SEO analytics
Also called Search analytics, SEO measurement
SEO analytics is the measurement of what organic search produces for a business. Done properly it connects a search query to a session, a session to an enquiry, and an enquiry to revenue. Done badly it reports rankings and sessions, which cannot tell you whether the work paid for itself.
Written and maintained by the SEO Engico team
Two mechanical engineers turned marketers, and the people who do the work. Meet the team
Last reviewed 4 August 2026
Why are rankings a poor measure?
Because a ranking is an input, not an outcome. Positions move for reasons unrelated to your work, results now vary by person and location, and a large and growing share of searches end without a click at all. A rankings table can improve in a month where enquiries fell, and it happens often enough that reporting one as though it were the other is a real problem.
Impressions have a related trap. Automated tools, rank trackers and AI grounding queries all generate Search Console impressions that no human will ever click, and on smaller sites they can dominate the total. Prioritising content off raw impressions sends you after traffic that does not exist.
What should actually be measured?
Enquiries and revenue attributed to organic, at minimum. That needs GA4 key events configured with real values, Enhanced Conversions so the conversions Safari and iOS drop are recovered, call tracking because most reporting never sees the phone, and ideally CRM stitching so a closed deal traces back to the query that began it.
Alongside that, track position on the specific commercial queries that matter rather than a site-wide average, because a site-wide average mostly measures how your long tail is behaving.
How do you measure AI search visibility?
Search Console now reports impressions from AI surfaces, but without clicks or queries attached. For anything beyond that you need to query the models directly with a fixed prompt set and record whether you are cited. Because the answers vary between runs, a single check is close to meaningless. Run the set repeatedly and report a rate with an interval.