Meta Claimed 1,996 Conversions. Real Orders: 1,406.
Confidential DTC Brand, 8-figure DTC health and wellness brand running high-volume quiz-funnel ad campaigns on Meta, spending well into six figures a month across dozens of active ad sets.
Ad-level reconciliation that matched every claimed conversion to a verified first-party order.
Killed the duplicate pixel and built a daily Slack report to catch the gap live.
The Setup
This brand runs quiz funnels. A prospect clicks an ad, takes a quiz (menopause symptoms, weight loss goals, hormone health, whatever the funnel targets), and lands on an offer. It's a volume game: hundreds of ad variants live at once, new batches launched weekly, media buyers making spend decisions every single morning based on what the numbers say worked yesterday.
Which means the numbers have to be right. If they're not, a media buyer scales the wrong ad, kills the right one, and burns real dollars finding out which was which.
The Break
It started on one campaign. The brand's own backend showed 8 conversions on a specific ad. Meta's Ads Manager showed 14. Six phantom conversions, one ad, one day.
Then it got worse. A different campaign: backend counted 78, Meta counted 78, dead even, for weeks. Then the team launched a batch of new ad variants at midnight, and the gap exploded. The next day's numbers: 187 real orders, but Meta's own dashboard showed two different totals depending on which report you pulled, one of them nearly double the other.
By the time it reached us, one high-spend ad set alone had racked up roughly 360 phantom conversions in a single day. Across the account, Meta was claiming 1,996 conversions where only 1,406 could be matched to an actual order. That's a 42% overclaim, concentrated rather than spread evenly: 28 specific ads accounted for 623 claimed conversions against 304 real ones, representing over $21,000 in daily spend riding on numbers that were roughly double what actually happened.
The brand's head of growth put it plainly in our shared Slack channel: "There are tons of ads in these accounts and there will be a lot more. We can't expect the media buyer to look in the account and then go back to our reporting and back to the account and find the sub id on hundreds and hundreds of ads each day."
He was right to be frustrated. Nobody should have to reconcile hundreds of ads by hand every morning just to know if yesterday's spend worked.
The Diagnosis
Here's the part that actually mattered: before pointing at anything else, we audited our own tracking first. Every claimed conversion from Meta got matched, ad by ad, against verified first-party orders using a 7-day last-click window in the brand's own data. Our own system wasn't double-sending anything. That ruled us out in a single pass.
Once that was off the table, the pattern pointed one direction: two Meta pixels, firing in parallel, on the exact same campaign. The brand had set up a duplicate pixel months earlier for a reason that made sense at the time, then quietly stopped using one of them without ever actually turning it off. It kept counting in the background the whole time, invisible unless you were matching claimed conversions against real orders and asking why the gap only showed up on two specific campaigns.
Separately, about a third of the flagged ads had zero matching UTM traffic at all, a naming drift between the ad name in Meta and the tracking parameter, left over from a bulk upload. Easy fix once it's visible. The hard part was ever seeing it.
The Fix
We didn't hand the brand a spreadsheet and wish them luck. We built the answer directly into their workflow, and then we used it to close the loop live.
Every morning at 6:15 AM PT, an automated report posts straight into the client's Slack: the previous day's ad-level performance, Meta's claimed conversions next to our verified, order-matched numbers, side by side, for every active ad. Any ad running a sustained ~2x gap gets flagged automatically. That flag is what surfaced the two problem campaigns in the first place.
From there it took one working session to find the actual cause. Rule out our own tracking, isolate the pattern to two campaigns, trace it to the duplicate pixel, and turn it off. Done, live, on the call. No multi-day forensic exercise, no waiting on a ticket queue. The brand's own data team could have found this eventually. They have smart people and they were digging hard. But it would have taken days of manual cross-referencing, repeated every time a new ad batch launched. We turned it into a number that shows up automatically every morning, and a fix that happened in real time once that number pointed somewhere specific.
Where It Stands
The redundant pixel is off. The daily reconciliation report is live and running in production, and the next few days of data will confirm the gap closes for good now that there's only one pixel reporting. We're also auditing a fresh batch of UTM parameters the brand's own analytics team sent over, just to rule out anything else hiding in the account. That's the real distinction infrastructure gives you: not a promise that nothing ever breaks, but proof of exactly what broke, why, and when it got fixed, down to the ad ID and the dollar.
Why It Matters
This is the trust deficit in its purest form: marketing sees one number, the backend sees another, and the media buyer is left guessing which one to trust with real budget. Most tools would have handed this brand another dashboard. We handed them an audit trail, matched at the ad level, delivered automatically, with the exact 28 ads and the exact dollar figure behind every discrepancy named.
Renting a dashboard means hoping the vendor's math lines up with your bank account. Owning the infrastructure means proving it, ad by ad, every single morning, without anyone having to ask.