Home/How we measure
How we measure
Every number on this site names the test that produced it, the date it ran, and what it does and does not prove. This page defines each instrument so the figures can be checked rather than trusted.
The record we keep for every result
A published figure without its method is decoration. For each result we keep and can show:
- Test name: which instrument ran
- Test date: the day it ran, not the day we wrote it up
- URL tested
- Scope: what the test examined and what it ignored
- Before result
- Changes made between the two runs
- After result, on the same instrument
- What it proves
- What it does not prove
- The report or screenshot behind the number
Two figures from different instruments are never compared. A readiness score is never presented as a ranking.
The three instruments
Full audit: “Google can read the site”
Can Google crawl the pages, understand what they say, and index them? Covers crawlability, indexability, titles and descriptions, canonical tags, internal linking, structured data and mobile rendering. Scored out of 100.
It does not measure where you rank, how much traffic you get, or how good your offer is.
Full audit: “AI assistants can read it”
Can AI answer engines reach the page, understand it, and quote the business correctly? Covers whether AI crawlers are permitted, whether content exists in the HTML rather than only after JavaScript runs, whether passages are self-contained enough to quote, and whether structured data identifies the business. Scored out of 100.
It does not measure how often an assistant actually cites you. That depends on the assistant.
Quick scan: “site built clean”
Fourteen pass/fail construction items: status code, title, meta description, canonical, social preview image, viewport, robots.txt, sitemap, structured data present, an entity type in that data, product data where relevant, an H1, HTTPS with HSTS, and whether the page is password-gated. Reported as a percentage of items passed.
It does not measure content quality, rankings, traffic, or whether the business will sell anything. A site can score 100 on this and have no demand at all.
Analytics figures
Where we publish a conversion figure we state the property, the window, and the event definitions. A “contact event” is a tracked key event in Google Analytics; calls, form submissions and message events are counted separately in the source record and are not silently blended into one number.
Limits we state openly
- Scores from different instruments are not comparable. We do not put them in the same table.
- A single before-and-after pair is evidence, not proof of causation. Search results move for reasons outside any engagement.
- Where a figure is ours rather than a client’s, we say so.
- Where we have no measurement, we describe the work instead of estimating a number.
Naming an outcome is not the same as having moved one
Before we agree to build anything we name which of the three outcomes it is meant to move: win more work, earn more per customer, or spend less doing the same work. Work that cannot be tied to one of them cannot be proved afterwards, and unprovable work is how firms keep clients who are getting nothing.
Naming the outcome only says what we are aiming at. It is not evidence. So each outcome we talk about has to arrive with one of two things attached, and never with neither:
- A dated measurement on a named instrument, the same shape as every other figure on this site.
- Or a plain statement that we hold no measurement for it. Where that is the case we describe the work instead of estimating a result.
As of 2026-08-15 that leaves us with dated evidence for win more work (Allday Fence, Search Console, 2 clicks to 176) and no dated figure of our own for either of the other two. There is no hours-saved, error-rate or rework figure for any property on this site, including our own. The Permit Solutions Automation record describes a system that exists and runs; it is not a savings claim, and reading it as one misrepresents it.
The reason we are strict about this is that an unqualified outcome reads as a guarantee. Once a claim is heard as a promise, whether it was measured stops being an academic question.
Common questions
What does "Google can read the site" measure?
Whether Google can crawl the pages, understand what they say, and place them in its index. It covers crawlability, indexability, titles and descriptions, canonical tags, internal linking, structured data and mobile rendering. It is a readiness score. It does not say where the site ranks.
What does "AI assistants can read it" measure?
Whether AI answer engines can reach the page, understand it, and quote the business accurately. It covers whether AI crawlers are allowed, whether the content is present in the HTML rather than only assembled by JavaScript, whether passages are self-contained enough to be quoted, and whether structured data identifies the business. It is a readiness score, not a measure of how often you are cited.
What does "site built clean" measure?
A shorter check of construction quality: status codes, title and description, canonical tag, social preview image, viewport, robots.txt, sitemap, structured data presence, an H1, and HTTPS with HSTS. It is fourteen items, pass or fail. It is not a full audit and it is not a ranking.
What is the difference between a full audit and a quick scan?
The full audit is the scored instrument behind the Permit Solutions Services and MyHausFax numbers; it examines content, technical health, structured data and AI readiness in depth. The quick scan is fourteen pass/fail items used to check construction before or between engagements. The two produce different numbers and are never compared to each other.
Why do you separate readiness from ranking?
Because they move for different reasons. Readiness is what we control: whether the site is built so it can be found and understood. Ranking depends on competitors and search-engine changes nobody controls. Averaging the two lets a bad month look like a good one, which is why we publish them apart and label which is which.