Results
Everything on this page is work I did, and none of the clients are named. That is deliberate. They did not hire me so their competitors could read about their account. What I can show you instead is the problem, what I changed, the number that came out of it, and how that number was measured.
Where a figure comes from a client’s CRM instead of an advertising platform’s own dashboard, it says so, because those two sources disagree more often than the industry likes to admit. Where a comparison has limits, the limits are printed next to it instead of left for you to find.
The first two are CRM-attributed, compared against the prior full year on the same account. The benchmark in the third is the client’s own corporate program: my figure counts completed online bookings tracked in Google Ads and theirs counts confirmed visits in their medical-records system, so read that gap as directional rather than exact. All four are broken down in full further down the page.
Case 01
An industrial manufacturer · CRM-attributed, full year against full year
The problem. Spend was climbing and the platform reporting looked healthy, but a conversion in the ad account and a customer in the business were not the same thing. Nobody could say with confidence which campaigns were producing deals, which meant nobody could defend the budget internally either.
What I did. Rebuilt the segmentation around what the CRM said was closing, in place of what the platforms were claiming. Cut the segments that were not converting, moved the money into the ones that were, and switched the reporting to closed deals and cost per customer. The plan going into the year was never to spend more.
Cost per lead
Deals closed
CRM-attributed, full-year comparison against the prior full year.
Case 02
The same industrial manufacturer
The problem. Phone leads were landing in the CRM with no source attached to them. Sales worked the calls, some of them closed, and none of it counted toward paid search. The program was being judged on a lead number that was too low, and every decision built on that number was slightly wrong.
What I did. Pulled the call records out of the advertising platform and matched them against CRM records by area code and timestamp, one at a time, across an eighteen-month window. Tedious, and it is the only way to prove a call belongs to a campaign when the CRM never captured the source.
Pulled from the advertising platform, per call.
Against CRM records on area code and timestamp.
Re-attributed across an 18-month window.
Case 03
An urgent care group with several locations · Verified at event level
The problem. The account was reporting about 4,000 conversions a month. Most of them were page views and clicks on driving directions. Genuinely booked appointments were mixed in with everything else, so nobody could answer the only question that mattered: how many people actually booked.
What I did. Traced the real booking-confirmation event at data level, checked that it fired only where a booking had genuinely happened, and rebuilt the reporting on that single event across every one of its locations. The inflated count went away and the number that replaced it was smaller and true.
What the account reported each month
One caveat on that comparison. My number counts completed online bookings tracked in Google Ads. The corporate program’s benchmark counts confirmed visits in their medical-records system. Those are two different events being counted, so read the gap as directional rather than exact.
Case 04
A technical consultancy
The problem. Their buyers are attorneys and insurance adjusters, people who pick an expert by asking someone they trust. Increasingly that someone is an AI assistant. The client wanted to know whether they were in those answers, and no existing report they had could tell them.
What I did. Wrote eight questions their buyers actually ask, then ran every one through ChatGPT, Perplexity, Gemini, and Google’s AI Overviews. Fresh session each time, and never the client’s name first, because a model will talk about any company you introduce it to.
Measured June 2026. AI answers shift over time, so this is a snapshot of that month and not a permanent state.
On several runs the engine searched the client by name, reviewed their history, weighed them against a competitor, and left them out of the answer anyway, which rules out awareness as the explanation. The full method, and what fixes it, is on the AI Visibility Audit page.
Shorter findings
These are diagnostics and infrastructure fixes, well short of outcomes I bill against. They earn their place because they unblock the numbers above.
On the same manufacturer, the ad platforms' own conversion counts ran 63% and 44% above the leads that actually reached the CRM over the same three months. Nothing about it looked broken, which is why it had not been caught.
An inherited Shopping campaign could only advertise a quarter of the catalogue, because of a product list set years earlier and never revisited. Found by opening the account and reading the settings, which is what the first month is for.
One landing page for a leadership development firm, after FAQ and metadata fixes. Its key query went from 0.17% to 3.1%, roughly eighteen times.
A design and architecture firm had exactly one URL indexed. After the sitemap was repaired, several hundred.
Relationships
I am the person in these accounts every week. The work described on this page was designed and executed by the same person you would be talking to.
Send a month of spend and what your CRM recorded against it. That is enough to show you where the two stop agreeing.
Book an intro call