How to label a chart so nobody has to ask what it means
Most confusing charts are not badly drawn, they are badly labelled. Three habits fix nearly all of it.
Direct labels beat legends
A legend makes the eye bounce between the key and the data, and every bounce costs comprehension. Put the series name at the end of the line, on the bar, or next to the node. Legends only earn their place when there are too many series to label directly — which is usually a sign of too many series.
Units go in the axis title, not every tick
Repeating a euro sign on every gridline is noise. Write Revenue (€ thousands) once on the axis and let the numbers be numbers.
The title should state the finding
Quarterly revenue by region is a description. Iberia overtook DACH in Q3 is a finding. If the chart exists to make a point, the title should make it and the chart should prove it.
Annotate the anomaly
When a spike has a known cause, say so on the chart. One short note prevents the same question in every meeting where the chart appears.
- Series names on the data, not in a key.
- Units once, in the axis title.
- Title states the conclusion.
- Known anomalies annotated in place.
A good test: show the chart to someone with no context for ten seconds, then hide it and ask what it said. If they answer with the title, the labelling is doing its job.
Frequently asked
Should charts have legends?
Only when direct labelling is impossible. Legends force the reader to hold a colour mapping in memory while reading the data.
What makes a good chart title?
A statement of the finding rather than a description of the data. The axis labels already describe the data.