−38% CAC, 2.4× spend

The brand’s new-customer cohorts paid back on order 3. That single fact made platform ROAS the wrong dashboard, and platform ROAS was the only dashboard the account had ever run on. The CEO was throttling spend during acquisition peaks to protect margin, which meant leaving growth on the table in the only months that mattered for this vertical.

Three agencies had held the account before us. All three ran it to platform-reported ROAS and killed anything under 1.8×, and none of them checked what a cohort did after day 30. When we rebuilt measurement around 90-day cohort payback, the account’s supposed worst campaigns turned out to carry a 3.1× payback by day 90, because the subscribers they acquired stuck around 2.2× longer than branded-search buyers. The 1.8× rule had been systematically shutting off the best acquisition sources in the account for two years of agency handoffs.

The person who built the audit ran the account, and the rebuild started in week one. Once the payback view cleared it, 38% of budget moved out of branded search, where the brand was paying to intercept people already typing its name, into non-brand prospecting. The audit had flagged 4 underperforming evergreen ads; those died the same week. By week four, half of the evergreen creative had been replaced with founder-led scripts tested against three new hooks, on a weekly release cadence instead of ad hoc drops.

The dashboard changed too. 90-day cohort payback replaced platform ROAS as the number the CEO saw every week, which ended the throttling reflex. When payback held through a spend increase, spend increased again. The decision loop got boring, which is what a decision loop should be.

Ninety days in, spend was running 2.4× the day-zero level, CAC was down 38%, and blended ROAS was up 62%. No magic ad. The account stopped killing the campaigns that were working and funded them instead.