DECISION BRIEF
Give FACET one question. Not your entire repository.
A bounded pilot shows whether the extra structure is useful for your team. Aim for one evidenced answer about an important feature, with visible effort and permission to stop afterwards.
question: Can the same discount apply twice?rule: A code applies once per orderevidence: - test: discount.spec.ts › applies a valid code onceopen: - Does this still hold after a cancellation?budget: 1h
Illustrative example · a draft record after the first pilot hour
THE PILOT IN FOUR STEPS
One hour. One evidenced answer or a clear no.
This is a time limit for your experiment, not a promised setup time or time to results. Pick one feature and one concrete question. Stop at the limit even if the answer is incomplete.
Capture the question and baseline: for example, “Can the same discount apply twice?” Note how you currently find the answer and confirm the expected rule.
Prerequisites
Access to source and existing tests, FACET already configured, a coding agent for drafts and someone who can confirm the business rule. Use a separate branch. Setup and approvals may require additional time.
WHAT YOU BRING BACK
Four things. Then you decide.
- One confirmed rule with source references.
- Actual test evidence or an explicitly named gap.
- The effort to capture, review and make an example update.
- A recommendation: continue selectively, revisit later or skip it.
Continue when …
another person can follow the evidenced answer, the knowledge will be reused on the next change and an owner accepts the maintenance cost. Agree on your own effort threshold before expanding.
Stop when …
the answer is just as easy to get from existing code and tests, records only duplicate familiar knowledge or nobody can maintain their correctness. The hour you invested is not a commitment to adoption.