Dashboard design that people actually use
Most dashboards are built to show everything and end up showing nothing. The design decisions that turn a data screen into something people open on purpose.
The most common dashboard failure is not ugliness. It is a screen full of accurate, well-rendered charts that nobody looks at, because none of them lead to a decision.
Start with the decision, not the data
Before any chart exists, answer: what will someone do differently after looking at this?
If the answer is "be informed", the dashboard will not get used. Real answers sound like: "decide whether to chase this account", "spot which region is behind and call the manager", "know if last night's job failed".
Every element should serve a decision. Anything that does not is decoration with a maintenance cost.
One screen, one audience
An executive wants three numbers and a trend. An operations manager wants a queue and exceptions. An analyst wants filters and raw export.
Trying to serve all three produces a screen that serves none. Build separate views. They can share components and data; they should not share layout.
The inverted pyramid
Structure by decreasing urgency:
- Status. Is anything wrong right now? Usually one to three items, prominent, and absent when everything is fine.
- Headline metrics. Three to five numbers, each with a comparison. A number without context is trivia — "£48,200" means nothing; "£48,200, up 12% on last month" means something.
- Trends. Where things are heading.
- Detail. Tables, breakdowns, and filters, below the fold, for people who want to dig.
Most dashboards invert this and lead with a dense table.
Design rules that consistently help
Show comparison by default. Versus last period, versus target, versus the same period last year. Pick one and apply it consistently.
Make the units unmistakable. Currency symbols, percentages, and clear time windows. "Revenue: 48.2" causes meetings.
State the freshness. "Updated 4 minutes ago" prevents an entire category of misunderstanding.
Design the empty state. A new account's first view is empty — one of the floor features every SaaS needs before it can charge anyone. That is your onboarding moment, and it is usually left as a blank grid.
Design the error state. When a data source fails, say so on the affected tile. Silently showing stale numbers is worse than showing nothing.
Use colour to mean something. If red means "bad", it cannot also be your brand accent on a neutral chart. Pick one semantic system and hold it.
Cap the chart count. More than six or seven per screen and people stop reading them individually.
Interaction that earns its complexity
- Filters that persist. Nobody wants to reselect their region every visit.
- Drill-down on everything. A number you cannot click to explain is a number people distrust.
- Export. Someone will always need it in a spreadsheet. Fighting this wastes goodwill.
- Shareable URLs. A filtered view someone can paste into a message is far more useful than a screenshot.
Performance is a design feature
A dashboard that takes eight seconds gets checked less often, which defeats the point. Practical measures:
- Pre-aggregate rather than computing from raw rows on every load — and keep per-tenant query caps in place so one customer cannot slow everyone down.
- Load the headline numbers first and stream the rest.
- Cache aggressively and show the freshness timestamp so caching is honest.
- Paginate or virtualise long tables.
How to tell if it works
Instrument the dashboard itself. Which tiles get looked at, which filters get used, which views get opened weekly versus never.
Then remove what nobody uses. A dashboard that lost half its tiles and kept all its usage is a better dashboard, and the removal is the hardest part to get agreement on.
If that dashboard is an internal tool other companies keep asking about, there is a separate decision to make.
If you have a dashboard people ignore, we usually start by watching five people use it for ten minutes each. The findings are rarely subtle.
