Zappush detects whether each purchase is a genuinely new customer server-side and sends only New Customer Purchase as the optimization event to Meta and Google. The algorithms train on net-new buyers instead of returning customers. First-order detection runs on your reconciled first-party profile, not a browser cookie that misses device switches and lapses.
From $59/month · 14-day free trial · Cancel anytime
Diego Flores
First-time buyer · via Meta
+41% new-customer revenue
When every sale fires the same generic Purchase event, Meta and Google optimize for maximizing total conversions. Returning customers are simply the cheapest, highest-probability conversions on the board. The features that predict a repeat buyer (past visits, engagement, cart adds) overlap almost entirely with the features that predict a new buyer, so the algorithm resolves the ambiguity by chasing people who already know and trust you. The pixel misses 30-40% of conversions and cannot reliably tell a first order from a fifth, so even brands that want to optimize for acquisition feed the model a polluted signal. That inflates reported ROAS with revenue you would have captured through email or organic anyway, while accretive net-new revenue stays flat. New-customer ROAS (ncROAS) is the canary. When it drops below 1 you are losing money on first activation. Every cycle the model drifts further toward existing customers, true CAC climbs, and the top of the funnel starves behind a healthy-looking blended number.
At purchase, Zappush checks the buyer against their unified first-party profile and full order history, not just a 7-day cookie, to determine if this is a genuine first order. It resolves new-vs-returning deterministically on order_id and SHA-256-hashed email, phone, and customer ID. A customer who bought last year on another device is still correctly flagged returning.
Zappush emits a distinct New Customer Purchase event through Meta Conversions API and Google's Data Manager, deduped against the browser event via a shared event_id. Returning purchases are suppressed or routed to a separate non-optimized event. You point the ad set's optimization at that single event, so Meta and Google train exclusively on net-new buyers.
Zappush hashes reconciled existing-customer identities and syncs them as Meta and Google exclusion audiences, so prospecting ads stop serving people who already bought. That is the reinforcement the conversion event alone cannot provide. Every order carries a new/returning tag, so you can read new-customer ROAS and new-customer CAC separately from blended numbers and watch the canary metric directly.
Classify each order against the buyer's full first-party purchase history by hashed email, phone, and customer ID, not a decaying cookie. First-order detection stays accurate across devices, long gaps, and ad blockers.
Emit a dedicated New Customer Purchase conversion via CAPI and Data Manager, and suppress or split off returning purchases. Meta and Google train the delivery model exclusively on net-new buyers.
Sync reconciled existing customers as real-time suppression lists across Meta and Google, so prospecting spend stops re-serving people who already bought. That is the reinforcement the event alone cannot provide.
Tag every order new or returning at the source, and surface new-customer ROAS and new-customer CAC separately from blended metrics. The canary metric is visible before blended numbers hide the erosion.
Ship the new-customer event server-side with SHA-256 identifiers and a shared event_id, keeping Event Match Quality at 8+ and preventing the browser copy from double-counting the same order.
REPEAT-PURCHASE DTC
Coffee, supplements, and skincare brands run high repeat rates, so returning buyers are the cheapest conversions on the board. Zappush checks each order against the full first-party purchase history and emits New Customer Purchase only on genuine first orders. Point the ad set at that event and prospecting trains on net-new buyers.
CROSS-DEVICE
A customer who bought last year on mobile and returns on desktop looks brand new to a browser cookie. Deterministic matching on order_id and SHA-256-hashed email, phone, and customer ID resolves them to one profile, so the order is correctly flagged returning. Existing-customer exclusion lists then keep prospecting ads off them.
AGENCY REPORTING
Blended ROAS hides acquisition erosion behind revenue from customers you already owned. Every order carries a new or returning tag resolved at the source, so new-customer ROAS and new-customer CAC report separately from the blended number. When ncROAS falls below 1, you see it in the account rather than a quarter later.
Start free in minutes, or see it live on a call.
From $59/month · 14-day free trial · Cancel anytime
