Google Called It a $25 Customer. It Was Actually $125.
Viva Raw Pet, a subscription e-commerce brand selling premium raw dog and cat food, recurring monthly orders.
First-click order data that isolated new customers from returning subscribers.
Rebuilt the campaigns around a verified new-customer-only event.
The Setup
This is a subscription business, the kind of company where most of the money made in any given month comes from customers who already signed up months ago and keep reordering. That matters, because it's exactly the condition under which Google's Performance Max looks the most impressive and lies the hardest.
The account's Performance Max campaign was optimizing toward a standard "purchase" event. On paper, the results were outstanding: about $44,000 in spend over two months, reportedly acquiring new customers at roughly $25 each. A number good enough to make anyone want to pour more budget in immediately.
The Break
The problem: "purchase" means any purchase, not "new customer," including the thousands of existing subscribers reordering the same product they'd already been buying for months. Performance Max doesn't distinguish, finding any sale anywhere in the account's orbit and claiming credit for driving it, whether it touched that customer for the first time or the four-hundredth.
Isolating the number that actually mattered, new customers only, verified through first-click order data instead of Google's own "purchase" event, changed the picture completely. Of roughly 2,000 purchases the campaign had taken credit for, only 339 were genuinely new customers. Real cost to acquire one: about $125-130. Five times the reported number.
The gap wasn't evenly distributed either. The dog food line of the business was especially bad. One segment was paying roughly $223 per new customer, another as much as $524, hidden inside a blended account average that looked fine because the cat food side of the business (less competitive, cheaper to acquire) was propping up the average.
The Diagnosis
Performance Max is, structurally, a black box: you hand it a conversion event and a budget, and it goes and finds whoever it thinks will trigger that event again, using every signal it has. If the event you hand it is loose, "purchase" rather than "verified first-time buyer," it will spend real money re-marketing to your own existing customers and report it back as new-customer growth. The sale it reports is real. The identity of the buyer usually isn't the one Performance Max implies.
The tell was hiding in plain sight the whole time: spend and revenue moved independently of each other. Cut the budget in half, and the reported results barely moved, because the "customers" being counted were going to buy anyway, subscription or not. That's the signature of attribution counting demand it didn't create.
The Fix
The account was rebuilt around one change: stop optimizing toward "purchase," and start optimizing toward a verified, first-click, new-customer-only event, imported directly from first-party order data rather than Google's own event tracking. Two focused campaigns replaced one blended one, each measured only against real, deterministic new-customer counts for its own product line.
The shift showed up immediately. In the week right after the change, the dog food campaign's cost dropped 21% while new customer purchases quadrupled in the same week. Nothing about Google got smarter. The account was finally asking it to solve the actual problem instead of an easy, misleading one.
Where It Stands
Fully verified, cost per newly acquired customer landed at roughly $100-145 across both product lines, a real, sustainable number the business could actually plan around, replacing a fake $25 figure that was never going to survive contact with the P&L. The account continues to scale on the corrected event, and the gap between the fake number and the real one stopped being a mystery once it was visible.
Why It Matters
A platform reporting a $25 cost to acquire a customer, when the real number is $125, is the difference between a metric that gets you promoted and one that gets you fired, not a rounding error, on the exact same account, depending only on whether anyone verified it. Google didn't need a single real click from a stranger to claim the win, only a loose conversion event and a subscription business with enough existing customers to quietly credit itself with. Verified, first-click, deterministic tracking is the only thing that makes the gap visible before the budget gets burned finding it the hard way.
That visibility only holds up because the tracking lives in infrastructure Viva Raw owns, not a rented reporting layer Google controls end to end. The same platform doing the counting has no incentive to flag when its own definition of 'purchase' is quietly wrong.