else, truncating the axis, makes the bars' lengths no longer proportional to their values.

The course's practice orders make the effect easy to measure. Monthly orders in 2025 range from 147 in February to 178 in December: December is about 21% higher. On a bar chart whose axis starts at 140, February's bar is 7 units tall and December's is 38, so December looks more than five times larger. Same data, and a completely different impression.

Line charts are different, because they show change through slope and position rather than length. A line chart's axis need not start at zero, and zooming in to show variation in a stable measure, such as a temperature, is often appropriate. The axis must still be clearly labelled, and zooming should not be used to make noise look like a trend.

A broken axis, with a visible gap in the scale, is sometimes used to fit one very large value alongside small ones. It should be marked unmistakably, and often a separate chart or a log scale is clearer.

Dual axes and invented correlation

A dual-axis chart plots two series against two different vertical scales, one on the left and one on the right. It can be useful for showing two related measures in different units, such as revenue and order count.

It is also one of the easiest charts to manipulate. Because each axis can be scaled independently, almost any two series can be made to appear to rise and fall together, or to cross at a chosen point, just by adjusting the ranges. The visual correlation is a product of the scaling, not of the data.

When a dual-axis chart is used, both axes should be clearly labelled, and the reader should be warned against comparing the heights of the two series. Often two separate charts, stacked and sharing the time axis, show the same information without the risk.

3-D effects, area and perception

3-D effects on bar and pie charts distort values. Perspective makes nearer elements look larger than farther ones, so in a 3-D pie the front slice appears bigger than an identical slice at the back, and the depth of 3-D bars makes their tops hard to read against the axis. 3-D adds no information to data that has only one or two dimensions, and should be avoided.

Area causes a related problem. When a value is shown by the size of a shape, such as an icon or a circle, doubling the value should double the area. A common mistake is to double the icon's height and width, which quadruples its area, so a doubling looks like a fourfold increase. More generally, people judge lengths accurately and areas and volumes poorly, which is why bar charts are usually preferred to pies, bubbles and pictograms for precise comparison.

Cherry-picked time windows

The time range shown on a chart can change its story entirely. A stock that has fallen over five years can be shown rising, by starting the chart at the most recent low. A sales figure can be shown "up 20%" by comparing with a particularly weak month, as with February against March in the practice data, where the difference in month length alone explains a large part of the rise.

Signs of a cherry-picked window include an unusual start date with no stated reason, a comparison with a single exceptional period, and a window that stops just before an unfavourable change. The honest approach is to choose the range the question calls for, show enough history for context, use year-over-year comparisons for seasonal data, and state why the window was chosen.

Related selective choices include showing only some categories, only the regions that improved, for example, and using different time ranges on charts that are meant to be compared.

Checking a chart against the numbers behind it

The most reliable defence against misleading visuals is to check a chart against its data, whether it is your own chart or someone else's.

Useful questions include:

• Where does each axis start, and is the scale linear or logarithmic?
• Do the visual sizes match the numbers? A bar twice as long should mean a value twice as large.
• What is the base? Percentages of what, counts out of how many, per capita or total?
• What time window and which categories are shown, and what was left out?
• Is correlation being implied by placement or scaling, rather than shown?
• What is the source, and how recent is it?

When building charts, reading the underlying numbers out loud beside the chart is a quick test: if the chart gives a noticeably stronger or weaker impression than the numbers, the design is doing something the data does not justify.