LabPlanPlan

CPM now, CPA when the payload is honest

Beeswax CPA on this seat requires cpa_target, cpa_event, and max_bid. Live tests run as CPM with conversion measurement attached. CPA pacing exists in the catalog. It has not been write-verified on a live line item.

Published 2026-09-09 · Search job: CPA optimization programmatic

This page is a plan, not a result. We have not judged this test yet. The protocol is here so the next flight can be read. Numbers below are setup notes or prior adjacent evidence, never a claim that this test already ran.

CPA optimization on programmatic display is available as a bidding strategy on this seat. It is not how live Adprime Zero tests run today. Live tests run as CPM, with conversion measurement treated as a separate meter, until the CPA payload is write-verified.

GET /rest/v2/ref/bidding-strategies lists CPA (id 5, paced=false) and CPA_PACED (id 9, paced=true). The CPA value schema requires cpa_target, cpa_event, and max_bid. The paced sibling requires the same three fields. That is catalog fact, not a live line-item proof.

Why we will not flip the strategy yet

The CPM pacing mistake taught the cost of treating a catalog row as a shipped control. CPA_PACED exists and is marked paced. The CPA payload shape has never been validated against a live line item on this product path. Conversion bidding stays deferred until corrected click reporting is reconciled with conversion signal. Adding CPA pacing without that smoke would be guessing.

The mapper that selects CPM_PACED for even delivery deliberately does not route CPA or ROAS campaigns. That is a plan boundary, not a claim that CPA is impossible.

What "honest" means here

Honest means the line item asks the exchange to bid on a CPA basis only when all three required fields are real: a numeric target, an event the seat can fire, and a max bid. It also means conversions are a measured meter, not a renamed click.

Until that write is proven, do this:

  1. Traffic as CPM (or CPM_PACED when the buyer wants even delivery).
  2. Attach conversion events and keep them as their own column.
  3. Judge traffic on IVT and click-to-page-view, not on a CPA the exchange was never asked to hit.

This is a plan, not a Lab result. The next judged step is a paused write-and-restore of CPA_PACED with real cpa_target, cpa_event, and max_bid, then restore the object. Do not publish a CPA win from a CPM campaign that happened to record conversions.