Most dashboards fail for a simple reason. Someone picked charts because they looked impressive, not because they answered a question. The result is a screen of colour that takes effort to read and gives no clear answer.

The reliable way to choose a chart is to start from the question, not the data. This guide pairs common questions with the charts that answer them, and lists the mistakes that make a chart misleading.

Start with the question

Before you draw anything, write the question in one sentence. "How are sales this month compared with last month?" is a question a chart can answer. "Show sales data" is not. If you cannot state the question, you are not ready to choose a chart.

Matching questions to charts

How much, or how many? (a single number). Use a large number with context: the change from last period, a target, or a small trend line next to it. A number alone does not say whether it is good.

How do categories compare? Use a bar chart, sorted from largest to smallest unless the categories have a natural order. Horizontal bars handle long labels better than vertical ones.

How does something change over time? Use a line chart. Time runs left to right, and the eye reads a line as continuity. For a few time periods, such as four quarters, bars are also fine.

What is each part of the whole? Use a stacked bar or a single bar split into parts. A pie chart can work for two or three parts that differ clearly. Beyond five slices it becomes impossible to compare them, and a sorted bar chart is better.

How are values distributed? Use a histogram to show how many items fall into each range, or a box plot if your audience is used to reading them.

Is there a relationship between two measures? Use a scatter plot, one dot per item. Add labels only to the outliers that matter.

Where is something? Use a map only when location is the point. If you only need to compare regions, a sorted bar chart is easier to read than a map in which large areas dominate.

Do people need exact values? Use a table. Tables are underrated. A well-formatted table with aligned numbers beats a chart when the reader must look up specific figures.

Mistakes that mislead

  • Truncated axes on bar charts. A bar's length stands for its value, so the axis must start at zero. If starting at zero makes the differences invisible, use a different chart, such as a line or a dot plot.
  • Dual axes. Two vertical scales on one chart let you make two unrelated lines appear to move together. Split them into two charts that share a time axis.
  • Three-dimensional effects. They distort proportions and add nothing.
  • Too many colours. If every bar is a different colour, colour carries no meaning. Use one colour, and a second to highlight what matters.
  • Too many lines. Beyond four or five, a line chart becomes spaghetti. Show the few that matter and let people filter the rest.
  • Hidden scales and units. Always state what is measured and in what units.
  • Cherry-picked time ranges. A period that starts at a low point makes any trend look like growth. Choose ranges for a reason and say what it is.

Habits that make charts easier to read

  • Label directly. Put the name next to the line or the bar, instead of making the reader look back and forth to a legend.
  • Write a title that states the finding. "Support tickets fell 18 per cent after the new help centre" tells people what to see. "Support tickets" does not. (The figure is invented, as an example.)
  • Use consistent scales across charts that are meant to be compared.
  • Add context. A target line, last year's value or an annotation marking an event turns a shape into a story.
  • Round sensibly. Show as many digits as the decision needs.
  • Order with intent, by value, by time or by a sequence that people expect.

Make charts work for everyone

In many populations roughly one in twelve men and one in two hundred women have some form of colour vision deficiency, so never rely on colour alone. Differences should also show up in position, labels, patterns or markers. Check that text and key marks meet contrast guidelines, and provide the same information in another form, for example a data table or a short written summary. Tooltips that appear only on hover do not work on touch screens or with a keyboard.

Design the rest of the state, too

A chart is only the happy path. Decide what it shows while loading, when there is no data yet, when the data failed to arrive and when a single value dwarfs all the others. Empty, loading and error states explains how to design these, and the first screen of a dashboard covers which charts deserve the top of the page.

A quick test

Show the chart to someone who has not seen it, for five seconds, then hide it and ask what they saw. If they cannot state the main message, the chart needs a clearer question, a simpler form or a better title.