Turn the symptom into a testable decision.
High inventory and poor availability can coexist when stock is in the wrong item, place or state. Aggregate inventory turns hide the decision that failed. The sequence below is designed to preserve definitions, expose alternative explanations and lead to a decision that can be validated.
- Define the service promiseChoose the customer-facing unit: fill rate, in-stock rate, on-time-in-full or lost demand. State whether backorders and substitutions count.
- Rebuild the inventory ledgerReconcile opening balance, receipts, reservations, sales, transfers, adjustments and closing balance by SKU and location.
- Find the first failed decisionTrace each stockout backward to forecast, safety stock, order quantity, allocation or execution.
- Segment the economicsSeparate A-items, long lead-time items, volatile launches, substitutes and low-margin tail inventory.
- Replay policy alternativesTest reorder points, allocation priorities and transfer rules with the same lead times, capacity and demand history.
Ask for the minimum data that can change the answer.
Begin with read-only access and a field-level purpose. Reconcile samples before scaling extraction, preserve event time and source provenance, and record missingness rather than silently filling it.
Validate the claim before changing the operation.
Replay one replenishment or allocation rule
Freeze a historical period and reproduce the observed baseline first. Then replay a candidate policy using only information that would have been available on each day. Report service level, working capital, expedites, aged stock and violations by SKU segment. A policy that improves overall availability but starves priority stores or consumes unacceptable cash has not passed.
What makes the diagnosis look right and still fail.
- Using monthly averagesA three-day stockout disappears inside a healthy monthly average even if it loses the launch.
- Calling all unmet demand forecast errorLate purchase orders, reserved stock and allocation rules can create the same symptom.
- Leaking future informationA replay that uses the final month’s demand or actual lead time overstates performance.
- Optimizing availability aloneMore safety stock can hide bad placement while increasing cash and obsolescence.
- Ignoring inventory accuracyA planning model cannot repair stock that exists in the system but not on the shelf.
Primary and official references
These sources define the measurement, control or operating context. They do not replace validation on the company’s own data.
Questions enterprise teams ask first.
What is the right grain for analysis?
SKU-location-day is a practical default; faster operations may require event time, while slow industrial items may use weeks.
How should lost sales be estimated?
Combine censored sales with search, views, substitutions, backorders and comparable in-stock periods. State the uncertainty range.
What makes a replay credible?
It reproduces the actual baseline, respects information timing and operational constraints, and reports errors by segment.
Should the first PoC use optimization or simple rules?
Use the simplest policy that can falsify the hypothesis. Optimization is justified when interacting network constraints materially affect the decision.