Keep the promise consistent to test platforms, not copy drift.

normalize spend envelopes to make channel performance comparable.

Determine the quality-adjusted decision metric before launch.

Count verification and account overhead as part of channel cost.

Review session behavior and device cuts to find the truth.

Shut off channels with high activity but low-fit behavior.

Double down only where user behavior proves real intent.