Case #036: Finding the one source that mattered

A mobile content offer ran across one big traffic source at a blended 12% ROI — the kind of number that's neither alive nor dead. The blend was lying to me. Buried inside were sources printing money and sources lighting it on fire, averaging out to mediocre. Here's the source-isolation playbook that found the gold.

— Demand source-level reporting before you launch: if a network won't break performance down by sub-source, you're flying blind by design. I only run where I can see the breakdown.

— Set a per-source budget cap on day one: each sub-source gets a fixed $50 test allowance. No single source can eat the budget before proving itself.

— Let them run to a sample, then sort hard: at $50 each I had a real read. Top three sources: 40-55% ROI. Bottom eight: negative. The blend of 12% was a mirage.

— Whitelist the winners, blacklist the losers, permanently: I built a clean whitelist of the three winners and a blacklist that carries to every future campaign on that network.

— Re-test the graveyard once a quarter: source quality shifts. A dead source can revive; a winner can rot. I re-check, but cautiously.

Rebuilt on the three-source whitelist only: $2,180 spent, $3,560 back, 63% ROI — from the exact same offer that read 12% when smeared across everything.

The lesson: a blended ROI is an average of heroes and arsonists — until you isolate by source, you're paying the losers out of the winners' pockets and calling the leftovers a campaign.