Every company has one: the dashboard someone built eight months ago, opened daily for two weeks, and now visited only by the person who made it. The charts aren't wrong. They're just not answering a question anyone is actually asking.
Our fix is to start from the decision, not the data. Before designing anything we ask: what will someone do differently after looking at this? If the honest answer is 'nothing, but it's good to know,' the metric doesn't make the cut. A dashboard that supports two real decisions beats one that displays forty interesting numbers.
Then the discipline gets boring, deliberately. One number per question. The comparison that gives it meaning (versus last month, versus target) right next to it, not two clicks away. Colors that mean something consistently (if red doesn't always mean bad, red means nothing). Labels a new hire understands without a glossary.
Under the visuals, trust is a pipeline problem. The fastest way to kill a dashboard is one wrong number at the wrong moment. After that, everyone quietly goes back to their own spreadsheet, and no redesign wins them back. Freshness indicators, data tests, and a visible 'last updated' timestamp do more for adoption than any visual polish.
The test we use before shipping: watch someone open it cold and narrate what they think it says. If they hesitate on any chart for more than a few seconds, that chart gets simplified or cut. The analytics dashboard behind this site's staff login went through exactly that process, which is why it shows five numbers, not twenty-five.
