Capacity stories that skip theater metrics

Theater metrics are easy to spot: they rise monotonically, nobody can explain what bad looks like, and they never connect to a release decision. We ask teams to pick one user journey and three signals that would change a ship date if they crossed a threshold. Everything else is context, not headline.

Once those signals exist, we practice describing them in plain language. “Checkout adds two hundred milliseconds at p95 when the cache partition is cold” beats “CPU is fine.” The second sentence tells product what risk they are carrying.

We close with a rehearsal where a non-engineering partner asks blunt questions. If the answers require opening three tabs, the story still needs editing.