Dashboard

A dashboard gathers the few measures someone needs to watch onto one screen, usually as numbers, charts and status indicators. The test of a good one is simple: can the person tell within seconds whether something needs their attention, and what?

What it is for

A dashboard answers a recurring question. How is the system doing, is the campaign on target, is the machine running within limits. It is for monitoring and triage, not exploration. The moment someone has to dig, they should be able to click through to detail, but the dashboard itself stays shallow on purpose.

Dashboard versus report

An analytics report interprets a period of data and argues a point. A dashboard has no argument; it shows the current state and lets the viewer decide. If you find yourself writing paragraphs of commentary on a dashboard, you wanted a report.

What goes wrong

The usual failure is the dashboard built from whatever data was available rather than from the decisions the user makes. Twenty tiles of equal weight create a wall that is technically complete and practically unreadable. Use visual hierarchy: the one number that matters most gets the most space, and anything abnormal should look different from anything normal.

Design for the bad day, not the demo. Show what the screen looks like when data is missing, stale or wildly out of range, because that is when someone is relying on it. In industrial settings, where a human-machine interface is a dashboard in all but name, a misleading display is a safety issue, not a styling one.