Inventory policies that trade service for cash — deliberately

Safety stocks and reorder points linger from a planner who left two years ago. ABC classifications are stale. Multi-echelon networks push inventory into the wrong nodes. Finance demands lower working capital while sales demands higher fill rates — and neither sees a coherent policy trade-off.

ERP min/max fields are manually edited under fire. Expedites hide chronic under-buffering on A items. Slow movers accumulate because nobody owns a kill decision. Warehouse space and cash are consumed by SKUs that should have been discontinued after the last season.

Demand noise, supplier unreliability, and MOQ constraints interact. A single global service target wastes cash on C items and still fails on critical SKUs. Without segmentation, “optimisation” becomes a black-box recommendation nobody trusts enough to implement.

Inventory planning owns policies; demand planning owns forecasts; procurement owns supplier lead-time reality; finance owns cash targets; operations owns warehouse constraints. Anti-patterns include changing every SKU overnight, ignoring MOQs, and celebrating model recommendations that violate shelf-life or storage limits.

Scenario thinking belongs in the same workflow. Planners need to see what happens to cash and service if targets tighten, if a supplier’s lead time worsens, or if a promotion doubles demand for a narrow window. Without scenarios, “optimisation” is a single opaque recommendation that finance cannot challenge. Network effects matter too: fixing buffers at a DC while ignoring store or plant constraints simply moves the shortage.

AI inventory optimisation should propose policy parameters by segment, expose the service–cash trade-off, and feed replenishment — while planners remain accountable for adopting or rejecting recommendations with an audit trail. Adoption is the real KPI: a model that recommends perfect policies nobody loads into ERP creates no inventory change, so planner change management is part of delivery.