In June we ran an audit. Across engineering, product and operations, we counted 412 dashboards. Then we checked the access logs: only nine had been opened by more than three people in the previous month. The rest were abandoned — monuments to questions somebody once asked.
Dashboards are answers to old questions
Nobody builds a dashboard on purpose to be useless. Each one began as a good answer to a specific question during a specific incident or launch. The problem is that the question expires and the dashboard doesn't. Months later it's a graph of a number nobody is responsible for, one of several that look similar and disagree slightly.

The cost is attention, not storage
Dashboards are cheap to keep and expensive to trust. When there are forty ways to see signups, a conversation about growth begins with ten minutes of arguing about which chart is right. A mature team doesn't have more instruments; it has fewer, and it knows exactly what each one means.
What we did
- Archived every dashboard with fewer than three viewers in 90 days, with a one-click restore.
- Gave each remaining dashboard a single named owner and a one-sentence purpose at the top.
- Replaced the 'overview' dashboards with a weekly written summary that explains what changed.
Three weeks later, the number of questions in our #metrics channel dropped by roughly a third. Nobody asked for a restored dashboard. The nine that survived are better maintained than the 412 ever were, and more importantly, people look at them with a question in mind.
If you do one thing this quarter, run the audit. You'll find the dashboards you actually use, and the exercise will tell you what your team believes is important.



