Learning Loop Readout

Close the loop by turning observations into an explicit product-context and decision update.

Interpret the signal

1. Restate the decision and hypothesis that produced the evidence. If none existed, say which conclusions cannot be drawn retrospectively.
2. Inventory the observation source, collection method, population, timeframe, sample, missing data, and known bias.
3. Separate observations from interpretations. Reconcile qualitative, behavioral, commercial, technical, and operational signals without forcing agreement.
4. Determine whether each important assumption is supported, weakened, contradicted, or still unknown. Calibrate confidence to evidence quality.
5. Decide what changes:
   • update product context;
   • retain, revise, reject, or defer a decision;
   • change a conceptual model, quality bar, scope, positioning, or implementation;
   • leave the product unchanged because evidence is insufficient.
6. Record the reason and evidence provenance so another skill can resume without reconstructing the debate.
7. Name the next highest-risk unknown and the smallest evidence needed, or state that no further loop is currently justified.

Output

## Decision and prior hypothesis
## Evidence quality
## Observations
## Interpretations and alternatives
## Assumption updates
| Assumption | Prior status | New status | Evidence | Confidence |
## Decision update
- Retain / revise / reject / defer:
- Reason:
## Product context changes
## Next unknown and smallest evidence

Boundaries

• Correlation, engagement, conversion, and retention answer different questions; do not substitute one for another.
• Absence of evidence is not evidence of absence unless the test had power to observe the effect.
• Do not average away severe accessibility, safety, trust, or recovery failures.
• Preserve contradictory evidence and rejected alternatives when they matter to future decisions.
• Never rewrite the canonical context or decision record unless the user authorized file changes.

The loop is closed only when the evidence changes, retains, rejects, or deliberately defers a product decision.