A reader once left a message on my answerphone with a very simple question: how do you measure SEO? She needed figures for an internal report and worried it sounded naive. It was not: it is the most useful question anyone can put to a consultant, and the easiest to answer badly.
The question has not aged. In the French study Réussir avec le web, published by Afnic in 2025 (around 2,500 French micro-businesses and SMEs surveyed between September 2024 and August 2025, self-reported answers), 27% of owners judged the time spent on their online presence to be profitable, and 39% said they did not know how to assess it. Those are French figures about French firms, and the survey covers nothing outside France. The situation it describes is familiar enough.
Learning how to measure SEO performance mostly means accepting what each tool can see, what it cannot, and in which order the effects appear. Here is the chain worth tracking, the tools covering each link, and the blind spots that make people read their data backwards.
What you are measuring, and in what order
Organic search produces a chain of effects, not a single number. Each link needs a different tool, and each loses a little information compared with the one before.
- Visibility: your pages appear in the results (impressions, position).
- Clicks: a share of those appearances becomes a visit.
- Sessions and engagement: what visitors do once they are on the site.
- Enquiries: a form, a call, a quote request, an order.
- Revenue: what those enquiries turn into, often weeks later.
The classic mistake is to follow only the first link, rankings, or only the last, this month's revenue. The first moves long before sales do; the second depends on things unrelated to Google. Useful measurement connects the two ends.
| Tool | What it gives you | Main blind spot |
|---|---|---|
| Search Console | Impressions, clicks, CTR, average position, queries, pages | Nothing after the click; rare queries are withheld |
| Analytics (GA4, Matomo) | Sessions, engagement, key events, acquisition channels | Depends on consent and tagging; blind to impressions |
| Rank tracking | Position on a list of queries you have chosen | Results are personalised and localised; the sample is your own |
| CRM, till or invoicing | Qualified enquiries, sales, margin | Rarely linked back to the first contact |
Search Console: the reference source on the search side
It is the only source that tells you which queries Google shows your site for. Four metrics structure the Performance report: clicks, impressions, CTR and average position. Be careful with the last one. In the chart it reflects your best result, aggregated at property level; in the table it applies to the row displayed. It is not comparable to a third-party reading.
The branded and non-branded filter
Google announced the branded queries filter on 20 November 2025, then confirmed on 11 March 2026 that it was available for all eligible sites. It separates searches containing your brand name — variants, misspellings and identifiable product names included — from everything else. Non-branded growth shows whether new people are finding you; branded growth reflects awareness or advertising.
Three caveats, documented by Google: it is offered only for domain or top-level prefix properties, not a subdomain or directory; it needs sufficient query volume; and the classification is AI-assisted, so it can misread individual queries. Search Console help notes 16 months of history.
Limits to know before you draw conclusions
- Anonymised queries: Google has documented this since 2022 — queries issued by too few users over a two to three month window are not shown. They count towards the chart totals but not the table, so the sum of the rows is always lower than the total.
- Export volume: 1,000 rows from the interface, up to 50,000 rows per day, per site and per search type through the Search Analytics API. On a large catalogue, that export is not enough.
- Freshness and history: data is generally available within two to three days, and the most recent days may still be preliminary. Export every month, or connect the API, if you ever want to compare this year with three years ago.
One habit is worth building early: falling clicks with stable impressions is a different problem from falling impressions. The first points at your titles, your snippets or the results page itself; the second at your pages and their status in Google's index.
Analytics: from the click to the enquiry
Search Console stops at the click. For what follows you need an audience tool such as GA4 or Matomo. Two points deserve attention.
The channel. Isolate organic search before you read anything: direct traffic, social, email and paid campaigns follow different logics. Check that your campaigns are tagged. Untagged ones land in the wrong bucket and contaminate every later comparison.
Attribution. A visitor rarely arrives through the channel that triggers the purchase. Since the removal of the linear, time decay, first click and position-based models in November 2023, GA4 offers only three models: data-driven attribution, last click across paid and organic channels, and Google paid channels last click. All ignore direct visits unless the whole path is direct. The default lookback window is 90 days for most key events and 30 for acquisition events, and conversions can be reattributed up to seven days after being recorded. Never judge a month in the days right after it closes.
A last trap, usually found too late: user and event data retention is set to either 2 or 14 months. It does not affect standard aggregated reports, but it limits explorations and funnel reports, and it cannot be applied retroactively. Move it to 14 months now, or switch on the BigQuery export.
Seasonality and year-on-year comparison
Comparing September with August means very little in most trades. Compare the same period year on year, and check three things before interpreting the gap.
- The calendar: working days, school and bank holidays, and the movement of Easter or Black Friday from one year to the next.
- Changes on your side: a redesign, a migration, a new consent banner, a change to your tagging. A collapse in audience is sometimes a collapse in measurement, which is why a technical check often precedes any conclusion about traffic.
- Changes on Google's side: algorithm updates and shifts in how results are displayed. A falling CTR at a stable position usually says the results page has changed, not your site — AI-generated summaries above the classic links being the clearest example.
Keep a dated log of what you do: publications, link acquisition, template changes, an advertising burst. Without it, a curve proves nothing.
Rankings: measure progress, not only results
Moving from position 100 to 80 brings neither traffic nor revenue. It is still information: the work is being registered. But you need a tool capable of seeing it.
I once followed a site launched in February, with its first links built shortly afterwards. For two months my rank tracker showed nothing, because it only recorded the top twenty. A second tool, reaching position 100, showed steady progress from month one. Same site, same period, opposite readings; the difference was measurement depth.
Results also vary with time of day, location, device and search history. A rank reading is a photograph taken under specific conditions, not a fact about your site. Real clicks in Search Console are a better guide, and they should drive which pages and queries you prioritise in your content plan.
A monthly dashboard that fits on one page
For a small or medium business, six indicators are enough, recorded on the same day each month, over a rolling year and against the same month last year.
- Non-branded organic clicks (Search Console): your acquisition indicator.
- Branded organic clicks (Search Console): your awareness indicator, read separately.
- Impressions and CTR on your ten to twenty priority queries.
- Number of pages generating at least one click: how broad your visibility is.
- Key events from organic search: forms, calls, quote requests, orders.
- Enquiries you actually handled, from your CRM or your inbox, with their stated origin.
The last one is the most neglected and often the most instructive. Asking "how did you find us?" corrects many attribution illusions, especially when the conversion ends in a phone call no script will see. If the list feels like too much, start with the first two lines and add the rest over a quarter, the same staged logic that applies when you scope a diagnosis.
Finally, accept the delay. Organic search produces its effects over months, and a measurement taken too early is the most common reason people stop something that had just started to work.
Common questions
How long before the effect of SEO can be measured?
Impressions and deep positions can move within a few weeks. Clicks and enquiries follow later, usually over several months, depending on the competition and on how established the site is. A single month in isolation cannot be interpreted.
Why do Search Console and analytics show different numbers?
They do not count the same thing. Search Console counts clicks from Google results; analytics counts sessions recorded on the site. Consent, script blocking, redirects and processing delays create a gap mechanically. What matters is that each source stays consistent with itself over time.
What is the branded and non-branded filter for?
It separates searches containing your brand name from all the others. The non-branded share measures your ability to be found by people who do not know you yet. It is limited to top-level properties with enough query volume.
Do I need to pay for a rank tracking tool?
Not necessarily for a small organisation: Search Console gives you real clicks, impressions and average position at no cost. A paid tool helps to watch competitors, follow positions beyond the top twenty, or automate reporting. Check its tracking depth before subscribing.
How do I connect SEO to revenue when the sale closes on the phone?
Add an origin question to your intake and record the answer in your CRM, even roughly. Track clicks on phone numbers and form submissions as key events. Cross-referencing the two gives a reasonable estimate; exact attribution is not available offline.
Is the average position in Search Console reliable?
It is accurate but frequently misread. In the chart it reflects your best result across all queries combined, so a worsening average may simply mean you are visible on more queries, including some near the bottom. Read it query by query.