Information design

Free lesson

What information design is

Lesson 1 of 12 in Information design.

The beauty of data visualization - David McCandless · TED-Ed
Lesson

is the practice of arranging information so somebody can understand it and do something with it. Charts are the famous part. The larger part is timetables, forms, labels, instructions, signage and every document that exists to be used rather than admired. A timetable you have to read three times, or a form you sign without being sure what you agreed to, is information design done badly. Much of this work is done by people who would never call themselves designers, such as the official who lays out a ballot paper or the engineer who writes the manual, and that is one reason so much of it goes wrong.

The test is rather than appearance. A chart that is beautiful and misread has failed, and a plain table that answers the question in four seconds has succeeded, which is an uncomfortable standard for a field that mostly judges itself by how things look. So the first question is always what the reader needs to find out, decide or do. The layout follows from the answer.

Pictures have changed decisions before. In 1858 Florence Nightingale en.wikipedia.org published a diagram of the causes of death in the British army in the Crimean War, drawn as wedges around a centre, and it showed that far more soldiers were dying of preventable disease than of their wounds. She made charts like it for members of Parliament and civil servants who were unlikely to read a table of figures. She was using a picture to win an argument about sanitation.

In Vienna from 1925, the philosopher Otto Neurath, Marie Reidemeister and the artist Gerd Arntz built Isotype en.wikipedia.org, a system of standard for explaining social and economic facts to the general public. Its central rule still holds. A bigger quantity is shown by more symbols of the same size, never by a bigger symbol, because readers cannot judge area reliably. Reidemeister's job even had a name, the transformer, meaning the person who turned a table of figures into a chart anyone could read.

Every chart is a translation. A number becomes a length, a position, an area or a colour, and people read some of these far more accurately than others, which is why a bar chart usually beats a pie. Edward Tufte's The Visual Display of Quantitative Information, which he published himself in 1983, made a generation of designers argue about this, and its advice to strip out every mark that carries no information is still a good first edit. Tufte coined the word chartjunk for the decoration that crowds a chart, such as heavy grid lines, patterns that shimmer and needless 3D, and the word stuck. Much of the craft is in deciding what to leave out.

The are unusually concrete. In September 2006, at Methodist Hospital in Indianapolis, six babies nbcnews.com were given heparin at 10,000 units per millilitre instead of the 10 units per millilitre they should have had, and three of them died. The vials were the same size, and both labels were blue, one a darker shade than the other. Nobody who designed those labels meant any harm. They made two very different doses look alike, and on a busy ward that was enough. Medicine already uses a typographic fix for a related problem. Drug names that look alike are printed with their differing letters in capitals, as in predniSONE and prednisoLONE, a practice called tall man lettering en.wikipedia.org that the US Food and Drug Administration encourages manufacturers to use.

A working information designer starts from the reader's question and chooses the form last. Sometimes the answer is a chart. Just as often it is a sentence, a sorted table or a clearer label, and the chart is dropped. The GOV.UK design system design-system.service.gov.uk, the pattern library for British government services, starts forms from one question per page, because that helps people understand what they are being asked and concentrate on answering it, and it groups related questions only where research with real users supports doing so. After the form is chosen comes the edit, which removes every word, line and colour the reader does not need, and then the test on a stranger. None of this shows in the finished piece, and that is exactly why it gets skipped.

An ink drawing of a woman tracing a row on a printed timetable with her finger while the designer sits across the table silently taking notes.
Watching a stranger try to use the timetable you designed.

A real case

The butterfly ballot in Palm Beach County

Palm Beach County, Florida · 2000

What they did

For the US presidential election in November 2000, the county's supervisor of elections, Theresa LePore, spread the presidential candidates across two facing pages with a single column of punch holes down the middle. She did it so the names could be printed larger for the county's many elderly voters.

What happened

The holes alternated between names on the left and right pages. Al Gore was listed second on the left, but the second hole down belonged to Pat Buchanan, listed first on the right. Buchanan received 3,407 votes in the county, and a study in the American Political Science Review in 2001 found it the most anomalous result for him of any of the 3,053 US counties where he was on the ballot. It concluded that the ballot led over 2,000 Democratic voters to vote for Buchanan by mistake. George W. Bush won Florida, and with it the presidency, by 537 votes.

What to take from it

The reason behind the layout was kind and sensible, since older eyes need bigger type. What nobody did was put the ballot in front of a few voters and watch where they punched. A layout that makes two choices look like one is a factual error, and the only test that reliably finds it is a stranger with a real task.

Source: American Political Science Review, The Butterfly Did It (2001)

What to do it with

  • ClaudeFree tier, paid above it

    A language model that writes, edits, reads long documents and runs tools, including agents that work inside a codebase.

    Bad at It is agreeable by default, so it will improve a bad idea instead of telling you it is bad. Ask it to argue against you and the quality of the answer changes.

  • FigmaFree tier, paid above it

    The shared file where interface design happens: components, variables, prototypes and the handoff to code.

    Bad at Its generative features make plausible screens rather than considered ones, and a plausible screen is the hardest kind to argue with in a review.

  • InDesignSubscription

    Long documents, grids, parent pages and print-ready output.

    Bad at Its AI makes pictures and words, not layouts, and Generative Expand has handled linked images and colour profiles badly.

The whole tool directory

The words, in plain language

Information design
Arranging information so it can be understood and acted on.
Data visualisation
The subset dealing with numbers. The most discussed and not the largest.
Encoding
Turning a number into something visible, such as a length, a position or a colour.
Pictogram
A simplified picture standing for a thing or a person, as on an airport sign.
Comprehension
Whether the reader ends up with the right idea. The only measure that counts.
Consequence
What happens when they do not. Why this discipline has an ethics.

Why it matters

A badly designed ballot paper can change an election, as the case below shows, and a badly designed label can cause the wrong dose. Both have happened repeatedly. Most information design failures are quieter, a form abandoned halfway, a sign that sends people the wrong way, a chart that leaves a board with the wrong impression, and they are rarely traced back to the design at all.

It is the most reliably neglected design work in most organisations. Nobody is proud of the form, so nobody is assigned to it, and it is frequently the single document every customer has to get through. The typical mistakes follow from that neglect. People choose a chart for how it looks, decorate where they should simplify, and write for the expert who built the system when the reader is a stranger in a hurry. Start with the documents your own customers cannot avoid, such as the invoice, the returns form and the instructions in the box, because improving one of those often helps more people than a new campaign would.

Test it with somebody who does not already know the answer. Comprehension cannot be assessed by the person who made it, because they cannot unsee the structure, and five strangers with a specific question will find the failures in twenty minutes. Give them a task, such as finding the next train or the right dose, and watch where they hesitate. Say nothing while they do it. Note every place somebody stops, fix those places, and then test again with different people, because the same five will have learned your layout.

The module works through the tools. It starts with choosing a chart that fits the question, the three forms that do most of the work, and the scales and axes where charts most often mislead. Then come colour used for data, the distortions of every map projection, diagrams that explain how something works, and wayfinding in buildings. It closes on tables, forms and dashboards, and on the point where an accurate chart turns into propaganda.

Elsewhere


Type
term
Period
ongoing

Check what stuck

Elsewhere in the library

Hear about new courses.

Leave an email and we will tell you when something new lands. We do not write often.