Opening the book…
Real improvement usually has a cause you can name: a new process, extra staff, a fixed bottleneck. A jump with no such cause is more often a definition change, a double count, a reclassification, or a filter someone adjusted. Celebrating it publicly before checking locks you into defending a number you may have to retract.
When a measure moves sharply, ask what changed in the method before you ask what changed in the world, and go to the source records rather than the dashboard. Check whether the denominator shifted, whether anything was recategorised, and whether the cut-off date moved. Hold the announcement until the raw data supports the story.
Step changes from a known discrete event — a facility opening, a large contract landing, a system cutover — have their cause already identified and need only confirmation. Genuine breakthroughs do happen; the rule asks you to verify, not to disbelieve.