Do not audit a moving rollout as if it were finished
Google's broad core updates can roll out over days or weeks. During that period, results may move in more than one direction. Record what you see, check the official status, and resist making large reactive changes from the first downturn.
Once the rollout is complete, compare a stable post-update period with a comparable pre-update period. Use page and query groups rather than a sitewide average. A site can gain in one intent class and lose in another; collapsing both into one visibility score hides the decision you need to make.
Correlation earns investigation, not certainty
If the first sustained loss begins inside the rollout window, the update is a plausible explanation. It is not proof. Check deployments, indexing, migrations, analytics, demand, and manual actions around the same date. Then compare the affected result sets before and after the rollout.
Ask what changed in the winning pages: the problem they solve, the evidence they supply, how current they are, whether the result format changed, and whether the query now expresses a different intent. Do not reduce the comparison to word count, “authority,” or a generic quality score unless the evidence actually points there.
Improve the cause you can defend
Google says there may be nothing technically wrong with pages that perform less well after a core update. That does not mean doing nothing. It means the work should improve the page and site at the level revealed by the comparison: consolidate overlapping pages, replace recycled summaries with original evidence, correct stale information, fix inaccessible primary content, or make the page answer the current reader's decision.
Large deletion programs need more evidence than a red graph. Removing a useful section can make the next comparison worse while destroying the history that made recovery possible.
Measure the new site, not the promise of recovery
Google says improvements can take effect in days or require several months, and there is no guarantee of a noticeable ranking effect. Define leading checks you control: the repaired page is indexable, overlapping URLs are consolidated, original material is present, and internal links reflect the new ownership. Then track impressions, clicks, and qualified outcomes for the affected groups.
The update timeline helps establish when to compare. It cannot decide the repair for you.