For founders
Dashboards people actually read
Stephen Few defined a dashboard as a single-screen display of the information someone needs to do their job. Most dashboards fail that test on presentation, not data. Five rules from designing them, and the perception research behind each one.
Prashant Kumar Singh · · 4 min read
In short
- A dashboard is one screen that shows what a person needs to do their job; if it needs scrolling or explaining, it is a report.
- Number first, chart second. Bars and lines, not pies or gauges, because length and position are judged most accurately.
- One colour means "act now" and nothing else uses it. Every number shows its target and its trend.
Stephen Few's definition has held up for two decades: a dashboard is a visual display of the most important information needed to achieve one or more objectives, consolidated on a single screen so it can be monitored at a glance. His essay on why most dashboards fail makes the point that the failures are almost never about the data. They are about presentation: the wrong chart, too many colours, no context, no priority.
I have designed dashboards for admin consoles, sales teams and security analysts. These are the five rules that survived all of them.
1. Number first, chart second
The question a dashboard answers is "how are we doing", and the answer is a number. Show it large, with a small label underneath. The chart next to it shows how the number got there. A dashboard that leads with charts makes the reader do the arithmetic the dashboard was supposed to do for them.
2. Bars and lines, not pies and gauges
Nielsen Norman Group's work on preattentive attributes summarises the perception research: people judge length and position accurately, and judge angle and area poorly. A bar chart uses length. A line chart uses position. A pie chart uses angle and area, which is why nobody can tell a 24% slice from a 28% one. Gauges waste most of their pixels on the dial and show one number that would be clearer as text.
3. One colour for "act now"
Colour is the strongest preattentive signal on the screen, which is exactly why it has to be rationed. Everything in a normal state is grey or a single neutral. The one accent colour means something needs attention, and it appears only when that is true. A dashboard with six colours has no way to shout.
4. Context: target and trend
A number alone means nothing. 4,210 sign-ups this month is good or bad depending on last month and on the plan. Every headline number gets its target, or last period, beside it, and a small sparkline showing the direction. Few's bullet graph is the compact form of this and still the best one.
5. One screen, one job
If a dashboard needs scrolling, it is a report. If different people need different numbers, they need different dashboards, not one with tabs. The test is the five-second rule: can the person it is for look at it and know whether anything needs their action today? If not, remove things until they can.
| Instead of | Use | Because |
|---|---|---|
| Pie chart | Horizontal bar chart | Length is judged accurately; angle is not |
| Gauge | Number with target and sparkline | Same information in a tenth of the space |
| Six series colours | Grey plus one accent | The accent can mean "act now" |
| A number | Number, target, trend | A number without context is not information |
| Tabs | Separate dashboards per role | One screen, one job |
A note on the dense case
Some dashboards are for specialists who compare hundreds of rows: analysts, operators, reviewers. The rules still hold, but "one screen" means a dense table rather than a few big numbers. Density is not the enemy; decoration is. A dense table with fixed columns, one accent colour and no chart junk is a dashboard. A spacious card layout that shows twelve of a thousand alerts is not.
Questions people ask
- What is the definition of a dashboard?
- Stephen Few's: a visual display of the most important information needed to achieve objectives, consolidated on a single screen so it can be monitored at a glance. If it needs scrolling or explanation, it is a report.
- Why are pie charts bad for dashboards?
- Pie charts encode values as angle and area, which people judge poorly. Bar charts encode length and line charts encode position, both of which people judge accurately, per the perception research NN/g summarises.
- How many colours should a dashboard use?
- One accent colour reserved for items that need action, with everything else in grey or a single neutral. More colours remove the dashboard's ability to signal urgency.
If you're a founder
If your product has a dashboard, send a screenshot for the free teardown. Most of the fixes are removals, which makes them cheap.
If you run an AI lab or data team
The weekly programme report follows the same rules: the acceptance number first, its trend, the rejection causes as a bar list, and one colour for the thing that needs a decision.
Sources