2.1→4.3 ROAS, blended

Situation

A US womenswear brand at roughly $60K per month in revenue had stopped growing. Meta ROAS had declined for three consecutive quarters while ad spend stayed flat. Two agencies had already rotated through the account, each adding campaigns without removing any — by the time we audited it, thirty-one campaigns were competing for the same audience with no coherent structure.

Diagnosis

• The account was starved of signal: conversion volume was split across so many ad sets that nothing exited learning phase.
• Pixel-only tracking was losing roughly a third of conversions after iOS, so the algorithm optimized on incomplete data.
• Creative had quietly fatigued: the top three ads were over five months old, and CPMs had risen 40% against them.

System built

We consolidated thirty-one campaigns into three, each with a defined role: prospecting, retargeting, and retention. The Conversions API was rebuilt with event deduplication, restoring the lost third of the signal. A weekly creative cadence went live — new concepts tested against controls, winners spawning variations, everything retired on schedule before fatigue.

Scaling followed rules, not moods: budget increases were tied to a blended CPA ceiling derived from the brand's actual margins, checked weekly against revenue reconciled in the analytics layer.

Results

Blended ROAS doubled in seven months while monthly paid revenue more than doubled. The structure has held through two subsequent scaling pushes without resetting learning.

Before / After

Case studies are anonymized to protect client confidentiality. Metrics reflect actual account performance over stated periods.