Google Said It Lost 30% of the Leads. Counting Them by Hand Showed 10% More.
Confidential High-Ticket Services Brand, a high-ticket services brand generating leads on Google Search and Meta at the same time, with Google as the larger channel.
The Setup
This brand sells considered, high-value projects, so leads are the unit that matters. Google Search carries most of the spend, with Meta running alongside it. Search runs on target ROAS, which is an unusual choice for lead generation. The media buyer has stopped treating the ad platforms' own attribution as the answer, and checks the real lead rows directly.
The Break
In a single week, with no campaign change, Google's cross-channel report said leads had fallen about 30%. The actual lead rows said the opposite: about 10% more leads than the week before.
The report wasn't only wrong in total. Inside it, you could watch credit move. Conversions credited to Search dropped by about 60 while conversions credited to Meta rose by about 70, with some of the gains appearing in regional segments that had never produced anything before. Net result: a small real gain. The platform was reporting a collapse. What had actually happened was the same leads changing label.
The Diagnosis
The mechanic is the attribution layer itself. Google's cross-channel view assigns each lead to a channel after the fact, and it reassigned the same leads from Search to Meta mid-week, with no change to targeting, budget or structure behind it. The leads didn't move. The label on them did.
We can say precisely what is and isn't known. The reported decline didn't match the real lead count, credit shifted between channels inside the report, and nothing in the account had been changed. Why the report moved isn't something either platform can prove from the inside: it may be Google no longer claiming Meta-originated traffic, or it may be Meta influence that never shows up as a click. Both would produce the same report.
The Risk
A buyer who trusted the report would have cut or restructured a campaign that was performing, in response to a loss that existed only on a dashboard. This buyer held steady, because he was counting real rows. Most teams aren't, and as the account heads into its slow season, any genuine softness will be tangled up with seasonal movement that no single platform can untangle.
Why It Matters
Two ad platforms each report their own share of the same leads, and each report is a model. When the model reassigns credit mid-week, the dashboard shows a crisis while the business shows growth. A rented dashboard has no reason to tell you its number just moved on its own, because a stable-looking report is the product.
Not trusting the report isn't the same as replacing it. Counting rows by hand once a week is a person acting as the data layer. A client-owned record of every lead, tied to its actual source, does that job continuously, and it doesn't change when a platform changes its mind. That's the difference between renting a view of your business and owning the record of it.