Most chart problems come from a short list of habits. These data visualization dos and don'ts cover the ten that matter most, each with the reason behind it, so you can check a chart in a minute before it goes into a report, a slide or an article. Two live charts below use the same election data to show what one setting can do to the story.
Data visualization dos and don'ts in one paragraph
Pick the chart that answers your question, start bar and column axes at zero, sort and label bars, use colour only when it carries meaning, and put the finding in the title and the source underneath. Don't truncate axes, crowd a pie with slices, rely on colour alone, or leave out the context readers need to check your claim.
The dos
1. Pick the chart for the question
Comparing amounts across categories is a bar chart. Change over time is a line or column chart. Parts of a whole with two to five parts can be a donut. Two measures for the same items is a scatter plot, and a third measure turns it into a bubble chart. Start from the question, then pick the chart; our guide to choosing the right chart for your data walks through the common cases.
2. Start bars and columns at zero
A bar's length is its value, so the axis has to start at zero. The chart below shows the House seats each party won at the last six US general elections, from the Clerk of the House's election statistics. On a zero-based axis the close elections of 2020 to 2024 look close, because they were.
The step by step build of this chart is in how to create a grouped column chart.
3. Sort and label
Unless the categories have a natural order (age groups, months, survey scales), sort bars by value so the ranking is visible at a glance. Then print the value on each bar, so nobody has to read it off the axis. Our bar chart maker does both with two settings.
4. Give colour a job
Use one colour when there is one series. Add a second only to highlight something, or when a series already has a colour readers know, like a party or a brand. A rainbow of bars looks lively and tells the reader nothing. For palettes that work, see how to choose chart colors that do the legend's job.
5. Write the finding in the title and cite the source
"Control of the US House flipped twice in six elections" tells the reader what to look for; "House seats 2014-2024" makes them work it out. Put units and the time period in the subtitle and the source in the footer, so anyone can check the numbers.
The don'ts
1. Don't truncate a bar or column axis
Here is the same data with the axis starting at 180 seats instead of zero. In 2014 the Republican column now looks about eight times taller than the Democratic one, when it had 31% more seats (247 against 188).
Line charts are different: they encode change in position, not length, so a non-zero baseline is fine there as long as it is labelled. More examples of the trick are in the misleading chart tricks that fool you most.
2. Don't crowd a pie chart
Past five or so slices, readers cannot compare angles, and slices that do not add up to a meaningful whole should never be a pie. A sorted bar chart shows the same numbers more precisely, which is the argument in why bar charts still rule.
3. Don't rely on colour alone
Some readers cannot tell red from green, and printouts are often greyscale. The W3C's accessibility guidelines make this a rule: colour must not be the only visual means of conveying information. Add direct labels, shapes or patterns, and check contrast.
4. Don't use two y-axes to suggest a link
Two series on two independent axes can be scaled to look as if they move together or apart. If you need to compare two measures, index both to 100 at a common start, or show them as two charts side by side.
5. Don't hide the numbers behind the chart
A chart with no source, no units and no time period cannot be checked. Neither can one that shows totals where rates matter, such as cases by state without population. Give readers the denominator and the source, and link the data when you can.
A quick checklist before you publish
| Check | Do | Don't |
|---|---|---|
| Chart type | Matches the question | Picked because it looks impressive |
| Axis | Bars and columns start at zero | Truncated to exaggerate a gap |
| Order | Sorted by value or natural order | Left alphabetical by default |
| Labels | Values printed where they help | Readers measure against gridlines |
| Colour | One colour, plus one for emphasis | A rainbow, or red against green |
| Title | States the finding | Only names the metric |
| Source | Named under the chart | Missing |
FAQ
What are the most important data visualization dos and don'ts?
Start bars and columns at zero, choose the chart for the question, sort and label your bars, use colour with purpose, and state the finding and the source. The most common don'ts are truncated axes, crowded pie charts, colour as the only cue and missing context.
Is it ever fine to start a chart axis above zero?
Yes, for line charts and scatter plots, which show position rather than length. For bars and columns, start at zero; if the differences are small, print the values or chart the change instead.
How many colours should one chart use?
As few as the data needs: one for a single series, and about six at most for categories. Beyond that, readers cannot match colours to the legend, so label the series directly or split the chart.
How do I make charts accessible?
Do not rely on colour alone, keep text contrast high, label series directly, and write a title that states the main point so screen-reader users get it from the text. A colourblind-safe palette helps; see our colorblind-friendly palettes guide.
