Business validation

1. Identify current stage and the decision that must be unlocked.
2. Separate direct evidence, proxy evidence and inference.
3. Record evidence with URL, capture date, claim and confidence.
4. Score the complete opportunity card; enforce the 35/50 and
   minimum-dimension gate.
5. When METRICS_INGESTION_ENABLED=false, do not define or run visitor, reader,
   audience-response or social-performance experiments. Record the proposed decision
   and evidence need for Phase 3 instead. After a future owner unlock, define one
   falsifiable experiment: segment, offer, channel, metric, baseline, target, review,
   cost/max loss and stop condition.
6. Prefer the cheapest test that can change a decision. Do not build the full
   product to test demand for the product.
7. Preserve the original hypothesis/forecast. At review choose exactly one:
   continue, change, stop or pivot.
8. Recognized revenue and recorded cost remain valid business facts. Qualified
   actions, traffic and other audience signals remain unavailable until Phase 3.
9. Paid scale requires an earlier signal and working measurement.
10. If evidence is insufficient, return NO_ACTION/research; never fill gaps
    with plausible prose.