If a number on your dashboard would not change what anyone does, it does not belong there. That single rule removes most of what sits on a typical screen.
Crowded dashboards are rarely built carelessly. They accumulate. Each addition seems reasonable, and nothing is ever removed, until the display is complete and useless at the same time.
Start from the decision
For each panel, ask what decision it informs, who makes that decision, and how often. If you cannot answer all three, the panel is decoration.
Working backwards from decisions usually produces a short list. That brevity is the point. A screen with six things on it gets read. A screen with thirty gets glanced at.
Build role specific views
An owner, a supervisor and a coordinator need different things. Trying to serve all three on one screen produces a compromise that serves none of them well.
- The person doing the work needs today queue and anything blocked.
- The person running the team needs where work is stalling and where capacity is short.
- The person running the business needs direction of travel and the few figures that signal trouble early.
Separate views are more work to design and much less work to read.
Show exceptions, not totals
Totals are reassuring and rarely actionable. Exceptions are where attention belongs: items overdue, jobs without an owner, records missing required information, anything stalled longer than it should be.
Pair each exception with an action. A list you can click into and resolve is a tool. A count you can only look at is a worry.
Leading and lagging
Lagging indicators describe what already happened. They matter for accountability, but by the time they move, the decision window has closed.
Leading indicators are the earlier signals: enquiries this week, jobs entering the pipeline, response times, capacity booked ahead. They are less definitive and considerably more useful for deciding anything. Most dashboards are heavy on the former and thin on the latter.
The numbers are only as good as the record
A dashboard is a view over a source of truth. If the underlying records are incomplete, entered inconsistently or updated late, the display will present that unreliability with a confident chart and people will stop trusting it.
When a figure looks wrong, the fix is almost always upstream, in how the data is captured, not in the presentation layer.
A short pruning exercise
- List every panel currently displayed.
- For each one, name the decision it changes and the person who makes it.
- Delete anything without an answer.
- Of what remains, mark which are leading and which are lagging.
- Replace at least one total with an actionable exception list.
- Review again in three months and delete whatever nobody has used.
It is normal for this to reduce a dashboard by half. Nobody misses the removed panels, which is the clearest evidence they were never doing anything.