Winner Repurposing بدون Copy

1) Is the winner a transferable idea, or just a high-performing asset?

A 30-second customer story outperforms the rest. The team proposes six copies with the same script, resized canvases and new hashtags. That may extend distribution, but it does not show that the underlying mechanism transferred. It also obscures provenance, rights and the measurement contract of each derivative.

In this lesson we build Winner Repurposing Derivation Map v1. The goal is not to duplicate the winner. The goal is to state what appears to have won, bind that hypothesis to source evidence, then rebuild expression for each target channel.

Central rule: preserve the tested mechanism, not necessarily the literal wording, sequence or visual treatment of the source artifact.

2) COM-013 scope: repurposing, not channel playbooks or the performance loop

COM-013 owns derivation from a winner. COM-012 owns channel-format playbooks and COM-014 owns the performance feedback loop. Those outcomes remain separate.

A “winner” here is not causal truth. Distribution, timing or audience conditions may have contributed. winning_hypothesis therefore stays bounded and traceable to source_experiment_ref.

3) Mental model: Winner Evidence → Hypothesis → Provenance → Transformation → Fresh Test

Ask five questions in order: what was the source, in which experiment did it win, what mechanism is hypothesized to transfer, what will deliberately change in the derivative, and how will the derivative be measured on its own surface?

If source_experiment_ref or derivation_relation disappears, the variant is no longer traceable to the evidence that created the winner claim.

4) What PROV establishes — and what it does not

W3C PROV-O provides provenance concepts for Entity, Activity and Agent and includes prov:wasDerivedFrom for derivation. This FACT gives us a standard way to think about a derivative's trace back to a source.

PROV does not define a marketing winner, a similarity threshold or a repurposing policy. Those remain local product/governance choices.

5) Separate the winner hypothesis from the artifact

source_asset_id identifies the asset. source_experiment_ref binds the evidence context. winning_hypothesis states the bounded mechanism the team believes may transfer.

“Problem-to-outcome customer proof” can be a mechanism hypothesis. Exact script lines or the same scene order are not automatically the mechanism unless the experiment isolated them.

6) Channel-native transformation: make a real change, not a resize

Every variant_id needs explicit changed_elements: hook, structure, duration, framing, visual treatment, CTA or sequencing. If changed elements are effectively empty and the script remains the same, the state should become BLOCKED_LITERAL_COPY.

Healthy repurposing preserves intent/mechanism while rebuilding expression for the target surface.

7) Provenance chain: every derivative should trace backward

derivation_relation shows what the derivative came from, while evidence_ref makes the decision reproducible. If the source experiment reference is removed, the winner label is no longer auditable and BLOCKED_NO_EXPERIMENT_REF is safer than continuing production.

Provenance does not by itself prove quality or legality; it makes the chain visible.

8) Rights: prior publication is not automatic permission for a new derivative

Current YouTube Shorts guidance says copyright-protected material created elsewhere must be approved for use on YouTube; invalid rights can lead to Content ID claims or copyright takedown consequences.

In MAKNOW, rights_state is checked per derivative. If a retained music clip, image or testimonial ingredient has unclear rights, the variant becomes BLOCKED_RIGHTS. Approval of the source does not automatically clear the derivative.

9) Duplicate risk: do not hide near-verbatim reuse

duplicate_risk is an explicit gate. We do not invent a universal legal similarity percentage. We make near-verbatim reuse visible and compare it with the declared changed elements.

If the team only resizes the canvas while preserving the same script, sequence, hook and CTA, it is not a fresh transfer test. Call it copy, not new evidence for the hypothesis.

10) Give every variant a fresh measurement contract

Do not automatically reuse the source success metric on the target surface. A Short, email, landing page and carousel can have different exposure and action semantics. measurement_contract should define the target metric, window and success rule for that variant.

If the source metric is copied across channels without redefinition, REVISE_MEASUREMENT is safer than manufactured success.

11) Major failure modes and the correct typed state

• source experiment removed → BLOCKED_NO_EXPERIMENT_REF
• near-verbatim script with cosmetic changes → BLOCKED_LITERAL_COPY
• retained ingredient has unclear rights → BLOCKED_RIGHTS
• source evidence is stale/unrecoverable → STALE_SOURCE_EVIDENCE
• source metric reused without a target contract → REVISE_MEASUREMENT
• second winner uses a materially different mechanism → NEW_HYPOTHESIS_REQUIRED
• a decisive field is actually missing → HOLD_MISSING_EVIDENCE

Typed states stop ambiguity from being converted into “ready to publish.”

12) Lab and transfer: a second winner with a different hook

COM-013-LAB1 contains eight offline fixtures. The learner derives state from visible evidence; the previous label is never the oracle.

In the transfer case, a second asset wins on the same topic through a different hook. If the mechanism is materially different, do not merge it into the first winner. Create a new hypothesis or keep the relationship disputed, then define a fresh measurement contract.

13) Bounded decision, freshness and what this lesson does not prove

This lesson does not prove causation, grant legal clearance or publish live content. Its output only determines whether the current evidence supports bounded repurposing or must remain HOLD/BLOCK/REVISE.

PROV-O is a stable W3C Recommendation. YouTube's Shorts/copyright guidance is vendor-current and should be rechecked by 2026-10-18 or earlier after a material change. Winner extraction, duplicate-risk gates and approval states are MAKNOW DESIGN.