The rules
- Every figure names its source and the date it was read, in the same place the figure appears. Not in a footnote, and not on a different page.
- Search performance is compared as full calendar months. Partial periods produce changes that are artefacts of the dates.
- Speed figures are described as what they are — single-visit lab measurements on a stated date — and never presented as Core Web Vitals field data, which is a different measurement Google takes from real Chrome users over 28 days.
- A number we did not measure is labelled as such, in the sentence, every time.
- A number we cannot reproduce on request is not published. That rules out AI visibility scores, and it is why there is no such score anywhere on this site.
- Case-study figures are re-measured rather than left to age. Where a figure has been re-read, the date changes with it.
- No rating, review count, award or "trusted by" claim appears unless it can be shown verbatim from its source. Where a rating exists elsewhere, this site links to it rather than restating it.
What is deliberately not here
The absences are as deliberate as the figures, and each one is an absence because the alternative would have been a claim rather than a measurement.
- A star rating or review count. The previous version of this site carried an aggregate rating claiming 47 reviews against three testimonials in the code. It has gone, along with the page it was on.
- An AI visibility score. Every one I have seen sold is either a proprietary index nobody can reproduce or a sample small enough to swing weekly. The method used here is published in the AI guides instead; the score is not.
- A Google Ads campaign result. There is no published client campaign performance figure, because none has been cleared for publication. The Google Ads page says so on the page rather than leaving you to notice.
- Search performance figures for Vehicles 2 Lease. Cybrial rebuilt that site and works on it monthly, but no search figure for them is published here — there is not yet a window long enough to report honestly.
- Industry averages dressed as our results. Where a market range is given — in the cost guides — it is labelled as a range, with its basis stated.
Why analytics totals are treated as floors
A UK site with a consent banner only records an analytics session once the visitor accepts. Everyone who declines, ignores it or leaves first is missing. That makes every GA4 traffic figure a floor rather than a total, and it makes every conversion rate calculated from one an overstatement — often by several multiples.
Measured on one e-commerce site I run, over a single 28-day window: Merchant Centre recorded 337 Shopping clicks while GA4 recorded 54 organic Shopping sessions, and 288 sessions across all channels. Both measurements are correct. Only one of them is a total, and quoting the other as though it were is how a business ends up pricing an ad campaign against a conversion rate that does not exist. There is more on how to fix that in the tracking guide.
When something here turns out to be wrong
It gets corrected on the page, and the date at the top changes. Not quietly deleted. The whole point of publishing the method is that it makes the claims falsifiable, and a policy that only survives while nothing is ever wrong is not a policy.
