Start with the question, not the chart
A chart is an answer to a question. Before opening any tool, write the question in plain words: Which category is largest? Is this number rising? How much of the total does each part represent? Do these two things move together? Once the question is clear, the right chart form is usually easy to name, and many decoration decisions disappear.
Designers sometimes begin with a chart style they like and then try to fit the data into it. That tends to produce graphics that look interesting and explain little. Working from the question keeps the focus on the reader.
Matching chart forms to questions
Most everyday questions fall into a few groups.
Comparing categories. Which is bigger, and by how much? Bar charts are usually the clearest choice because the eye compares lengths along a shared baseline very accurately. Horizontal bars work well for long labels. Sorting bars by value, unless the categories have a natural order such as months, makes the ranking obvious.
Showing change over time. A line chart connects points and makes direction and rate of change easy to see. When there are only a few time periods, or when each period matters on its own, columns can work too. Keep time moving left to right.
Showing parts of a whole. A pie or donut chart can work when there are very few slices and the point is a rough share, such as “about half.” With many slices, or when slices are similar in size, a bar chart is generally easier to read. A single stacked bar can also show composition, though only the first segment shares a baseline for easy comparison.
Showing distribution. If the question is how values are spread, a histogram or similar chart shows clusters, gaps and outliers better than an average alone.
Showing relationships. A scatter plot places each item as a point by two measurements. It helps reveal whether two things tend to rise together, though it cannot prove that one causes the other.
Showing exact values. Sometimes a small table is the best choice. If readers need to look up precise numbers, do not force them to estimate from bars.
Scales and baselines
The most common way charts mislead is through the scale.
- Truncated bar baselines. Bars encode value by length, so they should usually start at zero. If a bar chart starts at a higher number, small differences look dramatic. As a hypothetical example, bars showing 98 and 100 on an axis starting at 96 would appear to have heights of 2 and 4, making one look twice as tall as the other when the real difference is small.
- Line charts are more flexible. Because lines show shape rather than length, a non-zero baseline can be reasonable, as long as it is clearly labeled. Even then, a very narrow range can make ordinary fluctuation look like a crisis.
- Uneven intervals. If the time between points is irregular but spaced evenly on the axis, the trend is distorted.
- Dual axes. Two vertical axes in a single chart let a designer make two lines appear to cross or track each other simply by choosing the scales. Use them rarely and label them clearly.
- Area and size. If circles or icons are scaled by diameter instead of area, a value that is twice as large looks about four times as large. Scale by area, or choose a bar chart.
These problems are not always intentional, but readers are affected either way. Misleading graphics and ethics in design are wider topics, and the short rule is: a chart should not make the data look more or less dramatic than it is.
Reduce clutter so the data stands out
Once the form and scale are right, remove anything that does not help.
- Light gridlines or none at all when values are labeled directly.
- No heavy borders, shadows or three-dimensional effects. Perspective distorts comparison.
- Direct labels on lines or bars instead of a distant legend where possible.
- Limited color. Use one neutral color for context and one accent for the thing that matters.
This is the same restraint that applies to minimalist design in general: each element should earn its place.
Use color and emphasis deliberately
Color in charts has jobs: separate categories, highlight a point or show an ordered scale. Do not use many colors just for variety. When a chart has more than a handful of categories, color alone becomes hard to follow, and a different form or a grouped layout may be better.
Check that important differences do not rely only on red versus green, because color blindness affects a meaningful portion of readers. Add labels, patterns or different lightness. Good color contrast against the background also matters for thin lines and small text.
Emphasis should follow the question. If the point is that one region stands out, make that bar the accent color and leave the rest quiet. A clear visual hierarchy within the chart, from title to highlighted data to supporting detail, keeps readers from hunting.
Titles, labels and context
A good chart title states the finding or the topic in plain language. Label axes with units. Note the source of the data and the time period when relevant. If numbers are estimates or the sample is small, say so briefly. Context protects readers from drawing conclusions the data cannot support.
Text in charts should be readable at the size it will appear. Small axis labels that look fine on a large monitor can become unreadable on a phone, so test charts at real size.
Accessibility
Provide a text summary of the main takeaway for people who cannot see the chart, and consider offering the underlying numbers in a table. Avoid conveying meaning only through color. Chart descriptions are part of accessible design, not an optional extra.
Where charts meet infographics
A single clean chart answers one question. When several charts, text and illustrations combine into a larger visual explanation, that moves into infographic design, which adds structure and sequence of its own. If you are just starting, practice with individual charts first: pick a small set of numbers, write the question, try two chart forms and see which one answers it faster.