Inventory policy: how much service is another day of coverage worth?¶

Reducing inventory releases cash. Increasing it protects sales and operating continuity. Neither statement, by itself, tells a planner how much of each item to hold at each location.

An inventory policy starts with a better question: what inventory is needed to sustain an explicit service promise, given the replenishment process, uncertainty, and economics of that item-location?

Inventory is not one homogeneous block¶

Four components are commonly mixed in aggregated reports:

• Cycle stock comes from purchase, production, or transfer lot size and replenishment frequency.
• Pipeline stock covers expected demand while supply is in transit or production.
• Safety stock protects against deviations in demand, lead time, yield, or supply availability.
• Anticipation stock is deliberately built before seasons, campaigns, shutdowns, or capacity constraints.

Each component answers a different decision. Reducing lot sizes changes cycle stock. Shortening lead time reduces pipeline and often the buffer. Better forecasts reduce part of the uncertainty. Network optimization can reposition risk between echelons.

That is why a generic “reduce inventory by 15%” target can remove capital from the wrong place while preserving excess that creates no service.

The policy must exist at item-location level¶

Financial targets still matter, but they work as constraints and outcomes—not as substitutes for operating parameters. Four definitions should precede any formula:

1. What protection period is required? Under continuous review, it is usually lead time; under periodic review, it also includes the time until the next review.
2. Which service metric matters? Cycle service level is the probability of completing a replenishment cycle without a stockout. Fill rate is the share of units immediately served from stock. They are not interchangeable.
3. How does replenishment work? Minimum lots, calendars, capacity, and multiples change both inventory and exposure to risk.
4. What happens when stock runs out? A unit may become a lost sale, backlog, substitution, emergency shipment, or line stop. Each consequence has a different economic cost.

Why protection has diminishing returns¶

The interactive illustration below uses a normal-demand example with mean demand of 10 units/day, daily standard deviation of 5 units, mean lead time of 5 days, lead-time standard deviation of 2 days, and a lot size of 100 units.

Under independence assumptions, uncertainty during lead time is:

σDL = √(L·σD² + μD²·σL²)

The chart computes cycle service level with Φ(z) and fill rate through the normal loss function. Move the slider and compare the two curves.

The pattern is the point: every additional unit buys less protection than the previous one. Getting close to 100% requires increasingly large buffers, and a normal distribution still does not provide an absolute guarantee.

The best point depends on the economics of shortage¶

The service curve shows risk; it does not choose a policy. The decision must compare two marginal effects:

• the cost of protection: capital, storage, and the expected loss of inventory that expires or becomes obsolete;
• the cost of shortage: lost contribution, emergency freight, substitution, penalty, backlog, or operational interruption.

For a planner, it is useful to keep four economic components separate:

Separating these terms prevents double counting. Lost sales should normally be valued through lost contribution, not full revenue, and an obsolescence or write-off term should not also be hidden inside a generic holding-rate assumption.

In a simple continuous-review model, the relevant annual cost can be expressed as:

Cost(SS) = SS·h + ES(z)·(Annual demand/Q)·p

Here, h combines capital and storage for one unit-year and p is the expected economic consequence of one unit short. If expected write-off changes with the policy, it must be added as a separate term. Cycle and pipeline stock may be included in the financial view, but if they do not vary across the safety-stock scenarios, they do not change the marginal buffer choice.

This explains why one service target for every item often destroys value. A critical, hard-to-substitute item with a high shortage cost may justify more protection. A slow-moving item with high obsolescence and fast replenishment may justify less.

From reactive management to network optimization¶

A practical maturity path has four steps:

1. Standardize replenishment. DRP or Deployment replaces ad hoc decisions with explicit, auditable parameters.
2. Recalculate from zero. Policies stop being incremental edits of current balances; item-locations are segmented, and single-stage models size buffers from observed data.
3. Optimize multiple echelons. Multi-echelon models evaluate where inventory best protects the network and avoid adding independent buffers at every level.
4. Redesign the operating model. Once policy parameters are governed, teams can revisit the push-pull boundary, postponement, channels, frequencies, lots, capacity, and supplier collaboration.

Multi-echelon is not automatically better for every context. Model sophistication should follow the actual network complexity and the maturity of the process that will execute the recommendations.

Data needed to move from curve to policy¶

The chart is didactic. A governed policy needs connected operational data:

A useful implementation checklist¶

1. Do we have an unambiguous service metric for each segment?
2. Can we separate cycle, pipeline, safety, and anticipation inventory?
3. Are parameters recalculated with observed forecast error and lead time, or only changed after a crisis?
4. Are shortage and holding costs explicit and free of double counting?
5. Can DRP, purchasing, production, and distribution execute the recommendation?

The objective is not simply to hold less inventory. It is to use inventory as an economic resource: enough to protect the service promise, positioned at the right echelon, and revised when uncertainty or the operating model changes.

Continue with the durable product concepts and review workflow in Inventory policies and optimization, then see how replenishment executes those parameters in DRP and Deployment. To organize that evolution by process capability, read From inventory target to executable policy. The economic decision data model shows how policy simulation and Cost-to-Serve share physical facts without becoming the same calculation.

Further reading¶

• Farasyn et al. (2011), “Inventory Optimization at Procter & Gamble: Achieving Real Benefits Through User Adoption of Inventory Tools”, Interfaces, 41(1), 66–78.
• Prak, Teunter & Syntetos (2017), “On the calculation of safety stocks when demand is forecasted”, European Journal of Operational Research, 256(2), 454–461.