The dashboard nobody read

A team built a dashboard with forty-two charts. Six months later, logs showed that eleven people had opened it in total and that the median session lasted nine seconds.

The post-mortem identified a specific failure rather than a general one. Each chart had been requested by someone, and no chart had been designed against a decision. Asked what action a figure would change, the team could answer for four of the forty-two, and those four were the ones occasionally used. The rest answered questions nobody was going to act on, which is a reasonable description of most reporting.

The replacement had five charts and one rule: every chart carries, in its subtitle, the decision it informs and who makes it. 'Queue length by hour — tells the duty manager whether to open the second desk' is a specification, and a chart that cannot be given such a line is removed. The constraint did more work than any design improvement: four of the five charts were redrawn once the decision was written down, because the original form could not answer the question it turned out to be for.

The nine-second median is the other lesson. A dashboard is read at a glance by someone who came for one number, so the layout question is not what to include but what the eye lands on first. The team put the single most decision-relevant figure at the top left in large type, with everything else below, and usage rose to a point where people complained when it was slow.

That complaint was the success metric. Nobody complains about a tool they were not using.