What is a dashboard?

Why most dashboards go unread

Because they answer no question anyone was asking. A dashboard built by listing every available metric produces a wall of charts that is comprehensive and useless — the reader has to work out which numbers matter before they can think about them.

The dashboards people actually use are built backwards: start from the decision, work out what would change it, and show only that. Everything else belongs in an ad hoc query, not on a screen someone is supposed to check daily.

What separates a useful one

• A stated purpose. Who reads it, how often, and what they might do differently as a result.
• Comparison built in. A number alone is uninterpretable; against last period, target or segment it means something.
• Definitions visible. If nobody can say what a metric includes, nobody will trust it enough to act on it.
• Honest uncertainty. Provisional data marked as provisional — see temporary attribution.
• Fewer charts than seem necessary. Removing metrics usually increases how much a dashboard is used.

The trap

A dashboard makes some things visible and everything else invisible. Over time, teams optimise what is on the screen because that is what gets discussed, which means dashboard design quietly becomes strategy.

That is worth being deliberate about. If attributed ROAS is on the dashboard and incremental lift is not, the organisation will optimise attributed ROAS regardless of what anyone believes about measurement.

The corrective is to put the uncomfortable numbers on the dashboard too — the ones with wide uncertainty and slow feedback — rather than confining them to a quarterly deck nobody acts on.