Common Pitfalls

How charts lie

Even accurate data can mislead when visualized carelessly — sometimes by accident, sometimes on purpose. Knowing these traps makes you both a better creator and a skeptical reader.

Misleading axes

Truncating the y-axis is the classic trick. Two bars that differ by 2% can look wildly different if the axis starts at 98 instead of 0. Always check the axis range before trusting a dramatic-looking chart.

   Same data, two y-axes:

   [0–100]  → bars look almost identical  (honest)
   [98–100] → bars look dramatically different  (misleading)

If a chart's axis doesn't start at zero (for bar charts), ask why. The choice of scale is itself a claim about what matters.

Cherry-picking data ranges

Showing a stock's rise over a carefully chosen week while ignoring a decade-long decline. Always disclose the full context — the time window, the comparison group, the sample.

Correlation vs causation

Two variables moving together does not mean one causes the other. Ice cream sales and drownings rise together — but both are driven by summer heat. Look for confounders, mechanisms, and (ideally) controlled experiments.

Simpson's paradox

A trend that appears in the whole dataset can reverse within every subgroup. A famous example: a university seemed to admit men at a higher rate than women overall, yet each individual department admitted women at equal or higher rates. The aggregate hid the department-level truth.

Dual-axis dangers

Two y-axes on one chart lets you overlay unrelated metrics and rescale them to "line up" — manufacturing a correlation that isn't there. Use dual axes rarely, and label them exhaustively when you do.