Review SEO changes without confusing timing with causation

Save a baseline before release

Before release, save the URL, title, main headings and known indexing state. A versioned copy and clear reason for the edit are more useful than pasting the entire body into a report. Export search metrics with explicit periods and filters, marking unavailable data as unavailable. Decide what you intend to inspect before seeing the outcome. This reduces selective use of attractive dates and makes a later reviewer able to distinguish the original hypothesis from an explanation invented after the chart moved.

Describe the actual intervention

List every element changed in the release: title, body, internal links and URL if applicable. “Rewrote the article” is too vague to connect a result to an intervention. Note concurrent migrations, template updates or measurement corrections that complicate interpretation. Do not compress several changes into a single confident causal story. Instead, use the review to decide whether future work can be separated into more inspectable steps, while recognizing that practical site releases sometimes necessarily combine several related fixes.

Compare matching windows and filters

Match window lengths and record country, device, search type and page filters. Consider weekday composition and seasonal demand; where useful, add a comparable prior-year period as context. Across a domain migration, map old and new URLs before comparing properties. Do not add overlapping exports or combine figures collected under incompatible filters. A report can contain several clearly labeled comparisons, but should not disguise them as one continuous series when the measurement boundary changed during the period.

Inspect several signals together

Read clicks, impressions, CTR and average position together. New query exposure can alter aggregate CTR or position without showing that the answer deteriorated. Inspect the main question groups for the page and whether the intended audience is being reached. If the movement cannot be explained, preserve that uncertainty. A useful review states the next inspection needed instead of choosing whichever single metric makes the change look successful. Aggregate charts are starting points for investigation, not a substitute for examining the queries and pages underneath them.

Use surrounding pages as context

Check whether the movement is isolated or shared by related pages and the wider site. Surrounding pages do not establish a controlled experiment, but can suggest broader demand changes or technical problems. Choose reasonably comparable purposes and periods rather than forcing unrelated articles into a control group. If a shared component changed across many URLs, record a site-level intervention instead of attributing the same effect independently to every article. This keeps the evidence proportional to what the actual release and available comparisons can support.

Separate observation, interpretation and action

Write three separate statements: what changed in the data, what might explain it and what should happen next. For example, impressions rose; additional query exposure may contribute; inspect whether the page answers those new questions. Combining these into a single sentence can make an interpretation sound established. Keep uncertainty visible even in internal reporting. When evidence remains sparse, continuing observation can be a legitimate next step. Repeatedly changing titles to manufacture a new intervention may make the situation harder to interpret.

Example: an edit during a migration

Consider a hypothetical pricing rewrite released in the same week as a domain move. Rising clicks on the new URL may partly reflect traffic moving from the old URL. Save the mapping, redirect state, query composition and dates of indexing observations. Do not advertise a short comparison during migration as the effect of the rewrite alone. A later review should ask whether evidence now separates the migration process from changes in the answer. The purpose is a sound next decision, not an impressive percentage detached from its context.

• Why SEO tool numbers disagree: reconcile rank, traffic and keyword data
• Map URLs for a site move: retained pages, consolidation and checks
• Read Search Console queries and pages together before editing