Generation backlog¶

Open¶

ENH-009 - Deepen deterministic retail scenario realism¶

• Priority / effort: Idea / L
• Outcome: Stable customer, vendor, lead-time, promotion, weather, and holiday states create visible, reproducible operational effects.
• Acceptance: Scenario presets preserve determinism and move documented KPI targets in predictable directions.

Settled — do not reopen¶

• Generation is Spark-native and deterministic.
• Lakehouse Delta tables replace local DuckDB/parquet as the active contract.
• New data columns use snake_case; legacy TMDL-bound exceptions remain documented until migrated.
• Shared business invariants are enforced by utility/src/retail_setup/generation/invariants.py: no sales while closed, no pre-launch sales, no same-day returns, product lifecycle dates present, and validated (non-degenerate) profile controls.