Front-loading a deal test inflates your read by ~25%

Front-loading a deal test inflates your read by ~25%

The advice: blast all your traffic at a new deal to test it fast. Speed corrupts the signal.

The distortion, observed across ~15 deal tests:
— A new affiliate often gets a temporary boosted CPA or relaxed gate for the first 30 days (operator land-grab pricing).
— Dumping volume into that window books a cohort at terms you won't keep, overstating EPC by ~20-30%.
— Worse, a fat early cohort with thin month-2 retention sets a bad blended quality baseline that follows you into renewal.

The correction: season the deal. Send 200-300 FTDs, hold, and read month-2 retention and NGR/FTD before scaling. You're not testing whether the deal pays today — you're testing whether the cohort retains, and retention is unknowable in week one.

A controlled ramp also keeps the operator's risk team calm; a sudden 10x spike is the classic trigger for a manual fraud review and a held payment.

Benchmark of the day: judge a deal on month-2 cohort retention from a 200-300 FTD seed, never on week-1 EPC.