[C6 HL-038] Instrument product learning and controlled experiments #48

Outcome

Learn from real usage without turning the artwork into a conversion dashboard.

Placement and activation

• Circle: C6 — Sustainable commercial and institutional operation
• System responsibilities: Commercial / Product / Assurance
• Capability coverage: 04 Data Platform; 06 Infrastructure & Runtime; 12 Growth, Analytics & Experimentation
• Planning status: Conditional; activate only when the requirement below is met.
• Activation: An actual product decision needs usage evidence; hosted instrumentation also passes privacy review.
• Parent epic: [EPIC C6] Sustainable commercial and institutional operation #9; master roadmap: [MASTER EPIC] Hole-loop: concentric evolution within Persona / Visual Form Canon #2.

Dependencies

Required prerequisites:

• A named product decision, explicit question/hypothesis, proposed evidence method, owner, and experiment/edition identities.
• Source/edition identity from HL-003 #12 and the applicable local-study evidence from HL-010 #20 or released-edition documentation from HL-017 #27.

Conditional dependencies:

• HL-037 #47 applies when studying a commercial offer, revenue, entitlement, or payment workflow; a free artwork or local author comparison does not require billing.
• HL-032 #42 must be satisfied before collecting hosted personal/session data.
• HL-041 #51 supplies applicable consent, privacy-notice, vendor, or institutional responsibility decisions before the corresponding collection begins.

Qualitative reader comparisons, opt-in local studies, and product decisions can proceed without monetization or hosted instrumentation. Revenue and attribution metrics are selected only for an actual commercial question.

Acceptance evidence

• Define useful product questions separately from artistic interpretation; specify minimal events, activation, retention, funnels/cohorts, attribution and revenue analytics only where meaningful.
• Never infer reading, comprehension, mood, vulnerability or intention from visits/dwell; document consent, minimization, sampling and retention.
• Define experiments/A-B tests, assignment, feature-flag ownership, hypotheses, guardrails, stop conditions and review; prefer author/preview studies when public art should remain a stable edition.
• Keep experiment ID distinct from edition identity, prevent incompatible patches entering one release silently, and record build → release → observe → learn → change decisions.
• Test opt-out/reset and event correctness; quantitative evidence complements documented qualitative review rather than rating artistic success.
• If hosted personal/session data is collected, satisfy HL-032 #42 before instrumentation; local qualitative composition comparisons do not require a SaaS analytics stack.

Completion and scope

Attach the actual artifact or commit, relevant verification, limitations, and an artist decision where a composition changes. This issue plans authorized work; creation does not claim implementation or deployment. Preserve independent work/source identities and private source boundaries. Repository creation requires the extraction evidence in HL-004 #13 and HL-045 #14.