Meta CAPI implementation and a creative testing framework for blended CAC

A D2C subscription product relied on Meta as its primary acquisition channel — pixel-only, with all purchase events firing from the thank-you page. Post-iOS-privacy-changes, reported performance had degraded and delivery optimisation with it: Meta was seeing a fraction of real conversions, and the fraction was biased toward non-EU, non-Safari users.

At the same time, creative had stagnated: a handful of long-running ads, refreshed ad hoc when someone "felt" fatigue, with no record of what had been tested before.

First, signal. We implemented the Conversions API server-side, fed from payment webhooks rather than page views, with event-ID deduplication against the existing pixel so the two streams reinforced instead of double-counting. Hashed emails and click IDs were passed on every event, and event match quality was monitored monthly as a managed metric. Trial starts and first payments became separate optimisation events, letting campaigns optimise toward payers rather than trial tourists.

Then, creative. We installed a weekly testing cycle: concept hypotheses drawn from customer research, a hook × format variant matrix, controlled cost-cap test budgets, and a creative ledger recording every verdict. Broad targeting carried delivery; the creative did the segmentation.

Blended CAC — not platform-reported ROAS — was the steering metric throughout, cross-checked with periodic holdout-based incrementality reads.

RESULT METRICS PENDING CLIENT SIGN-OFF FOR PUBLICATION.