A city publishes a 40-page report on transit ridership. Almost nobody reads it. Someone turns the three numbers that matter into a single graphic, and it is shared thousands of times. That is information design: arranging data and processes so a reader understands them faster and more accurately than they would from text. The second half of that sentence matters as much as the first. A graphic that is quick to read but leaves the wrong impression is worse than the report, because people trust pictures and rarely check them.
Gestalt: the eye groups before it reads
Before anyone reads a single label, their visual system has already sorted the page into groups. The Gestalt principles, described by German psychologists in the early twentieth century, name the cues it uses. They matter in an infographic because a reader who groups things wrongly will misread the data even if every number on the page is correct. You use them on purpose: put things together that belong together, and make the grouping visible without a single word.
| Principle | What the eye does | How an infographic uses it |
|---|---|---|
| Proximity | Treats things that are close together as one group | Keep each number right beside its label; leave clear gaps between sections |
| Similarity | Groups things that share colour, shape or size | Give every 'Toronto' data point the same colour in every chart on the page |
| Enclosure | Groups whatever sits inside a shared box or tint | A light panel around a sidebar of related facts |
| Continuity | Follows lines and smooth paths | A connecting line guiding the reader through steps 1 to 5 of a process |
| Closure | Completes shapes that are only partly drawn | Simple icons that suggest a shape with a few strokes |
| Figure–ground | Separates a subject from its background | Enough contrast that the chart is clearly in front of any decorative background |
Icon systems
Icons save space and cross language barriers, but only when they work as a system: a family that looks like it was drawn by one hand to one set of rules. A single mismatched icon, filled when the others are outlined, makes the reader wonder whether it means something different. Consistency is what lets the reader stop looking at the icons and start looking at the information.
- One grid. Draw every icon on the same square grid (for example 24 × 24 units) with the same padding, so they sit at the same optical size.
- One stroke weight and one style. All outlined or all filled, with the same corner radius and line ends.
- One meaning per icon. If a bus icon means 'transit' on one panel, it cannot mean 'school trips' on the next.
- Label anything ambiguous. A floppy disk still means 'save' to many people and nothing at all to others. A short label costs little and removes the guess.
- Test at the smallest size. Detail that looks charming at 200 px turns to mush at 16 px. Simplify until it survives.
Choosing the right chart
| The question the data answers | Chart to use | Why it works |
|---|---|---|
| How do these categories compare? | Bar chart (horizontal if labels are long), sorted by value | Lengths on a shared baseline are the easiest visual quantity to judge |
| How did this change over time? | Line chart | The slope of the line shows the rate and direction of change |
| What share of the whole is each part? | Pie (five slices or fewer), stacked bar, or a 10 × 10 waffle grid | Each shows parts adding to 100%; the waffle makes percentages countable |
| Are two variables related? | Scatter plot | Each point is one case, so clusters and trends appear on their own |
| How are values spread out? | Histogram | Bars show how many cases fall in each range |
| Where does it happen? | Map shaded by rate (per capita), not by raw count | Raw counts mostly show where people live |
Honest scales: truncated axes and proportional area
A bar chart encodes value as length, so the bars must start at zero. Start the axis at 45 instead, and a result of 52 against 48 is drawn as bars of 7 and 3 units: one looks more than twice the other when the real difference is about 8%. The statistician Edward Tufte called the size of this distortion the lie factor. A line chart is different, because it encodes value as position and slope, so a non-zero baseline is acceptable there if it is clearly labelled. Even so, the reader must be able to see where the axis begins.
The same trap appears with proportional symbols. If one city has four times the population of another and you draw its circle with four times the radius, the circle's area is times larger, and area is what the eye reads. To make the area proportional to the value, scale the radius by the square root of the ratio. For three-dimensional pictograms the error is worse still, because volume grows with the cube of the height.