Compare like with like
Choose a period long enough to reflect the page’s normal demand and compare it with a relevant earlier period. Note seasonality, campaign activity, tracking changes and product events. Separate query impressions, clicks, position and qualified actions; these can move in different directions for different reasons.
For example, fewer clicks with stable impressions may warrant inspecting the results page and snippet. Fewer relevant impressions may suggest demand or coverage has changed. A loss of enquiries with steady traffic may be a fit or conversion problem. Record the pattern before labelling it an algorithm issue. Avoid explaining a decline with whichever industry headline appeared most recently.
Rule out technical and measurement faults
Open the page, test its response and inspect its canonical and indexing instructions. Check whether a redesign removed useful links, changed a URL or broke an asset. Confirm that the analytics or conversion definition did not change during the comparison period.
Review historical redirects and mobile behavior. A page may appear healthy at its new address while older referring links lead somewhere irrelevant. An embedded form may fail only on a smaller screen. These problems call for repairs, not another thousand words of content. Preserve the distinction in your report so a technical recovery is not attributed to an editorial experiment.
Inspect the current reader task
Read the queries that now bring people to the page and the questions sales receives. Has the buying problem changed? Have new options altered the comparison? Is the article teaching a retired feature or referring to an old plan? Compare your content with current official product information rather than copying a competing article’s structure.
Mark specific gaps: missing prerequisites, unsupported claims, outdated screenshots, unhelpful examples or a mismatch between the headline and the answer. Choose improvements that change what a reader can understand or decide. A new date and a rewritten introduction do not constitute a meaningful refresh when the core problem remains.
Choose refresh, merge, redirect or retain
Refresh a page that still owns a useful question but needs better information. Merge pages when they compete for essentially the same reader task and one stronger destination can preserve their value. Redirect a retired URL when there is a genuinely relevant replacement. Keep a useful page when a short-term fluctuation lacks a clear problem.
Document the reasoning and preserve a copy of the old page. If you remove content, check which internal pages and external links still depend on it. Update links to the final preferred destination. Do not treat deletion as a shortcut to a cleaner reporting chart; losing a useful resource can create a different discovery gap.
Measure the refresh as a change, not a promise
Record the edit date, affected sections and intended benefit. Choose a follow-up measure that fits the diagnosis: indexing state after a technical correction, relevant queries after an intent update, or enquiry quality after a product explanation changes. Keep unrelated releases and distribution activity in the notes.
Review the result over a sensible observation window and preserve uncertainty. A page may improve while demand rises across the category; that does not isolate the effect of your edit. The next decision should follow what you can observe. Continue improving when a known gap remains, test a different explanation when evidence contradicts it, and avoid repeated rewrites simply to show activity.