Product Sense · easy
What makes a chart misleading, and how do you avoid it?
Asked in BI Developer interviews.
Short answer
Truncated axes that exaggerate change, dual axes that imply correlation, pie charts with too many slices, a rate shown as a raw count. Pick the chart for the question: trend to a line, comparison to a bar, part-to-whole to a stacked bar. Label units, start counts at zero, and drop the decoration: the goal is an honest, fast read.
How to answer it
The principle
A chart is misleading when the visual impression differs from what the numbers say. Most of the time this is careless rather than dishonest, which is why knowing the specific patterns matters more than good intentions.
The common patterns
Truncated y-axis a bar chart starting at 90 turns a 2% difference
into an apparently enormous one. Bars encode
length, so they must start at zero. Line charts
encode position and may be truncated, with the
axis clearly labeled
Dual axes two series on different scales, where the crossing
point is chosen by the author. Almost always
better as two charts, or one indexed chart
Cherry-picked window a range chosen to show a trend that reverses on
a longer view
Wrong chart type a pie chart with nine slices; area encoding a
single value, where doubling the radius
quadruples the perceived size
Missing denominator raw counts across groups of different sizes.
More cases in a bigger city is not a rate
Aggregation hiding an overall trend concealing a reversal in every
subgroup, which is Simpson's paradox on a chart
No uncertainty a point estimate drawn as fact, so noise looks
like signal
Rainbow color a spectrum implying an order that does not exist,
or unreadable for the color-blindHow to avoid it
- Bars start at zero, always. Lines may be truncated, and labeled.
- One axis, or two charts. If you must use dual axes, say so on the chart.
- Show the full available time range, or state why the window was chosen.
- Rates, not counts, whenever groups differ in size.
- Show uncertainty on anything estimated: an interval, a band, or the sample size.
- Color carries meaning, and check it in greyscale and for color-blindness.
- Give the chart an honest title. "Revenue up 4%" is a claim; make sure the chart supports it.
The best test
Show it to someone who was not involved and ask what they conclude, without narrating. If their conclusion differs from what the data supports, the chart is misleading even if every number on it is correct. That test finds more problems than any checklist.
Key points
- Misleading means the visual impression and the numbers disagree.
- Bars from zero; lines may be truncated with a clear label.
- Dual axes let the author choose the story. Avoid them.
- Counts across unequal groups need to become rates.
- An estimate without uncertainty invites people to read noise as signal.
FAQ
- Is truncating always wrong? Not for line charts, where small changes matter and position is the encoding. It is wrong for bars.
- What about 3D charts? Avoid. Perspective distorts the comparison, which is the whole job of the chart.
- How do I show a small but important change? Plot the change or the percentage difference directly, rather than distorting the axis of the level.
The 30-second answer
A chart is misleading when the visual impression disagrees with the numbers, and usually that is carelessness rather than intent.
The patterns I watch for are a truncated axis on a bar chart, which turns a two percent gap into a visual chasm, dual axes where the author effectively chooses where the lines cross, a time window picked to show a trend that reverses on a longer view, and raw counts compared across groups of very different sizes when what is needed is a rate. I would add showing an estimate with no uncertainty at all, because that invites people to read noise as signal.
The test I would actually use is showing it to someone uninvolved without explaining it and asking what they conclude. If their conclusion is not what the data supports, the chart is misleading even though every number on it is right.
In the interview
- What they are checking: the axis and dual-axis cases, and a rule for choosing chart types.
- Common mistake: defending the truncated axis as "showing the detail". It shows a detail at the size of a headline.