Choose a chart that answers the comparison question
Match the chart to the pattern you need to show, check the scale and baseline, and keep the display focused on one clear message.
About 6 min read · Allow extra time for the practical checks
Choosing a chart can feel like a software question: open a chart menu and pick a familiar option. A better starting point is the comparison the reader needs to understand. Are you showing values across categories, change over time, a distribution, a relationship, or parts of a whole?
The UK Office for National Statistics advises selecting a chart according to the data and the relationship or trend to highlight, using familiar forms where possible, and keeping the display focused on the most important relationship. The Government Analysis Function also recommends deciding the message before the chart and checking that the chosen form fits the statistical relationship. These principles support a selection process; they do not make a chart choice automatic.
State the message before selecting the display
Write one plain sentence describing what the reader should compare. For example: “The median completion time differed across the three task formats in this sample.” This is clearer than “show the results”. If you cannot state the comparison, inspect the question, variables and analysis before designing the figure.
The example is fictional. Its task formats, participants, results and values are invented for illustration and are not findings from a real study.
Match the question to a chart family
On a narrow screen, scroll sideways to compare each question with a possible chart and its selection check.
| Reader’s question | Possible chart | Check before using it |
|---|---|---|
| How do values compare across a few categories? | Bar chart or dot plot | Use a common scale; order categories to support the comparison; start bars at zero unless a clearly justified alternative is required. |
| How did a measure change across ordered time points? | Line chart for a continuous time sequence; columns for separate periods or amounts within intervals | Label the time period and clarify whether values are a stock at a point or a flow during an interval. |
| How are numeric observations distributed? | Box plot or dot plot | Choose a form that shows the feature you need; avoid hiding small groups or important clusters. |
| How do two numeric measures vary together? | Scatterplot | Describe association only; a pattern does not by itself establish cause. |
| How is one total divided among a few components? | Stacked bar, or a pie/doughnut chart when segments are few and legible | Check that categories form the stated whole; if comparing groups or segments, use a common baseline or a simpler display. |
The mapping is a starting point, not a rule that overrides the structure of the data. Some chart forms can show more than one relationship, but a complex display may obscure the comparison. If the figure has competing messages, simplify it or use two clear displays. The ONS guidance specifically warns that multiple relationships can reduce clarity and notes the value of consistent baselines when comparing components.
Work through one fictional comparison
Imagine a teaching team records completion time for three versions of a practice exercise. The question is whether the observed times differ across versions. A bar chart of the mean would hide the spread and could be distorted by unusually long times. A dot plot of individual times, grouped by version, or a set of box plots may make the distribution and differences more visible. The right choice depends on the sample size, distribution and purpose of the figure.
Now suppose the team wants to show monthly counts of completed exercises. A column chart can make separate monthly totals easy to compare. If the data instead record a measure observed continuously over time, a line chart may better show the pattern. State whether the value is a count occurring during each month (a flow) or a total held at month end (a stock); those quantities answer different questions.
These are invented examples. No sample sizes, patterns or results are implied. The examples show how the chart follows the reader’s question and the data’s structure.
Check the scale, baseline and categories
For magnitude comparisons, keep the scale interpretable. Bar lengths encode value from a baseline, so a truncated axis can exaggerate differences. If a nonzero baseline is justified for another chart form, make it visible and explain it. Use consistent intervals and units, and show uncertainty when it matters to interpretation.
For time series, use the same periods and scale across compared series. Avoid connecting values when the intervals or order do not support continuity. For part-to-whole charts, make sure the categories are mutually exclusive and cover the whole you name. Do not use a stack simply because the software offers it: segments away from the baseline are hard to compare precisely.
If the figure is crowded, has many categories, combines unrelated units, or needs several footnotes to be understood, step back. A table may be better for exact values; a chart is most useful when the visual pattern is the point. The separate guide to building a descriptive-statistics table explains how to present the values themselves.
Keep interpretation within what the chart shows
A chart can show that two measures move together or that one group has a higher observed value. It cannot by itself show why the pattern occurred, whether a difference is statistically reliable, or whether the result generalises beyond the data. Keep interpretation tied to the design and analysis. For inferential claims, consult the relevant methods and report assumptions and uncertainty; see the guide to interpreting a confidence interval.
Finish with a reader check
Before using the figure, ask:
- Can I state its main message in one sentence?
- Does the chart make the needed comparison easy to see?
- Are the axes, units, categories and time periods clear?
- Is the baseline appropriate to the form and comparison?
- Does the title describe the figure’s point without claiming more than the data show?
- Can a reader who cannot see the chart access the same important message in text or a data table?
The last question matters because a graphic should have an equivalent text alternative. The Government Analysis Function recommends a written description or accessible table that conveys what the chart is meant to communicate. After selecting and building the display, use the separate guide to write a clear table or figure caption.
The ONS and Government Analysis Function guidance provides a starting point for explanatory charts, not a universal rule. The choice still depends on the variables, sample, analysis and intended reader. The examples here are fictional and do not replace statistical advice or a journal’s figure requirements.
