How do you choose the right chart type for a given analytical question?
Match the chart to the relationship in the data: comparison across categories calls for bars, trends over continuous time call for lines, correlation between two numeric variables calls for a scatter plot, and distribution shape calls for a histogram or box plot. The question you are answering — not aesthetics — drives the choice.
How to think about it
Start with the question, not the chart
The chart type follows from what you want the viewer to perceive:
| Analytical goal | Primary chart |
|---|---|
| Compare magnitudes across discrete groups | Horizontal or vertical bar |
| Show change over ordered time | Line |
| Reveal relationship between two numeric variables | Scatter |
| Show distribution of a single variable | Histogram or box plot |
| Show part-to-whole (only 2–3 parts) | Bar with 100 % stacking, or a single pie |
| Show geographic pattern | Choropleth or dot map |
The four core questions
- Comparison — “Which region had higher sales?” → bar chart, sorted descending.
- Trend — “How did revenue change month over month?” → line chart; the connected line encodes continuity.
- Relationship — “Does ad spend correlate with conversion?” → scatter plot; add a regression line if the direction matters.
- Distribution — “Are response times skewed?” → histogram (shape) or box plot (spread + outliers across groups).
Quick decision flow
Ask three questions: (1) How many variables? (2) Are they categorical or numeric? (3) Do you want to show magnitude, change, relationship, or distribution? Most charts fail not because they are ugly but because they answer a different question than the one on the slide.