HomeBlog › Data Visualization Mistakes That Mislead Readers
data vizchartsbest practice

Data Visualization Mistakes That Mislead Readers

By ResearcherLife Academy · May 26, 2026 · 9 min read

Most misleading figures aren't fraud — they're avoidable mistakes. Here are the ten most common ones in scientific charts and exactly how to fix them.

A chart is an argument made of ink. Small choices can exaggerate an effect or hide a weakness without anyone intending to deceive. Reviewers increasingly catch these, so fixing them protects both your credibility and your reader.

1. Truncated y-axis

Starting a bar chart's axis above zero magnifies tiny differences. For bar charts, always start at zero. If small differences matter, use a line chart or dot plot where a non-zero axis is acceptable — and label it clearly.

2. Dual y-axes

Two different scales on left and right let you imply any correlation you like by sliding the axes. Avoid them; use two panels or normalise the data instead.

3. 3D and exploded charts

3D bars and pie slices distort the very quantities they're meant to show — perspective changes apparent size. Keep charts flat and 2D.

4. Pie charts with too many slices

Humans compare angles poorly. Beyond a few categories, a sorted bar chart communicates proportions far better.

5. Rainbow (jet) colormaps

The classic rainbow map creates false boundaries and fails for colourblind readers. Use perceptually uniform maps like Viridis — full reasoning in our colour palette guide.

6. Hiding the data behind summaries

A bar of means hides the distribution. Show the points: use box plots, violin plots, or jittered dot plots so readers see spread and sample size.

7. Overplotting

Thousands of overlapping points become a blob. Use transparency, smaller markers, density (2D histograms), or sampling to reveal structure.

8. Chartjunk

Heavy gridlines, drop shadows, background images, and redundant decoration compete with the data. Maximise the data-to-ink ratio — every mark should carry information.

9. Inconsistent scales across panels

When comparing panels, use the same axis range, or readers will misjudge magnitudes. If you must differ, make it obvious.

10. No indication of uncertainty

A point estimate without error bars, confidence intervals, or n implies false precision. Always show uncertainty and state what the bars represent (SD, SEM, CI).

The honesty test: would the figure tell the same story if you started the axis at zero and showed every data point? If not, rethink it.

Want to build these skills properly?

Our Blender for Scientific Illustration course takes you from zero to publication-ready renders, graphical abstracts, and journal covers.

Explore the Blender Course →