Most bad charts aren't badly drawn — they're the wrong type for the question. There's really just one question to ask: what do you want the reader to see?
The one-question test
- "Which one is biggest / how do items compare?" → Bar chart. Length is the easiest quantity to judge precisely.
- "How does it change over time / in sequence?" → Line chart. Direction and steepness show growth, dips and inflections instantly.
- "What's the split of a whole?" → Pie chart. Use only for parts of one total, ideally 3-6 slices.
When each type lies
Pie charts lie when slices are close
Humans can't compare angles precisely — 28% vs 32% looks identical in a pie but obvious in a bar. If your top slices are within a few points of each other, switch to bars.
Line charts lie without order
Connecting points implies continuity. Categories with no natural order (countries, products) connected by a line suggest a trend that doesn't exist — use bars.
Bar charts lie with a truncated axis
Starting the y-axis at 90 instead of 0 makes a 4% difference look like a doubling. Our tools always start bars at zero for exactly this reason.
Quick reference table
| Your data | Use |
|---|---|
| Sales by product | Bar |
| Revenue by month | Line |
| Traffic by channel | Pie / donut |
| Survey results | Bar |
| Budget breakdown | Donut |
Whatever you pick, the GPT Chart tool switches between all four types on the same data in one click — try your data as each type and keep the one that answers your question fastest. More in chart design tips.