After a new inspection workflow launched, an analyst picks a post-launch week with unusually low backlog as the 'baseline' and reports a huge improvement. What baseline practice should the information manager insist on?

Baselines for post-intervention evaluation must be chosen without cherry-picking post-hoc windows that create fake success. Selecting an unusually low-backlog week after a new inspection workflow launches as the permanent baseline manufactures improvement and misleads leaders about sustained performance and staffing needs. Always using the single best week after go-live as the comparison point institutionalizes that bias in every future chart. Ignoring pre-intervention history because “new systems make older data irrelevant by default” discards the contrast needed to evaluate whether the change worked. Setting the baseline equal to the target so every chart shows zero gap hides performance gaps rather than measuring them. Exam caveat: legitimate baseline choices include pre-period averages, matched seasonal weeks, or agreed control periods—and the rule should be documented before inspecting which post-period week looks best. Operational check: pre-register the baseline window and metric definition, compare the post-launch series to that baseline and to a longer pre-trend, display uncertainty, and refuse briefings that swap in best-week baselines after seeing results. In municipal practice, keep the rationale and rejected alternatives visible to on-call analysts and stewards so the next briefing cycle does not reintroduce the same failure mode for city services.