Reading Your Own Search Performance Data

The four numbers

• Impressions. Times a URL appeared in results for a query. It proves visibility, not interest.
• Clicks. Times someone chose your result. The closest thing to a demand signal.
• CTR. Clicks divided by impressions. Highly dependent on position and query type, so only comparable like-for-like.
• Average position. An average across many impressions in different contexts. A movement from 18 to 14 may mean nothing; a page moving into the top few positions matters a lot.

What the data cannot tell you

It cannot tell you why anything changed. It does not show every query — low-volume terms are withheld. It is delayed by a couple of days, so a “drop” at the end of a range is often incomplete data. And comparing a 7-day window to a 28-day window, or a period containing a holiday, produces conclusions that are artefacts of the comparison.

A weekly review that produces decisions

Observe. Write three factual statements: which queries gained impressions, which pages have impressions but almost no clicks, which pages have nothing at all.

Hypothesise. Write three possible explanations, labelled as guesses. “The title does not match the query intent” is a hypothesis, not a finding.

Act on one. Choose the single change with the best ratio of effort to information, make it, note the date, and check again in a few weeks. Changing six things at once means learning nothing.

Diagnostic patterns worth recognising

• Impressions, no clicks. Usually a mismatch between the title and what the searcher wanted, or a position too low to be seen.
• Clicks, no enquiries. The page ranks but does not answer the objection — price, scope or trust.
• Nothing at all. Either not indexed, or the page targets a query nobody searches. Check coverage before rewriting.
• Impressions for unintended queries. Often the most valuable finding: the market is telling you what page to write next.

Coverage and indexing

Use the page indexing report to distinguish “crawled, not indexed” from “discovered” from “excluded by noindex”. Each has a different cause and different fix. Submit the sitemap once, then rely on reports rather than repeatedly requesting indexing.

Keep your own record

Export the figures you care about into a spreadsheet on a fixed cadence, alongside a note of what you changed that week. The comparison between your change log and the data is where learning actually happens — and Search Console’s own history is limited.