Provenance distribution experiment

Purpose

This is a four-week learning cycle for bringing relevant provenance and museum-data practitioners into one measurable Meta Museum journey. It operationalizes the useful parts of the distribution handbook without adding session replay, a second analytics product, automated outreach, or broad social posting.

Primary journey

The primary funnel is:

1. `organic_landing_session`

1. `provenance_assessment_start`

1. `provenance_assessment_complete`

1. `provenance_assessment_recommendation_kit`

1. `research_kit_checkout_start`

Checklist downloads, review inquiries, and the other assessment recommendations remain supporting signals. They are valid user choices, not failed conversions. A checkout start is commercial intent only; it is not a sale, revenue, or product validation.

Attribution and privacy

Use only the generated links. The browser analytics boundary retains sanitized `utm_source`, `utm_medium`, and `utm_campaign` values for consented organic-funnel events. Arbitrary URL parameters and direct identifiers are not copied into analytics. Visitors who decline analytics remain outside the measured denominator.

Generate or refresh the packet with an explicit UTC start date:

pnpm organic:distribution:plan -- --starts-on=2026-08-24

The command writes content-addressed JSON and human-readable Markdown under `artifacts/organic-income/distribution/`.

Operating cadence

• Before week 1, verify the landing routes and exact links. Record the experiment window before interpreting results.
• During each week, an accountable human may separately approve individual posts or partner/newsletter submissions. Preserve the exact approved copy and destination outside this packet.
• At the end of each week, export aggregate consented GA4 rows by source, medium, campaign, landing route, and funnel event. Do not infer zero from an unavailable export.
• After week 4, compare qualified-action rate and assessment completion by source. Continue only relevant sources with genuine qualified actions.

Decision rules

• Do not expand distribution before 30 consented campaign landing sessions; the sample is otherwise descriptive only.
• If assessment completion is below 60% after 10 starts, inspect the four-question flow before increasing traffic.
• One checkout start or review inquiry demonstrates intent only. Provider-confirmed payment or a returned, qualified inquiry is separately required.
• Never describe clicks, page views, downloads, or form opens as demand, validation, revenue, professional review, or research impact.

Authority boundary

The generated packet prepares sequencing, landing links, and measurement. It does not authorize publication, email, direct contact, paid placement, partner representation, production configuration, or spending. Those actions require separate attributable human approval.

Technical recommendation packet

Impact and ownership

The public analytics service owns event eligibility and sanitized attribution. The client analytics component owns consent checks, session-scoped attribution continuity, and route/click emission. The distribution service and command own deterministic planning artifacts. No provider, database, route-handler, Linked Art model, or external-service boundary changes.

Highest coupling and failure risks

1. GA4 consent or configuration can leave a correct browser implementation with no production observations; unavailable data must remain unknown, not zero.

1. Campaign parameters disappear during ordinary internal navigation unless sanitized attribution remains session-scoped; the helper and regression test now protect that handoff.

1. Landing slugs can drift from the organic guide registry; the experiment test verifies every generated path against the canonical guide list.

Next actions

• Growth operator: export aggregate campaign/event dimensions after the closed window. Acceptance: source, medium, campaign, and event totals reconcile without identifiers. Validation: normalize the GA4 evidence and compare it with the generated campaign ID.
• Product owner: review assessment completion only after the declared minimum denominator. Acceptance: at least 10 starts before interpreting the 60% threshold. Validation: reproduce event totals from the retained export.
• Engineering: add an end-to-end consented campaign journey when the next browser-proof cycle runs. Acceptance: a generated landing URL emits one landing event and later assessment events keep the same sanitized attribution. Validation: `pnpm organic:funnel:browser-proof` plus the focused analytics tests.

Next-cycle hypothesis

The Linked Art and museum/provenance partner links will produce a higher qualified-action rate than owned or broad research-community links. Falsify it from a closed-window aggregate export if the niche-source rate does not exceed the comparison-source rate after the minimum denominator is met.