Almost every organisation has dashboards that were requested enthusiastically and are now opened by nobody. The build was not the problem; they usually work correctly and look reasonable.
They fail because they were designed around available data rather than around a decision.
A dashboard should support one decision
The commissioning question is usually "what would you like to see?", and the honest answer is everything. That produces a wall of charts where nothing is prominent because nothing was prioritised.
The better question is "what decision will you make with this, and how often?". A dashboard that answers one recurring question well gets opened. One that displays forty metrics does not.
Numbers need a comparison
A figure alone is close to meaningless. Revenue of 240,000 is good or bad depending on last month, on the target, and on the same month last year.
Every headline number should carry a comparison, otherwise each viewer supplies their own from memory, and they will not agree.
Show what to do, not just what happened
The most-used dashboards make the next action obvious: which accounts need contact, which orders are stuck, which alerts are unactioned. They are worklists as much as reports.
Purely retrospective dashboards get reviewed in a monthly meeting and ignored the rest of the time, which is a legitimate use but a much smaller one than intended.
Trust decides everything
One wrong number, discovered publicly, ends a dashboard's usefulness permanently. Recovering that trust takes far longer than building the thing did.
Show the refresh time, the definition, and the source. Where a figure is provisional, say so on the dashboard rather than in a footnote nobody reads.
Deliver it where the work happens
A dashboard requiring a deliberate visit competes with everything else demanding attention. Pushing the important change into email or chat, with a link for the detail, gets used considerably more.
The measure of a dashboard is not how much it shows. It is whether anyone opened it yesterday and did something differently as a result.