The right chart depends on two things: what type of data you have and what question you're trying to answer. Start with the question.
Step 1: What are you trying to show?
Ask yourself which of these best describes your goal:
- Distribution — "How spread out are my values? What's the shape of the data?" → Histogram, box plot, dot plot, stem and leaf plot
- Comparison — "How do these categories or groups differ?" → Bar chart, Cleveland dot plot, box plot
- Relationship — "Do these two variables move together?" → Scatter plot
- Composition — "What fraction of the total does each part represent?" → Pie chart, donut chart, stacked bar chart
- Flow — "How does a quantity move through stages?" → Sankey diagram
- Change over time — "How did this trend?" → Line graph, area chart
Distribution: showing how values are spread
Use a distribution chart when your data is a list of numeric measurements and you want to show the shape — where values cluster, how wide the spread is, whether there are outliers.
| Chart | Best for | Example |
|---|---|---|
| Histogram | Large datasets (50+ values) | Distribution of 500 exam scores |
| Box plot | Comparing distribution across groups | Test scores for 5 different classes |
| Dot plot | Small datasets (10–100 values) | 30 students' assignment scores |
| Stem and leaf plot | Classroom/hand-drawn, preserves exact values | 25 lab measurements |
Rule of thumb: fewer than 50 values → dot plot or stem and leaf; 50–500 values → histogram; comparing groups → box plot.
Comparison: showing how categories differ
Use a comparison chart when you have named categories and want to show which is higher, lower, or how much they've changed.
| Chart | Best for | Example |
|---|---|---|
| Bar chart | Comparing any number of categories | Sales by product |
| Cleveland dot plot | 15+ categories, or two values per category | Before/after comparison across 20 departments |
Rule of thumb: fewer than 10 categories → bar chart; more than 15 or you're showing a before/after gap → Cleveland dot plot.
Relationship: showing correlation
Use a relationship chart when you have two numeric measurements and want to know if they're related.
| Chart | Best for | Example |
|---|---|---|
| Scatter plot | Any two numeric variables | Height vs weight |
| Bubble chart | Three numeric variables | GDP vs life expectancy vs population |
Rule of thumb: two variables → scatter plot; three variables → bubble chart; time on X with one measurement → line graph.
Composition: showing parts of a whole
Use a composition chart when your values add up to a total and you want to show each part's share.
| Chart | Best for | Example |
|---|---|---|
| Pie chart | 2–5 categories, part-to-whole story | Market share with 3 major players |
| Bar chart (100% stacked) | Comparing composition across groups | Market share by region |
| Donut chart | Same as pie, with center summary | Budget split with total in center |
Rule of thumb: if you have more than 5 categories or the values are close in size, use a bar chart — bar lengths are easier to compare than pie slice angles.
Flow: showing quantities moving through stages
Use a Sankey diagram when you need to show how a total quantity flows through multiple stages, splits, or paths. Classic uses: energy flow, budget allocation, website traffic paths.
The "one chart per question" rule
The most common mistake is combining multiple questions into one chart. A single visualization should answer one clear question. If you find yourself adding a second Y axis, a size encoding, a color encoding, and trend lines to one chart — you probably have two separate charts trying to share the same space.
Quick reference
| Question | Chart |
|---|---|
| How are my values distributed? | Histogram |
| How do groups compare? | Box plot |
| What's the range and median? | Box plot |
| Do two things correlate? | Scatter plot |
| What share does each category have? | Pie chart |
| How did this change over time? | Line graph |
| Which category is highest? | Bar chart |
| How much did each category change? | Cleveland dot plot (dumbbell) |
| How do flows break down? | Sankey diagram |
| What does my small dataset look like? | Dot plot or stem and leaf plot |
FAQ
What's the most common chart mistake?
Using a pie chart when a bar chart would communicate more clearly. Pie charts are only better when you have 2–5 categories and the "share of total" story is the point. For comparisons, rankings, or more than 5 categories, bar charts almost always work better.
When should I use a table instead of a chart?
When exact values matter more than patterns — like a financial statement where readers need to check specific numbers. Charts communicate trends and shapes; tables communicate specific values. If your audience will look up individual cells rather than grasp the overall shape, use a table.
Can I use multiple chart types together?
Yes — combining a scatter plot with a regression line, or a bar chart with a dot indicating the mean, are common and effective combinations. The rule is that each visual element should add information, not just decoration. If removing an element doesn't change what the reader understands, remove it.
References
- Office for National Statistics: Choosing a chart type
- Office for National Statistics: Pie and doughnut charts
How this guide is reviewed
This guide is maintained by the ChartsMakers Editorial Team and reviewed for statistical accuracy, product behavior, and clarity. Sources used for factual checks are listed in the References section. See our editorial policy.