How do you isolate which of three simultaneous changes moved rankings?
- Published November 6, 2026
- Linked from 1 answer
- In a cluster of 77
After the fact, you can narrow it down but not prove it. Changes made together leave one combined result, so the only ways to separate them are the differences they left behind: pages that received some of the changes but not all, the dates Google recrawled each page, and the kind of metric that moved. If those differences are missing, the accurate record is that the three changes together moved alongside the result, and the next round should change one thing at a time. Even that joint record is a correlation: compare it with similar pages you did not change, since an update or a competitor’s move over the same weeks can account for part of it.
Evidence that can separate the changes
| Difference to look for | What it lets you see | Example |
|---|---|---|
| Pages that got only one of the changes | That change's effect on its own | The new titles went on twenty pages, the content rewrite on five of them; compare the fifteen title-only pages with the five |
| Different recrawl dates | Which change was in place when the movement started | URL Inspection shows the last crawl; movement before a page's recrawl cannot come from that page's change |
| The metric that moved | The mechanism involved | A title change acts on click-through rate at a steady position; a content change acts on position and on new queries |
| The queries that moved | Which change touched those searches | Gains only on queries named in the new headings point to the content, not the internal links |
For example, suppose you rewrote the titles, added a section and added internal links to a set of guides in the same week. If click-through rate rose while position held, the titles are the likely cause. If position rose on queries the new section covers, and the guides that received links but no new section stayed flat, the section is the likelier driver. None of this is proof, because an update or a shift in demand could have arrived the same week.
When nothing separates them
- Record the combined result with dates, pages and metrics.
- Keep the changes; undoing one to test it costs you whatever it was contributing and takes another recrawl cycle to read.
- Apply the change you most want to understand to a new group of similar pages, alone, and watch that group against pages left unchanged.
Step 3 rests on a control-group method, described step by step in how to measure whether an SEO change worked. Plan the next change so it can be read on its own.
Add a comment