feat(schemas): unify outcome measurement into the row-shape vocabulary (closes #3857)

[bokelley]

Summary

Closes #3857. Collapses outcome_measurement (a separate field with free-form strings) into the unified row-shape vocabulary established by #3576 / #3848 / #3877.

The conceptual case: outcome_measurement was always the same category as the seller-reported attribution-derived metrics already in delivery-metrics.json (conversions, conversion_value, roas, cost_per_acquisition). Two surfaces describing overlapping subject matter, split because the protocol predated the unified vocabulary. With the qualifier slot proven generalizable (#3877's completion_source joining viewability_standard), the collapse is mechanical.

What changes

New schemas:

• enums/attribution-methodology.json — closed enum: deterministic_purchase | probabilistic | panel_based | modeled

Vocabulary additions (available-metric.json):

• incremental_sales_lift, brand_lift, foot_traffic, conversion_lift, brand_search_lift
• No attributed_sales — that's conversion_value with qualifier.attribution_methodology: "deterministic_purchase". The unified pattern handles the methodology split via qualifier, not parallel metric IDs.

Qualifier slot expanded at all five sites (committed_metrics, buyer-side committed_metrics, metric_aggregates, missing_metrics, performance-feedback.metric) with two new keys:

• attribution_methodology ($ref attribution-methodology.json)
• attribution_window ($ref duration.json) — first object-valued qualifier key. Establishes that qualifier values can be structured, not just string enums. The join-on-(metric_id, qualifier) pattern handles same-metric-different-window correctly.

Deprecation:

• core/outcome-measurement.json — title and description marked deprecated; migration table in description
• core/product.json outcome_measurement field — deprecated, points at the new pattern
• Schema retained for one-minor backwards compatibility; removed at the next major

Doc updates:

• commerce-media.mdx — new pattern alongside legacy field for the transition window
• media-products.mdx — field description deprecation note
• create_media_buy.mdx — qualifier section adds attribution_methodology + attribution_window with conditional-required semantics
• get_media_buy_delivery.mdx — qualifier vocabulary names all four keys

Migration

Retail-media sellers using outcome_measurement continue to work for one minor. New implementations:

{
  "reporting_capabilities": {
    "available_metrics": [
      "conversions", "conversion_value", "roas",
      "incremental_sales_lift", "units_sold"
    ]
  }
}

…and on commit, qualifier carries methodology + window:

{
  "scope": "standard",
  "metric_id": "conversion_value",
  "qualifier": {
    "attribution_methodology": "deterministic_purchase",
    "attribution_window": { "interval": 14, "unit": "days" }
  }
}

Seller-as-measurement-vendor remains the dominant retail-media topology — vendor identity is implicit (the seller) when no separate performance_standards.vendor BrandRef is set. No retailer adoption gap: the get_adcp_capabilities.measurement.metrics[] surface already exists; retailers just populate it with their existing metric definitions.

What's deferred

reporting_frequency / reporting_format (the outcome_measurement.reporting field's dimensions) move to a follow-up extension on reporting_capabilities — they're a property of the seller's reporting infrastructure (daily API, weekly dashboard) rather than a per-metric concern. Existing outcome_measurement.reporting payloads continue to work for one minor.

Test plan

• Schema validates (build:schemas during commit hook)
• TypeScript build passes
• Doc renders cleanly

Related

• Restructure outcome-measurement.json onto BrandRef + measurement-agent pattern #3857 — origin (this PR closes it)
• feat(schemas): unify metric accountability — timestamped contract + symmetric audit #3576 — established the qualifier slot pattern
• feat(schemas): completion_source qualifier — seller_attested vs vendor_attested completion_rate (closes #3861) #3877 — proved the qualifier slot generalizes (completion_source as second key)
• Add metric_aggregates partition to aggregated_totals — qualifier-aware delivery rollups symmetric to committed_metrics #3848 / feat(schemas): metric_aggregates partition — row-symmetric delivery aggregation (closes #3848) #3876 — metric_aggregates row-symmetric delivery surface
• docs(measurement): add three-layer taxonomy doc #3843 — measurement taxonomy doc (the doctrinal source for this collapse)
• Reconcile metric-type.json enum — collapses metrics / verification / attribution into one list #3858 — performance-feedback metric row migration (the same pattern applied)

🤖 Generated with Claude Code