1. Truncating a bar-chart baseline
Bar length encodes value, so starting far above zero can make a small difference appear dramatic. If a non-zero baseline is analytically necessary, make the break unmistakable and consider a dot plot instead.
2. Using unequal time intervals as equal spacing
A line suggests continuous movement. Plotting January, February, and December at equal distance misrepresents the time between observations.
3. Hiding the denominator
“Conversion increased to 20%” means little without the starting population, sample size, time window, and definition of conversion.
4. Using a pie for overlapping categories
Pie slices must partition one whole. Select-all-that-apply survey answers are overlapping groups and can sum beyond 100%, so use bars.
5. Encoding meaning only with color
Color-vision differences, grayscale printing, and low-quality projectors can erase the distinction. Keep labels or other redundant cues.
6. Sorting away meaningful order
Value sorting helps ranking, but it damages ordinal scales, process stages, and chronological sequences. Preserve the order that carries meaning.
7. Omitting inconvenient observations
Removing outliers or missing periods may be defensible, but the decision must be explained. Silent filtering changes the claim the chart appears to support.