<b>Case #005: The whitelist that ate the margin</b>

<b>Case #005: The whitelist that ate the margin</b>

A push-traffic crypto offer, $1,600 budget, $1.80 per signup. Standard playbook: run broad, find the converting placements, build a whitelist of winners, then run only those. I did exactly that and watched it die.

The broad phase, days 1 to 4, spent $700 and returned $810. Thin profit but profit. I pulled the top 60 converting site-IDs, killed everything else, and scaled the whitelist hard, expecting the ROI to jump now that I was only buying winners.

It collapsed. Days 5 to 8 on the whitelist: $900 spent, $520 back. Minus 42% on the very placements that had just been profitable.

The trap was survivorship plus auction pressure. Those 60 site-IDs looked great in the broad phase partly by luck — small samples, a few hot conversions each. The moment I concentrated all budget on them, two things happened: the lucky ones regressed to their true mediocre rate, and my own bids inflated the auction on a now-tiny pool of inventory. I was bidding against yesterday's version of myself.

The fix was to whitelist soft — keep 40% of budget on broad discovery permanently, and cap any single site-ID at 15% of spend so no placement could be over-trusted.

Full arc, 15 days: $1,600 spent, $2,290 back. 43% ROI — recovered only after I stopped believing my own winners.

The lesson: a whitelist built on small samples is a list of lucky coin flips — keep discovery running or your winners will regress while you scale them.