How to Measure SEO

Coverage comes before everything else, because a page that is not indexed cannot produce any other metric. Search Console's indexing report shows which URLs Google has stored and, more importantly, which it has excluded and why — soft 404s, duplicates without a chosen canonical, crawled but currently not indexed. Each exclusion reason is a work item.

Only once coverage is healthy do the performance metrics mean anything. Impressions tell you how often a page appeared in results; clicks tell you how often the appearance was compelling enough to choose; average position tells you roughly where you appeared. The three together diagnose most problems: high impressions with low clicks is a title-and-description problem, low impressions with decent position is a demand problem, and plenty of both at position eleven is an authority problem.

Engagement is the layer most SEO reports skip. Organic sessions, time engaged, and the share of visitors who reach a second page separate rankings that earn attention from rankings that merely exist.

Total keyword counts are the classic one. Ranking for ten thousand phrases nobody searches is indistinguishable from ranking for nothing, and tools that report 'keywords ranked' without volume-weighting are reporting noise. What matters is movement on the specific queries each page was built to win.

Third-party authority scores — Domain Rating, Domain Authority and their cousins — are another. They are one vendor's model of link strength, useful for comparing two sites at a glance, useless as a KPI because Google does not use them and they can rise while your traffic falls.

Raw backlink counts round out the list. A hundred new directory links move nothing; one new editorial link from a relevant publication can move everything. Count referring domains of real quality, not links.

Connect the analytics to the thing the site exists to do. For an ad-supported content site that means engaged organic sessions and pages per visit. For a lead-capture site it means organic visitors who complete the form. For a shop, organic revenue. Define the conversion first, then report organic traffic against it every month without exception.

Segment before you conclude. Blended organic numbers hide everything: brand queries (people searching your name) grow with reputation, not SEO work, so split them from non-brand queries. Split landing pages too — one tool page taking off can mask twenty others quietly decaying.

Attribution will never be perfect, and that is fine. The question a monthly report needs to answer is not 'exactly which visit produced which dollar' but 'is organic search producing more of what we want than last quarter, and which pages are responsible?'

Search Console stops at the edge of the generated answer. A citation inside ChatGPT, Perplexity or an AI Overview may produce no impression and no click, so classical reporting will miss the win entirely. The supplementary measurements are direct ones.

Keep a written list of the questions each important page should be the answer to, and run them against the major engines monthly. Record three things per prompt: were you used as a source, were you named, and which page was cited. The trend in that table is your GEO scoreboard.

Watch assistant referral traffic in analytics as a second signal — visits from chat.openai.com, perplexity.ai and their peers are small in volume and disproportionately high in intent. A growing trickle from those domains is often the first measurable proof that citation work is paying off.

Fix the dashboard before the routine: one page showing indexed-page count, non-brand clicks and impressions, position for your twenty target queries, conversions from organic, and the vitals status of your ten most important pages. If assembling the numbers takes more than a few minutes, the dashboard is wrong.

Then the monthly hour: check coverage for new exclusions, scan the performance report for pages that gained or lost more than a quarter of their impressions, run the AI prompt list, and write down three sentences — what moved, why it probably moved, and what you will do about it. The written note is what turns measurement into learning.

Resist daily checking. Rankings wobble day to day for reasons that mean nothing, and reacting to noise produces the worst kind of SEO work: constant motion, no direction. Monthly trends, quarterly strategy reviews, and immediate action only on hard failures like deindexing or a vitals collapse.