<b>One campaign was +80%, one was -30%, the report said +12%, and I almost killed the wrong one</b>

<b>One campaign was +80%, one was -30%, the report said +12%, and I almost killed the wrong one</b>

Bundling for convenience nearly cost me my best performer. Averages are where good campaigns go to hide.

<i>The setup</i>
— Two offers, same vertical, lumped into one campaign for "simpler reporting."
— Blended ROI: +12%, ~$1,400/day. Mediocre, considering killing it.

<i>The move</i>
Before pulling the plug I split the report by offer. The blend was lying to both of us.

<i>The numbers (illustrative)</i>
— Offer A: ROI +80%, but small volume.
— Offer B: ROI -30%, eating most of the budget and dragging the average to mush.
— Killing the "campaign" would have murdered the +80% winner along with the loser.

<i>The lesson</i>
Never make kill/scale decisions at a level that blends profitable and unprofitable units. Aggregation is for accounting, not optimization. Every layer of averaging you optimize at, GEO, offer, placement, hour, is a layer where a winner can be invisible and a loser can hide.

<i>What I'd do differently</i>
Structure campaigns so the smallest decision unit is also the smallest reporting unit. One offer, one campaign. The "simpler reporting" I wanted was the exact thing that almost made me kill an 80% ROI machine.