This is a loss post-mortem, and the worst kind: the campaign looked profitable the entire time it was bleeding us dry.
The dashboard showed a finance offer at 60% ROI. We scaled it. We celebrated it. We told the chat about it. For seven days we ran $300/day on a campaign that was, in reality, losing money on every single dollar.
— Reported revenue: $9,540
— Actual revenue (from the advertiser's panel): $5,880
— Real ROI: -7%, not +60%
The gap was deduplication. Our pixel was counting browser conversions and the postback was counting server conversions, and a misconfigured event was double-firing on roughly 38% of leads. The tracker showed two conversions where the advertiser paid for one.
We only caught it on day 8 when the affiliate manager flagged that our claimed numbers didn't match their panel. That conversation saved us — and stung. We'd reinvested $2,100 of "profit" that never existed.
The rebuild was a single source of truth: we killed browser-side conversion counting entirely and trusted only the server postback that matched the payment. Numbers dropped overnight, which felt like a loss but was just the truth arriving late.
— True net loss for the week: -$410
— Cost of the lie: another $2,100 mis-allocated to a loser
The lesson: never trust a conversion number your advertiser didn't also count — reconcile the panel against the pixel before you scale, not after.
The Green Day
@greenday_roi
Case #004: The pixel that lied for a week
This post was published in the Telegram channel The Green Day (@greenday_roi). All its posts are on the channel page.
Open on Telegram