Threshold elicitation (before segment tables)

If the user has not supplied thresholds, run this four-question pass before producing Gold / Rising / Negative tables — in the same pre-table batch as brand_markers (one conversation turn if possible). Defaults below are fallbacks only when the user explicitly says “use defaults” — otherwise pause for answers.

Questions (in order)

1. Gold ACoS ceiling — “What maximum ACoS % still counts as gold for your account (same attribution as the file)?”

   • Single number (e.g. 25 means “at or below target efficiency”).

2. Negative spend floor — “What minimum spend (in account currency) should a term need before we flag it as a negative candidate?”

3. Minimum clicks for negatives — “What minimum clicks on a term before we recommend keyword negatives (not just ‘watch’)?”

   • Suggest: ~10 if window ≈ daily, ~20 if ≈ weekly, ~30 if ≈ monthly — user picks.

4. Minimum analysis window — “How many days does this file cover, and what is your minimum window for strict negative recommendations?”

   • If actual window < user minimum (or skill default 14 days when user has no opinion), do not output strict negative recommendations — see segment-rules.md (sufficiency gate).

After answers

Record all four in the report header. If the user refuses, print PLACEHOLDER for each missing value and run in degraded mode with prominent caveats (prefer elicitation over silent placeholders).