Dashboard
A saved screen of the same charts in the same places, checked on a schedule, so a change in the numbers jumps out.
See it
What it is
A dashboard is a saved set of queries plus a fixed layout. The value comes from the sameness: same charts, same positions, same default date range, checked on a rhythm, so a number that moved is obvious before you read a single label. Digging into why something moved happens somewhere else, in an ad hoc exploration. A dashboard answers one question: is anything different today.
Build one per audience and per decision, not one per team ego. The good ones fit a single screen without scrolling, run 5 to 9 tiles, put the headline number top-left where the eye lands, and pair every raw figure with a comparison (previous period, previous year) because a number with nothing to compare it to means nothing. If a tile only matters when it breaks, it is an alert, not a tile.
Gotcha: dashboards rot quietly. Nobody deletes a chart, so a clean six tiles becomes forty that everybody scrolls past, and two of them named 'active users' quietly compute it differently (that one is called metric definition drift). Sticky date pickers and leftover filters cause fake panic more often than real incidents do, so write the metric definition into the tile description and reset the range by default.
Ask AI for it
Design a dashboard for [audience] that answers [decision they make weekly]. Give me at most 8 tiles in priority order. For each tile: chart type, the exact metric definition (numerator, denominator, time grain), the comparison shown next to it, and one line on the action a viewer takes if it moves. Put the headline number top-left, default the range to the last 28 days, keep it to one screen with no scrolling, and drop any tile with no decision attached to it. Flag any metric whose name could be computed two different ways.