How to make a good chart: choosing the right type, axes, colour and labels

TUREFF• •0

A good chart answers one question at a glance. A bad one makes the reader work, or worse, leaves them believing something the data does not say. Most of the difference comes from a few decisions you make before you ever open a charting tool.

Start with the question

Write down the one thing the chart should show, as a sentence: "Sales fell after March", "Seoul has more cafés per person than Busan". Then pick the chart that answers that sentence most directly.

Comparing amounts across categories: a bar chart. Horizontal bars work well when the category names are long.
Change over time: a line chart. Time runs left to right.
Parts of a whole: a stacked bar or, for two or three parts, a pie can work (more on that below).
Relationship between two numbers: a scatter plot.
Distribution of values: a histogram or box plot.
Ranking: a sorted bar chart, largest to smallest unless order carries meaning.

If you have several questions, make several charts rather than one chart that tries to answer everything.

Axes: when zero matters

Bar charts should start at zero. The length of the bar is the value, so a bar axis starting at 90 makes 95 look twice as big as 92.5. This is the most common way charts mislead, whether by accident or design.
Line charts do not always need zero. They show change, and zooming in on a range can reveal meaningful movement, such as body temperature or interest rates. Label the axis clearly so readers know it does not start at zero.
Keep scales consistent between charts you place side by side, or readers will compare the wrong things.
Avoid dual axes (two different y-axes on one chart) where you can; it is easy to make any two lines appear related by stretching one scale.

Colour with a purpose

Use colour to highlight, not to decorate. Grey for most of the data and one strong colour for the thing you are talking about is often the clearest chart you can make.
For categories, use a small number of clearly different colours; past six or seven, readers cannot tell them apart. Group the rest into "Other".
For values from low to high, use a sequential palette, light to dark in one hue. For values that diverge from a midpoint (above and below average), use a diverging palette with a neutral middle.
Keep the same colour for the same thing across every chart in a report.

Labels over legends

A legend makes readers look back and forth to match colours to names. Where you can, label lines and bars directly: put the series name at the end of each line, and the value at the end of each bar if exact numbers matter. Other labelling tips:

Give the chart a title that states the finding ("Exports fell 12% in 2025") rather than a description ("Exports by year").
Include units on axes, and state the data source and date beneath the chart.
Remove what does not help: heavy gridlines, borders, background shading and unnecessary decimal places.

Avoid 3D, and use pies sparingly

3D effects distort size: bars at the back look smaller, slices at the front look bigger, and readers cannot judge values accurately. There is almost never a good reason to use them for flat data. Pie charts are hard to read because people judge angles and areas poorly. They can work for two or three parts of a single whole, with labels on the slices. For more slices, or for comparing pies with each other, a bar chart is clearer. Donut charts have the same problem.

Make it accessible

Do not rely on colour alone. Roughly one in twelve men has some form of colour vision deficiency, and red-green pairs are the most common trap. Add labels, patterns or line styles, and test your palette with a colour-blindness simulator.
Ensure enough contrast between text, lines and the background, including in dark mode if your chart will appear there.
Use readable font sizes; axis labels that need a magnifying glass help nobody.
When publishing online, provide alt text or a short summary of the finding, and a link to the underlying data if you can.

A final check

Can someone understand the main point in five seconds?
Would the chart still be honest if the axis started at zero?
Does every colour, label and line earn its place?
Have you shown it to someone else and asked what it says?

Share a chart you are working on, or one that misled you, and ask for feedback in the comments.