How many metrics should be on one dashboard?
For an executive view, roughly five to nine — about what a person can hold in their head and compare against expectation in one pass. Operational views can be longer when they are lists of work rather than gauges. The real constraint is not screen space: every metric needs an owner and a known normal range. If you cannot state what normal looks like for a number, it should not be on the screen.
The expected-range test
Take any number on your dashboard and try to say, without looking anything up, what a normal value is for a Tuesday in this season. If you can't, that panel is not being read — it is being glanced at and skipped, because interpreting it costs more effort than the reader has available.
Run this on an existing dashboard and you will usually be able to delete a third of it immediately.
Tier, don't cram
The fix for "but we need all of these" is depth rather than density. Put the headline metrics on the surface, supporting metrics one click down, and record-level detail one click below that.
This also improves the alerting story: a surface metric moves, and the drill-down already contains the decomposition that explains it. Building that hierarchy deliberately is most of the work in a dashboard that gets used.
How to prune one that has already sprawled
- Look at panel-level usage. Whatever nobody expands has already been voted off.
- Ask each viewer for one decision. "Name a decision you made in the last month because of this panel." Silence is data.
- Check for duplicates in disguise. Leads, calls and conversions are often three renderings of one underlying event.
- Delete anything with no owner. If nobody is accountable for the number, nobody is reading it either.
The counterargument, honestly
There is a legitimate case for a wide operational screen — a dispatch board, a queue view, a location grid — where the point is scanning many similar items for outliers. Density is fine when every cell is the same kind of thing and the eye is looking for the odd one out.
What does not work is a mixed screen: three gauges, two trend charts, a table and a funnel, all at different time scales. That is four dashboards stacked on top of each other, and it is why a written brief often outperforms it for daily use.
Layout is part of the meaning
Where a number sits changes how often it is read. The metric that most often triggers action belongs at the top left, at the largest size, because that is where attention lands first and where it stops if nothing looks unusual.
Group panels by decision rather than by data source. The instinct is to put all the Google numbers together and all the field service numbers together, which is convenient to build and useless to read — it forces the viewer to reassemble a question that spans both.
Topics: dashboard design · KPIs · reporting · information design
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.