A dashboard is a tool for answering recurring questions quickly. Most go wrong in the same way: they are built to show what the data has instead of what a person needs, and end up as a wall of charts that nobody opens after the first week.
Start with the person, not the tool
Before opening the BI tool, write down who will use the dashboard, how often, and what they will do after looking at it. A sales manager checking weekly pipeline needs something very different from an engineer watching error rates. When the audience wants "everything", build several focused dashboards, not one crowded one.
Layout: tell people where to look
People read a page from the top left. Put the headline figures there, followed by the charts that explain them, with detail underneath for those who want it. Group related things together and leave space between groups. Every extra chart competes with the important ones for attention.
Context turns numbers into answers
"Revenue: 482,000" tells you nothing. "Revenue: 482,000, 6% above target, 3% below last month" tells you whether to worry. Wherever you show a figure, show what it should be compared with.
Charts that tell the truth
- Bars start at zero. A truncated bar axis makes small differences look huge.
- Lines for time. Points joined by a line suggest a trend, so use them only when the horizontal axis is ordered.
- Pie charts only for a few slices that add to a whole. Past four or five, use bars.
- Avoid dual axes, which invite readers to see a relationship that is created by the choice of scales.
- Skip decoration. 3D effects, shadows and heavy gridlines add ink and no information.
Labels, colour and access
Say what each chart shows in its title, and state the units and the period. Show when the data was last refreshed: stale data presented as current is worse than no data. Use colour consistently (the same colour for the same thing everywhere), keep enough contrast between text and background, and do not rely on colour alone, because some viewers cannot distinguish certain colours. The Web Content Accessibility Guidelines are a good reference for contrast and colour use.
Trust
A dashboard that disagrees with the finance report once will be ignored forever. Before launch, check a few figures against the source for a known past date. Define each metric in one place, and show the definition so that people know what they are looking at. Name an owner who answers questions and fixes errors.
Keep it alive
Dashboards decay. Questions change, data sources change, and tiles stop meaning what they once did. Review each dashboard on a schedule. Look at usage if your tool offers it, ask the people who use it what they would remove, and take down anything that has gone unused.
Checklist
- Name the audience and the decisions the dashboard supports. If you cannot, it is not ready to build.
- Put the most important number or question at the top left, where people look first, and limit the page to the few things that matter.
- Show each metric with context: compared with a target, the previous period or a benchmark. A bare number has no meaning.
- Choose chart types for the question: lines for change over time, bars for comparison, tables for exact values. Avoid pie charts with many slices, 3D effects and dual axes that mislead.
- Label everything: titles that say what is shown, units, the time range and when the data was last refreshed.
- Use colour with a purpose and keep it consistent: one colour per meaning, enough contrast, and never colour alone to carry information (some viewers cannot tell colours apart).
- Make filters few, visible and obvious about their state. A hidden filter that changes every number is a source of wrong conclusions.
- Define each metric once, show the definition on hover or in a notes panel, and reuse it across dashboards.
- Check that it loads fast and works on the devices people use, and that the numbers match the source on a known date.
- Name an owner, schedule a review, and remove the dashboard (or the tiles) that nobody opens.
Frequently asked questions
How many charts should a dashboard have?
As few as answer its questions. A common guide is a handful of key figures and a few supporting charts that fit one screen. If people scroll through dozens of tiles, split it into several dashboards for different questions.
What is the difference between a dashboard and a report?
A dashboard is for monitoring and quick answers at a glance. A report is for exploring, explaining and recording in more depth. Trying to make one do both usually fails at both.
References
From Grokem: https://www.grokem.com/data/business-intelligence/dashboards/dashboard-design-checklist
