Case study · billing month one
A window tint and auto glass shop in Oregon, in a town of about 22,000. The owner was good at the work and known for it — the most reviewed shop of his kind in the county. The work was never the problem.
His host had suspended the site. The domain served an error page, the certificate was untrusted, and the ads he was paying for every month were still pointing at it. His old quote form had produced about five inquiries a week; that month it produced none, because there was nothing for it to sit on.
Meanwhile his listings were quietly working against him. His Yelp page was unclaimed and showed a 4.4 off nineteen reviews. Bing was republishing that same nineteen. And when you asked an AI assistant who to call, it read the 4.4 — not the 4.6 and 157 reviews sitting on his Google profile — and put him behind a competitor with ten reviews.
This is the month after.
There is no “before” to compare these to, and that is the point — the month before, his domain served an error page and Google had nothing to show. This is not a percentage improvement on a website. It is nothing, versus twenty-two quote requests in one month, in a town of about 22,000.
Where those 22 came from
Eighteen of the twenty-two came from unpaid Google search — the part I work on — off just over a third of his traffic. It is also the only source where more than half the visitors did anything at all. His paid channels sent 61 people and produced nothing.
What 22 would be worth in your shop
$2,640a month, at your numbers
The only figure here that is mine is the 22. The job value and the close rate are yours — I am not going to tell you what your work is worth.
The uncomfortable part
Same month, same shop, same phone number. Display ads brought 36 visits and not one engaged session. Paid social brought 25 more and produced zero inquiries. Sixty-one people, one of whom stayed long enough to count as engaged.
Unpaid Google search brought 139 — fewer than 40% of his traffic — and produced 18 of the 22 inquiries. It was also the only channel where more than half the visitors did anything at all.
I am not telling you your ads are wasted; I do not know what you are spending or who runs them. I am telling you that is a question worth asking out loud, because it was worth asking here, and nobody had.
The part that surprised me
On day zero, asked who to call for his trade in his town, Perplexity could only cite a third-party directory — because his own site was down, there was nothing of his to read. Three weeks later the same question returned “I’d call [him] first”, citing his own website as the source.
Not every assistant moved. ChatGPT still ranks him fourth, still reading the stale 4.4 from the listings rather than the 4.6 on Google. That one is not fixed and I am not going to pretend it is — it is next month's job, and it lives in the directories, not the site.
What actually got done
His domain secured and verified paid through 2031, because a lapsed domain is a business that disappears overnight. His Yelp claimed — free, ten minutes, and half the reason the AI was quoting the wrong rating. A site on his own domain with a valid certificate, live in about two days. A quote form tested end to end as a customer, not just installed. Individual pages for the services he actually sells, because Google cannot rank a service it has never been told about.
None of that is clever. All of it was missing.
What moved on Google
Traffic and leads from Google Analytics, 6 August to 3 September 2026. Rankings measured twice on the same searches, 22 August, against a baseline frozen 31 July. A “lead” means a tap on his phone number or a quote form sent — not a booked job; only he knows how many of those closed. Three other things I track — reply time, jobs booked, and customers who came back — do not have a clean number yet, and I would rather tell you that than estimate them. Rank readings go stale after a fortnight; ask me to run it again.