[Comment] § 5.6, § 5.7 — Support performance-based and outcome-based advertising campaigns #80

Activity

1. clerongi

   added

   commentGeneral comment on the draftGeneral comment on the draft

   on Oct 22, 2026
2. Sirajmx

   added

   section:5.6-commercial

   section:5.7-guarantees

   on Oct 23, 2026
3. monjerillc commented on Nov 10, 2026

   

   monjerillc

   on Nov 10, 2026

   More actions

   We support this proposal and suggest explicitly considering attribution to the content and context that influence purchases, including organic discovery outside paid advertising placements.

   In agentic commerce, a purchase may follow a recommendation informed by a film, article, creator video, or other original work. Where measurable and appropriately permissioned, the framework should preserve source references so agreed attribution and compensation can recognize contributing creators and rights holders.

   This also matters for the long-term sustainability of the ecosystem. Original content supports discovery and, where used for model training or retrieval, AI capabilities. Sustainable commercial participation helps preserve incentives to produce that content.

   Purchase attribution should remain distinct from permission to use content for AI training or retrieval. Both require clear rules, but a sales-attribution record should not itself imply permission for those other uses.

   Could the proposed attributes accommodate content-source references and agreed allocation of credit across paid and organic touchpoints, with clear measurement evidence and safeguards against double counting?

   This builds on the shoppable media interoperability and verifiable commerce outcomes discussed in our issue #159.
4. monjerillc commented on Nov 11, 2026

   

   monjerillc

   on Nov 11, 2026

   More actions

   To make our earlier suggestion concrete, consider a viewer who saves a cooking show’s shopping context and, a month later, asks an AI agent to purchase the ingredients. The purchase happens through a different interface, but the originating content remains a contributing source.

   At MarkitPlace, we provide content IP owners with shoppability metadata connecting their content and its context to products and shopping actions. We suggest an optional commerce extension that connects:

   • Content provenance: References to the originating work, relevant scene or moment, and rights holder, preserved across appropriately permissioned shopping and agent interactions.
   • Outcome measurement under § 5.7: A qualifying transaction definition, its measurement source, and a referenced methodology covering attribution terms, returns, cancellations, fraud, and duplicate counting.
   • Commercial terms under § 5.6: Transaction-based pricing that references the qualifying outcome and specifies how subsequent adjustments affect billable amounts.

   Preserving a source reference would not automatically establish causation or entitlement to payment. It would provide evidence that agreed attribution and compensation rules can evaluate, including where both paid and organic touchpoints contribute.

   For #80, the key requirement is that an outcome-based offer can reference both its measurement methodology and the supporting provenance. We propose developing the detailed commerce metadata and transaction definitions separately, linked to #159 and coordinated with #38, while keeping consumer identity and shopping history out of public proposal records.
5. franks33 commented on Nov 11, 2026

   

   franks33

   on Nov 11, 2026

   More actions

   Commenting as: Adgentek - buyer (AdCP-native buy-side / ORCA) and seller (publisher ad server / CTV path).

   Support #80. Buying agents already plan against declared outcomes (CPA, CPO, ROAS), not only against impressions or completion rate. PR #146 is the right place to land this: CPA is already a core cost_method, with action_type, attribution_window, and action_source. Three gaps still block a real outcome buy:

   1. Core action types. Don't invent a new list. Use the event_type enumeration from IAB Tech Lab Event & Conversion API (ECAPI) 1.0, commonly called CAPI, as the core action_type vocabulary, pinned to a version. That already includes the outcomes a buying agent actually prices (purchase, install, start_trial, subscribe, generate_lead, sign_up, and the rest of the standard and additional events). Not every event is a billed unit — page_view and ad_impression usually aren't — but the commercial term should name which ECAPI event is counted, not mint a parallel label. ECAPI's custom / custom_event already covers anything outside the enum.

   2. Price, target, and guarantee are different. A CPA rate, a CPA cap (at_most), and a minimum number of actions are not the same thing. Keep the rate on pricing[]. Let committed_metrics[] in PR #147 carry an outcome metric (actions, or CPA as at_most) with the same measurement-source binding as [Change] § 5.7: Bind each committed metric to the measurement source that settles it #38. A makegood of extra impressions is usually the wrong remedy for a missed outcome.

   3. Handoff. PR #145 maps a Deals API deal from a CPM entry only. If CPA is a core method, say how that deal is handed off, or state that CPA deals are out of scope for 3.0 so implementations don't invent a second mapping.

   On OQ-3 / #4: do not normalize CPA against viewable_rate or completion_rate in the spec. Standardize the inputs (metric, unit, source, window). The buying agent scores unlike proposals against the advertiser's declared outcome.
6. KonstantinMirin commented on Nov 11, 2026

   

   KonstantinMirin

   on Nov 11, 2026

   More actions

   Thanks, @franks33.

   1. Agreed. § 5.6, § 5.6.5 to § 5.6.7 — Pricing: cost methods, currency, cancellation, and exclusivity (#94, #98) #146 now takes the ECAPI 1.0 event_type values as the core action_type terms, so a CPA entry can name purchase, sign_up, or any other ECAPI event except custom (§ 5.6.5, ECAPI terms commit). An action outside the list is an extension term that names itself, such as shop.example:store_visit.
   2. On a CPA entry, a minimum number of actions is a commitment with basis: units, and § 5.7, § 5.7.1 — Settlement: who measures, the metric terms, and the terms in force (#93, #95) #147 now binds the count: counted_by MUST be the entry's action_source (§ 5.7.1, CPA source commit). Under guaranteed delivery, a missed minimum is a shortfall, settled by shortfall_remedy, which includes credit. On a line item priced otherwise, an action minimum or CPA cap needs an extension metric term.
   3. § 1.1, § 5.4, § 5.9 — Handoff to OpenDirect and the Deals API (#132, #133) #145 maps only a CPM entry, and says other cost methods have no mapping yet (§ 1.1). If the working group prefers, we'll add an out-of-scope sentence to § 1.1, § 5.4, § 5.9 — Handoff to OpenDirect and the Deals API (#132, #133) #145.

   Analyzed and generated with the help of Claude, fully reviewed by me.
7. monjerillc commented on Nov 12, 2026

   

   monjerillc

   on Nov 12, 2026

   More actions

   Thanks, @franks33. We support reusing ECAPI’s versioned event vocabulary and keeping the price, target, and guarantee distinct.

   For the commission model in #159, only a completed, retailer-confirmed purchase that meets the agreed attribution rules earns a commission. Views, clicks, sign-ups, and checkout starts can still be reported as part of the shopping journey. The agreement should also specify how cancellations, full or partial returns, fraud, and duplicate reports affect the amount owed.

   Could #146 pair the existing ECAPI purchase event with a structured, versioned reference to the rules that make it eligible for commission? That would let a buying agent identify the billable event and the rules governing it, including what evidence confirms the purchase, which content or context gets credit, and how later adjustments affect payment.