Core concepts

Key concepts

Concept	Description
Reorder point (s)	The inventory level that triggers a new order. Continuous review: the instant on-hand + on-order inventory drops to or below s, an order is placed.
Safety stock	The portion of s above expected demand during lead time — your buffer against demand variability.
Service level	The probability you don't stock out in a given period. Simulated directly (not a formula approximation).
Stockout-probability curve	The sweep of stockout probability across a range of candidate reorder points, letting you see the full risk/cost tradeoff.
Cost-minimizing mode	Activates only when you give both a holding cost and a stockout cost — surfaces the reorder point with the lowest total expected cost, alongside the service-level-based recommendation.

How the simulation works

Once demand, lead time, and a target service level (or costs) are set, the tool:

1. Sweeps a range of candidate reorder points around a practical range for your inputs
2. Simulates many replications of demand and lead time for each candidate, tracking on-hand + on-order inventory
3. Computes stockout probability (and cost, if given) at each candidate
4. Identifies the smallest reorder point meeting your target service level, and — if costs were given — the reorder point minimizing total expected cost
5. Runs a final high-replication confirmation pass at just those two points for a tighter, confidence-interval-backed estimate

This usually finishes in a few seconds. See Async Job Lifecycle for how the frontend polls for results.

How the AI assistant works

The AI has full read access to your current input configuration (demand source, lead time, service level, costs), your current results (recommendation, curve data), and the attached dataset's column names and dtypes. When you describe what's needed, it submits a run by emitting a validated configuration, which the backend checks before enqueuing the simulation — so a malformed request never silently produces bad results.