Deterministic calculations first. AI explanation second.
Nerqel is designed so operational numbers do not depend on a language model inventing calculations.
Demand and replenishment
Historical demand is normalized into an expected daily demand rate. When dated history is available, recent and earlier demand are compared to introduce a bounded trend adjustment. Lead time, service level and demand variability feed safety-stock and reorder-point calculations.
Reorder point = expected demand during lead time + safety stock.
Safety stock
Safety stock uses a service-level factor, observed or estimated demand variability and the square root of lead time. When history is limited, the system uses conservative fallback variability rather than claiming false precision.
Risk and confidence
Risk scores prioritize shortages relative to the reorder point and compare available days of cover with supplier lead time. Confidence reflects the amount and quality of historical evidence. A confidence score is not a guarantee.
Role of AI
The AI Operator is grounded in persisted runs and actions. It is instructed not to invent numbers. If an AI gateway is unavailable, Nerqel falls back to deterministic summaries based on stored business facts.
Important limitations
Outputs depend on input quality and assumptions. Promotions, substitutions, seasonality, minimum order quantities, capacity constraints and supplier-specific commercial terms may require additional context. Inventory recommendations should be reviewed by the responsible operator before execution.