They Scaled Meta to $1M and Never Touched Google. Google Got 25% More Efficient Anyway.
Eric Javits, a high-end designer hat brand ($300-400 AOV), grandfathered founding client.
Google's own change history, pulled and checked against spend and conversion data, that proved nothing on the Google side had changed at all.
Held Google as a deliberate, untouched control while scaling Meta 66x, then let an eleven-week window and a year-over-year comparison confirm the mechanism before acting on it.
The Setup
On May 14, 2026, the account was restructured around one deliberate constraint: scale Meta aggressively, and change nothing on Google, not a dollar, not a setting, so the Google side could serve as an untouched control. Meta spend on three core ad sets scaled from roughly $15,000 to $1,000,000 over the following eleven weeks, running just 33 ads pulled from a pool of 3,000 the team manually graded for hook rate, hold rate, and message fit, keeping only the 160 that qualified. Cost per acquisition held nearly flat across that 66x spend increase, moving from about $100 to $127, and produced roughly 7,800 additional purchases.
The Break
Google was never edited during the window, confirmed by pulling its own change history: no new campaigns, no budget changes, nothing. Spend moved by $400 total, a 0.15% increase. And yet Google's numbers moved anyway. Conversion value rose 103%, about $700,000, on a 4% increase in sessions. Isolating the account's non-branded Performance Max segment specifically, cost stayed flat, clicks were up only 2%, and CPC was actually down 7%, while cost per conversion fell 25% and conversions rose 25%. Fewer real signals of new traffic, meaningfully more sales, on a channel nobody funded.
The Diagnosis
The account's own search-term report rules out the obvious explanation: filtering non-branded Performance Max conversions for the client's name returns zero results. This wasn't brand search inflating the numbers. Google's automated Search and Performance Max bidding was functioning like a remarketing list sitting underneath a "bidding strategy" label, quietly prioritizing people who'd already been touched by another channel over genuinely new demand. As Meta's remarketing pool grew from the scaled spend, that same finite pool of people searching Google converted more efficiently, not because Google found more of them, but because more of the people it was already finding had already been sold on the brand by Meta first. Google's dashboard reports this as its own campaign improving. The change history says otherwise: nothing about the campaign changed. What changed was who was already primed to buy by the time they searched.
The Year-Over-Year Confirmation
Comparing May 14 through July 30 against the same window in 2025, combined Meta and Google spend was $255,899 lower this year (Google down $241,446, Meta down $14,453), while gross sales were up 11%, roughly $784,000 in incremental revenue, off a $7.13M trailing base. Spending less on both channels combined and making substantially more, in the same year Meta went from a rounding error to the majority of the budget.
The account deliberately did not reduce Google spend even after the pattern became visible mid-window, specifically to keep the control clean and prove the mechanism held before acting on it.
Why It Matters
This is the same mechanism behind other reconciliation cases in this library, Google's automated products reclassifying externally-generated demand as their own, but the receipt here is sharper because the test was designed as a control from day one, not discovered after the fact in a billing review. Eleven weeks of a genuinely untouched Google account, confirmed by change history, showing measurable gains it did nothing to earn, is close to as clean an experiment as a live ad account produces. The year-over-year comparison closes the loop: less combined spend, more combined revenue, the same year Meta went from a rounding error to the majority of the budget.
A rented dashboard has no reason to flag this. Google's own reporting has every incentive to let a 103% conversion-value gain read as Google's own success, because the platform gets paid either way and the client has no way to check the claim without a change history and a channel-by-channel view neither platform is built to provide together. Owning the layer that watches both channels side by side, on a timeline, is what turned "Google seems to be doing better" into "Google is being handed credit for work Meta already paid to do," and gave the account a reason to trust the shift enough to keep scaling it.