Skip to content

Proof

Verified at source, or not stated

Research and case files

Anyone can show you a number.

01

A percentage proves nothing

"+340% growth" with no starting figure is arithmetic theatre. Three bookings to ten is +233%.

02

A logo proves someone signed

A logo wall says a contract existed. It says nothing about what happened after the signature.

03

The work is the proof

What proves a firm can help you is watching it find the thing everyone else missed.

So here is the work, with the method attached — and where it is unfinished, the page says so, rather than waiting for a number that flatters us.

The file

One case, one study, one client. No borrowed statistics.

Every figure below traces to a dated scan or a live account we can produce on request. Nothing here is an industry benchmark wearing our name.

01
The event that poisoned two ad platformsA conversion event quietly redefined to fire on page dwell. Both platforms optimized against a signal that meant nothing — and both reported success for doing it.
Published
02
Does map rank predict the AI answer?Twelve businesses, twelve trades, twelve cities. Both channels, same grid, same day. Selection rule and kill condition written before the first scan ran.
In progress
03
An Advisory engagement, namedOur first Advisory client, named with his permission. Of the seven firms we surveyed, not one names a client at all.
Named

Case file 01 · Attribution

Both platforms said the campaigns were healthy. Both were wrong.

An account we operate was optimizing against a conversion event that had quietly stopped meaning what it once meant. No dashboard could have shown it. Verified

Before InitiateCheckoutscheduler interaction
After InitiateCheckoutpage dwell
Cost per acquisition, as reported $237.00 What both platforms believed, and celebrated.
Cost per acquisition, actual $324.65 Found in the page source, not the reports.

No platform shows you the definition of your own conversion event — someone has to open the page and read the code that fires it. That is half an hour of work almost nobody does, and it is the whole diagnosis. The repair: rebuild the event, reset the learning, demote the polluted signal. This is now the first thing we check in any account we take on.

Study 02 · Twelve markets · Pre-registered

Does your map rank predict what the AI says about you?

Every guide written about AI visibility assumes an answer to this. None of them tested it. So we are testing it: twelve businesses, twelve unrelated trades, twelve cities, measured on Google Maps and inside the AI answer. The Maps half so far:

Immigration law · Houston1.96 avg rank
On the grid49 of 49
Local voice85.71%
Roofing · Tampa6.04 avg rank
On the grid46 of 49
Local voice51.02%
Wedding photography · Richmond3.98 avg rank
On the grid49 of 49
Local voice46.94%
Chiropractic · Boise6.15 avg rank
On the grid40 of 49
Local voice26.53%
Personal training · Portland10.67 avg rank
On the grid3 of 49
Local voice2.04%
Bookkeeping · Charlotte11.35 avg rank
On the grid23 of 49
Local voice0.00%
Landscaping · Sacramento12.57 avg rank
On the grid23 of 49
Local voice0.00%
HVAC · Kansas City13.18 avg rank
On the grid11 of 49
Local voice0.00%
Study status · 09 July 2026

Eight markets measured on Maps. Three measured on both. Nothing concluded.

In all three markets measured on both channels, map standing and AI standing moved in opposite directions — and the business that owns its map most completely was not the one the AI named most often. That is three data points, and three data points are an anecdote. We are not calculating a correlation on three markets, and neither should anyone else. Four earlier markets will be re-pulled before publication, because a stale figure is not a figure.

3 complete on both channels · 5 awaiting the AI half · 4 dated, awaiting re-pull

Selection

We do not choose the businesses

The subject in each market is the first result the scanning tool returns for that trade in that city. Mechanical, reproducible by a stranger, and designed so we cannot pick the ones whose numbers flatter us.

Instrument

The same grid every time

Forty-nine points, eight-mile radius, centred on the city. The keyword is the trade and the city, nothing more. Both channels, same day. Anonymized by trade and city.

No exclusions

The awkward results stay in

One subject turned out to be a service-area business with no storefront. Another appears at only eleven of forty-nine points. Neither was swapped for a tidier example.

The kill condition

We wrote down what would prove us wrong

If the twelve markets show the two channels moving together, our central claim is false, we publish that, and the line comes off the site. Recorded before any result was read.

Local Falcon · 7×7 grid · 8-mile radius · 09 July 2026 · anonymized by trade and city. Share of local voice counts only top-three placements — which is why a business can appear at 23 of 49 points and still score zero. Being seen and being chosen are different measurements, and most tools report only one of them.

File 03 · A client with a name

Not an initial. Not a stock photograph. A person.

"I spent months trying to work out where my business had broken. I never would have found the reason on my own — and it turned out to be something I had done myself."

Mark Eadie

204 Photography · Winnipeg, Manitoba · Advisory client since 2026 · Named with permission

7 competing firms we surveyed, publishing dozens of testimonials between them.
0 of them name a client at all.

The fault in his business was one he had introduced himself, and he could not see it from the inside. Nobody can. That is not a failing — it is a structural fact, and it is the entire argument for having someone outside the account go and read the code.

Start

The same scrutiny, pointed at your account.

Twenty-five minutes, free, and we will tell you straight what we can see from the outside. If it is worth going further, the Growth Diagnostic is the next step — and it credits in full against month one.

Book the free 25-minute call

Or write to the principal: ryan@theorymediaco.com. Read in person, answered in person.