Axis Truncation Ethics: When Cutting the Y-Axis Is OK
Truncating a y-axis is misleading on a bar chart, because bar length encodes magnitude and a non-zero baseline exaggerates every difference — but it's often necessary on a line chart tracking small, meaningful fluctuations around a large baseline, like body temperature or interest rates. The rule depends on the chart type, not a blanket "never truncate."
Quick answer: Never truncate the y-axis on a bar chart — bar length is the visual encoding, and a non-zero baseline distorts the ratio the reader perceives. Truncating a line chart's axis is defensible when the meaningful variation is small relative to the baseline (e.g. a metric that moves between 98% and 99.5%), as long as the axis range and starting value are clearly labeled so the reader isn't misled about scale.
Why is truncating a bar chart's y-axis considered misleading?
Bar charts encode value as length from a zero baseline, so viewers judge magnitude by comparing bar heights directly; cut the baseline to, say, 80 instead of 0, and a bar that's actually 5% taller than another can look 300% taller. This is one of the most reliable ways to make a small difference look dramatic, whether intentional or not.
| Chart type | Truncate the axis? | Why |
|---|---|---|
| Bar / column chart | No | Length encodes magnitude directly |
| Line chart | Sometimes, if labeled | Position encodes trend, not a length ratio |
| Scatter plot | Rarely | Distorts the visual relationship between points |
| Area chart | No | Filled area behaves like a bar's length |
When is truncating a line chart's axis actually defensible?
When the metric naturally moves in a narrow band far from zero and the small movement is exactly what matters — server uptime between 99.9% and 99.99%, or a thermostat tracking 68-72°F. Starting that axis at zero would flatten the line into an uninformative flat streak, hiding the very variation the chart exists to show.
How do I truncate an axis without misleading anyone?
Label the axis start and end values explicitly, and consider adding a visual break marker (a jagged line or shaded band) at the base of the axis to signal "this doesn't start at zero." Some tools do this automatically for line charts; if yours doesn't, add the note in the subtitle: "Y-axis starts at 95% to show detail."
Bad: Bar chart, y-axis 80-100, no label on the axis break
Good: Line chart, y-axis 95-100, axis explicitly labeled,
subtitle notes "Axis starts at 95%, not 0%"What's the alternative if I need to show a small change honestly?
Show the zero-based version and add a second, zoomed panel next to it, or annotate the exact percentage change directly on the chart instead of relying on visual slope. A callout like "+2.3 points" next to a zero-based, nearly-flat line is more honest and often more persuasive than a truncated axis, because the reader can trust the number without needing to trust the axis choice.
Common mistakes
- Truncating a bar chart and adding no label, hoping the distortion goes unnoticed.
- Using different y-axis ranges across two charts meant to be compared side by side.
- Truncating just enough to make a change look statistically meaningful when it isn't — see confidence intervals before claiming a small movement is real.
- Auto-scaled axes in dashboard tools that silently truncate on every refresh as the data range shifts.
Pro Tip: If you're unsure whether a truncated axis is honest, build the zero-baseline version first and look at it. If the pattern you want to highlight is still visible at zero, use the zero baseline — truncation is a last resort for genuinely small-magnitude, high-precision metrics, not a default styling choice.
← Back to Visualization Tips