Analytics and Reporting

Measure what matters, in this order

Rankings are a means, not a result. The numbers that matter run from the bottom of the funnel up: revenue or booked jobs from search, leads from search, qualified visits, then impressions and citations, then rankings. We set up the bottom first, because a report that only shows positions cannot tell you whether the work is paying.

SEO reporting also has to cover two audiences now. Google still sends most visits, but a growing share arrives from ChatGPT, Perplexity, Copilot and Google's AI Overviews and AI Mode, where the metric is whether you were cited rather than where you ranked. Our reports show both, separately, so you can see which is moving.

Getting the setup right, step by step

1. Search Console. A domain property that covers www, non-www, http and https and every subdomain, plus URL-prefix properties where a section needs its own view. Sitemaps submitted, and access shared with the people who need it.
2. GA4. One property, one data stream per site, internal and known bot traffic filtered, time zone and currency correct, data retention set to the maximum, and key events (what GA4 now calls conversions) defined for forms, calls, bookings, chat and purchases. We test each key event by firing it ourselves.
3. Bing Webmaster Tools. Verified and sitemaps submitted, because ChatGPT search draws heavily on Bing's index and Bing's own AI performance report shows Copilot activity.
4. AI referral tracking. A channel group or segment in GA4 for traffic from chatgpt.com, perplexity.ai, copilot.microsoft.com and the other assistant referrers, so AI visits stop being buried in "referral".
5. Call tracking where phone leads matter, set up so your real number stays consistent everywhere it is listed.
6. Lead quality. A connection from your form or CRM back to the source, so we count qualified leads and not just submissions. For e-commerce, purchase events with value.
7. Consent. A consent banner configured so that measurement respects the choice, with consent mode where you use Google's tools, and an understanding of how much data that will cost you.
8. Baseline. Before we change anything, we record where every number stands, so month one has a fair comparison.

What the monthly report says

The report has five sections, and they are always in the same order so you can compare months without hunting.

1. Headline numbers: organic clicks and impressions, AI referrals, leads or revenue from search, pages indexed, against the previous period and the same period last year.
2. What we did and what it changed: each item shipped, with the date and the effect we can see so far. Fixes that have not had time to show are marked as such rather than claimed.
3. Where you are cited: the share of a fixed prompt panel where each assistant cites you, re-run on a schedule, with the caveat that AI answers vary and this is a sample.
4. What did not work: stated plainly, with the cause where we know it and the change we are making.
5. Next month: a short decision list for the call.

If something went down, the report says so on page one and says why. A dashboard is available any day for the raw numbers; the report is where the judgement lives. See the sample monthly report for the format.

Where the numbers come from, and their limits

Source	What it is good for	What it cannot tell you
Google Search Console	Clicks, impressions, position and queries per page, straight from Google	Very low-volume queries are hidden for privacy; data lags by a day or two; no revenue
GA4	Sessions, key events, revenue, landing pages, referrers including AI assistants	Depends on consent and ad blockers; some figures are modelled; counts sessions, not searches
Bing Webmaster Tools	Bing clicks and impressions; AI performance report for Copilot	Small share of traffic in most markets
Search Console AI performance report	Impressions in AI Overviews and AI Mode by page	No click data for those surfaces, so AI Overview click-through cannot be measured directly
Rank tracking	Position trends for a fixed keyword set	Results are personalised and vary by location; a position is an average
AI citation panel	Whether assistants cite you for a fixed set of prompts over time	Answers change between runs; it is a sample, not a rank
Server logs	Which crawlers fetched which pages, and how often	Nothing about what happened after the fetch

We show these limits in the report rather than hide them, because a number without its limits leads to bad decisions.

What we need from you

• Admin or editor access to Search Console, GA4 and Google Tag Manager if you use it, or the ability to add us. If a past agency owns the property, recovering it is the first task.
• A list of what counts as a lead and roughly what one is worth. Without this we can count submissions but not value.
• CRM or booking system access where you want revenue attribution, or an export each month.
• The site's tag or code access to fix tracking, or a developer who will take a tested ticket.
• Ten minutes a month to read the report and come to the call with questions.

What analytics can and cannot achieve

Good measurement tells you which pages and queries bring the leads, which fixes moved which numbers, whether a drop is a real loss or a tracking break, and what search is worth per month in currency. That last figure is what makes the budget conversation simple.

It cannot measure everything. Consent choices and ad blockers hide a share of visitors from GA4, and that share differs by audience. Google does not report clicks from AI Overviews, so their effect on your traffic has to be inferred. AI citation tracking is a sample, not a rank. Attribution for a business where people research online and buy in person or by phone will always be partial. We tell you the size of each gap instead of pretending the numbers are complete.

How the work itself is measured

• Tracking completeness: every lead type has a key event, and each one is tested quarterly by firing it and finding it in the data.
• Attribution coverage: the share of leads or revenue with a known source, which we try to push up over time.
• Data integrity checks: internal traffic filtered, no duplicate tags, Search Console and GA4 landing pages reconciled, a monthly check that the thank-you page still fires.
• Report on time, every month, with the decision list acted on.

Common failure modes

• The key event stopped firing. A form was rebuilt or a thank-you URL changed and nobody noticed for three months. Monthly firing tests exist for this.
• Two tags on every page. Sessions double, conversion rates halve, and the report looks great or terrible for no reason.
• Internal traffic counted. The company's own staff are the top landing-page visitors.
• Cross-domain forms. A booking tool on another domain breaks the session, so the lead is credited to the booking tool instead of search.
• Comparing the wrong periods. A 28-day window against a 31-day month, or a holiday month against a normal one, reported as a trend.
• Rankings only. A report of green arrows for keywords nobody searches, with no leads column.
• Consent banner set to block everything. Measurement drops by half after a compliance change, and the drop is read as a search loss.
• Dashboards nobody opens. Forty widgets and no sentence about what to do. A dashboard is a data source; the report is the product.

Deliverables

How it runs

1. Audit trackingWhat is measured, what is missing, what is wrong, and a baseline before anything changes.
2. Fix the setupKey events, filters, consent, AI referrals and attribution made reliable and tested.
3. ReportMonthly, in plain language, same structure every month, with the limits of each number stated.
4. ReviewQuarterly check of strategy and budget against what the numbers show.

Which plan includes it

Compare all plans →

Questions

Related