Cold start baselines that survive a debate
Cold start debates fail when each engineer measures a different phone on a different café Wi‑Fi. A baseline worth citing lists the device model, OS build, app version, whether the account was logged in, and whether caches were cleared.
Run the path several times and report a range, not a single heroic best attempt. Note if the first launch after install differs from the fifth — install-time work should not be mixed into everyday cold starts without a label.
Share the script so QA can reproduce after a suspected regression. If reproduction requires a special build flag, the baseline belongs in engineering notes, not in an executive slide.
Performance Baseline Reviews from NetApp Solutions treat those conditions as part of the deliverable, so the number remains usable when the next train ships.