Case #013: The buyers I was paying to ignore me

Case #013: The buyers I was paying to ignore me

A subscription box offer was stuck at a loss that looked like a math error. $4,600 spent, $3,400 back, 26% down, and every metric individually looked healthy. CTR was fine, lander converted, the offer was strong. The loss had no obvious home.

I broke the audience report down by whether the user had visited the site before. The split was the whole story. New visitors converted at 3.2% and were profitable. But a large slice of my budget — about 40% — was being spent re-showing the ad to people who'd already subscribed.

I was paying full ad prices to advertise a subscription to people who already had it. They clicked out of recognition, never converted (they couldn't), and quietly poisoned the campaign's average. The algorithm loved them because they engaged; my P&L hated them because they never bought.

The fix was one exclusion audience: everyone who'd already hit the post-purchase thank-you page, removed from targeting.

The campaign cleaned up overnight:

— wasted 40% of spend reallocated to new prospects
— cost per acquisition dropped from $58 to $26
— $1,100 spent, $1,800 back in the first three days

Full turnaround, two weeks:

— $9,200 spent
— $15,100 revenue
— 64% ROI, from a 26% loss

The lesson: if you're not excluding existing customers, you're buying applause from people who can't buy again — engagement and revenue are not the same audience.