Replenishment Policies under Uncertainty

Introduction and context

Inventory replenishment is straightforward only when demand and supplier lead time are known. In practice, customer orders fluctuate, transport times vary, forecasts contain error, deliveries may be incomplete, and operational disruptions can occur. A replenishment policy must therefore answer four questions:

1. When should an order be placed?
2. How much should be ordered?
3. What service level is economically appropriate?
4. How should the policy respond to uncertainty in demand, lead time, and supply?

This chapter develops the principal policies used in logistics systems: continuous-review systems, periodic-review systems, reorder-point policies, min–max systems, base-stock systems, and safety-stock methods. The emphasis is not simply on formulas, but on the assumptions behind them, their operational implications, and their suitability for different supply-chain environments.

Learning objectives

After studying this chapter, the reader should be able to:

• distinguish continuous-review from periodic-review replenishment;
• calculate reorder points under variable demand and lead time;
• estimate safety stock using service-level requirements;
• design min–max and base-stock policies;
• evaluate the effects of review frequency, order quantities, and lead-time uncertainty;
• select a policy appropriate to product value, demand pattern, and supply risk;
• diagnose limitations caused by forecast error, data quality, and correlated uncertainty.

1. Fundamental inventory concepts

1.1 Inventory position

Physical stock alone is an inadequate basis for replenishment decisions. The relevant control variable is inventory position:

where:

• IP = inventory position, in units;
• OH = on-hand inventory;
• OO = open or on-order inventory;
• BO = backorders or committed shortages.

Inventory position measures the stock that will be available after outstanding orders arrive and existing commitments are satisfied. A policy based only on on-hand stock can place duplicate orders when replenishment is already in transit.

The inventory position should be compared with a policy level, such as a reorder point or order-up-to level. Inventory on hand is still essential for warehouse operations, but inventory position is normally the correct quantity for replenishment control.

1.2 Demand during lead time

Let demand per period be the random variable D, lead time be L, and demand accumulated during lead time be D_L. If demand is constant at rate d and lead time is L, then:

For variable demand with a fixed lead time and independent periods, if mean demand per period is μ_D and standard deviation is σ_D, then:

The square-root relationship assumes independent demand observations with a stable variance. It should not be used mechanically when demand is strongly seasonal, autocorrelated, intermittent, or affected by promotions.

1.3 Cycle service level and fill rate

Two service measures are frequently confused:

• Cycle service level (CSL): probability that no stockout occurs during a replenishment cycle or lead-time protection period.
• Fill rate: proportion of demand supplied immediately from available inventory.

A high CSL does not necessarily imply a high fill rate. For example, one large stockout can reduce the fill rate substantially even if stockouts occur in only a few cycles. Managers should select the metric that reflects the customer promise and the economics of shortage.

2. Continuous-review replenishment

In a continuous-review, or Q,r, system, inventory position is monitored continuously. When it reaches the reorder point r, a fixed quantity Q is ordered.

The policy is appropriate when transaction data are available in real time and stockouts are costly. Typical applications include high-value spare parts, pharmaceutical items, industrial components, and automated distribution centres.

2.1 Reorder point with fixed lead time

With constant demand and fixed lead time:

where SS is safety stock. The first term is expected lead-time demand; the second protects against uncertainty.

If demand is variable but lead time is fixed, a normal-demand approximation gives:

where z is the standard-normal quantile associated with the desired cycle service level. For example, z ≈ 1.645 corresponds to a one-sided 95% service target under the normal approximation.

The units are consistent: μ_D and σ_D are units per period, while L is periods; therefore both terms in the reorder point are units.

2.2 Variable demand and variable lead time

When both demand and lead time vary, the variance of demand during lead time is more complex. Under the assumptions that demand per period and lead time are independent, demand observations are independent across periods, and lead time has mean μ_L and variance σ_L²,:

Thus:

The first variance component reflects demand uncertainty during the average lead time. The second reflects the effect of lead-time uncertainty on average demand. The expression is exact under the stated independence and conditional-variance assumptions when demand is accumulated over a random number of periods; it is not appropriate without modification when demand and lead time are correlated.

A broader approximation includes covariance:

The covariance term matters in environments where high demand causes congestion and longer lead times, or where seasonal demand and supplier capacity change together.

Worked example 1: continuous-review reorder point

A distributor sells an average of 80 units per day, with a daily standard deviation of 20 units. Supplier lead time averages 6 days and has a standard deviation of 2 days. Demand and lead time are assumed independent. The target CSL is 95%, so z = 1.645.

Expected lead-time demand is:

The variance of lead-time demand is:

Therefore:

Safety stock is:

The reorder point is:

The result should be rounded according to the item’s ordering and packaging constraints. The calculation suggests that an order should be triggered when inventory position falls to approximately 755 units. The large safety stock is mainly caused by lead-time variability: 80² × 2² = 25,600, compared with 2,400 from demand variance.

3. Periodic-review replenishment

In a periodic-review, or P,S, system, inventory is reviewed at intervals of P periods. At each review, an order raises inventory position to an order-up-to level S. No order is placed between reviews.

The protection period is not merely the supplier lead time. A stockout can occur after a review but before the next review, so the system must protect demand over:

With variable demand and fixed lead time:

With both demand and lead time variable, a common approximation is:

This expression assumes independence and stable demand parameters. The review interval must be expressed in the same time unit as lead time and demand statistics.

Periodic review usually requires less continuous data capture and can consolidate orders, making it attractive when suppliers have fixed delivery days, transport economies, or administrative ordering costs. Its disadvantage is additional uncertainty exposure and usually greater average safety stock.

Worked example 2: periodic-review order-up-to level

Daily demand has mean 50 units and standard deviation 12 units. Orders are reviewed every 7 days. Lead time is fixed at 4 days. For a 98% CSL, z = 2.054. The protection period is 11 days.

Thus the order-up-to level is approximately 632 units. If inventory position at review is 410 units, the order is:

The order may need to be adjusted to a case-pack multiple or supplier minimum.

4. Safety stock design

Safety stock is not a universal percentage of average demand. It is a deliberate investment against specified uncertainty and service consequences.

4.1 Normal approximation and service factor

If protection-period demand is approximately normal, safety stock is:

The value of z increases as the required CSL increases. Because the normal distribution has diminishing tail probabilities, moving from 95% to 98% or 99% service can require disproportionately more stock.

4.2 Demand uncertainty only

For a fixed protection period T and independent demand:

4.3 Forecast-error method

When a forecast system directly estimates the standard deviation of cumulative forecast error over the protection period, that estimate should be used instead of combining daily deviations again. Double-counting uncertainty is a common implementation error.

4.4 Non-normal and intermittent demand

The normal method can be unsuitable for slow-moving, intermittent, lumpy, or highly skewed items. Alternatives include:

• empirical quantiles from historical protection-period demand;
• Poisson or negative-binomial models for count demand;
• Croston-type forecasting for intermittent demand;
• simulation using observed demand and lead-time distributions;
• quantile forecasting that directly estimates a target service percentile.

For a target CSL, a direct empirical approach is to set the reorder point near the desired percentile of simulated or observed protection-period demand. This avoids imposing symmetry where none exists.

4.5 Economic service levels

The highest possible service level is rarely optimal. Additional safety stock has carrying, obsolescence, insurance, space, and financing costs. Shortage costs include lost contribution margin, expediting, production interruption, penalties, and reputational damage.

For a single-period decision with underage cost C_u and overage cost C_o, the critical fractile is:

where F(Q^*) is the target cumulative probability at the optimal stock level. This newsvendor relationship is not a direct replacement for a multi-period replenishment policy, but it provides an economic basis for choosing a service quantile.

5. Min–max and order-up-to policies

A min–max policy specifies a lower trigger level and an upper target level. When inventory position reaches or falls below the minimum, the system orders enough to restore the maximum:

The minimum often represents the reorder point, while the maximum covers expected demand over a replenishment or review cycle plus lead time and safety stock.

Min–max systems are common in warehouses and enterprise resource-planning systems because they are easy to explain and maintain. However, their quality depends on how the minimum and maximum are defined. A policy that sets minimum and maximum as arbitrary percentages of historical sales may ignore lead-time variability, order constraints, and changing demand.

A useful design is:

The exact maximum depends on whether the system reviews continuously, at scheduled intervals, or only when an order is generated. The terms must not be interpreted identically across software platforms; policy documentation is essential.

6. Base-stock systems

A base-stock system attempts to maintain inventory position at a target level after every demand event. If the base-stock level is S, each unit of demand generates an equivalent replenishment order, subject to batching and supplier constraints.

In the idealized form, the system resembles a continuous-review order-up-to policy with a one-for-one response. It is particularly suitable for high-value items, spare parts, service parts, and environments where demand information is immediately available.

The base-stock level should cover expected demand over lead time plus safety stock:

For several stocking locations, base-stock design must consider whether inventory is held centrally or locally. Risk pooling at a central location can reduce aggregate safety stock, but transportation time and response requirements may make decentralization necessary.

7. Comparing replenishment policies

Policy selection should consider demand regularity, item value, service criticality, supplier constraints, information systems, and the cost of monitoring. ABC classification alone is insufficient: a low-value item with severe production consequences may require tighter control than a high-value but noncritical item.

8. Order quantity and replenishment parameters

The reorder point determines when to order; order quantity determines how much. A common economic order quantity for constant demand is:

where D is annual demand in units, S is ordering cost per order, and H is annual holding cost per unit. This model assumes stable demand, instantaneous replenishment, no shortages, constant ordering and holding costs, and no quantity discounts. In uncertain environments, Q may instead be constrained by truckload economics, supplier minimums, shelf life, production batch size, or cash-flow limits.

Order quantity and safety stock serve different purposes. Increasing Q generally increases cycle stock; increasing safety stock protects against uncertainty. Confusing these roles leads to excessive inventory without a corresponding service improvement.

9. Implementation architecture and data requirements

A sound replenishment system requires more than a formula. It requires:

1. Accurate item master data: units of measure, pack sizes, shelf life, supplier, and lead-time calendars.
2. Reliable demand history: sales separated from returns, substitutions, stockout-censored demand, and exceptional events.
3. Lead-time measurement: actual order-to-receipt time, not merely a contractual promise.
4. Inventory-position logic: on-hand, allocated, in-transit, and backordered quantities must be reconciled.
5. Parameter governance: ownership, review frequency, approval thresholds, and exception handling.
6. Performance feedback: service, inventory, forecast error, supplier reliability, and obsolescence should be monitored together.

Modern warehouse-management and enterprise-planning systems can calculate dynamic reorder points and issue alerts. Machine-learning models may improve demand forecasts or estimate lead-time distributions, but they do not eliminate the need for business rules. A black-box forecast can produce inappropriate stock if it ignores minimum order quantities, expiry dates, supplier calendars, or strategic service commitments.

10. Illustrative case: hospital consumables distributor

Background

This illustrative case concerns a regional distributor of sterile surgical kits. Demand is variable, stockouts can delay hospital procedures, and products have expiry dates. The distributor historically used a monthly min–max review based on average sales.

Problem

The policy created two apparently contradictory outcomes: excess inventory of slow-moving kits and emergency purchases of fast-moving kits. Investigation found that the system used sales rather than unconstrained demand, ignored open purchase orders in some locations, and used promised rather than observed lead times.

Analysis

The distributor segmented items by clinical criticality, demand frequency, expiry risk, and supplier reliability. For a critical regularly demanded kit, it estimated daily demand variation and the empirical lead-time distribution. A continuous-review policy was selected because transactions were available in real time and shortage costs were high. For low-value, stable consumables delivered on fixed weekly routes, periodic review was retained.

The team also calculated protection-period demand rather than applying a blanket safety-stock percentage. For intermittent kits, it used simulation based on historical demand occurrence and actual lead times rather than a normal approximation.

Approach and solution

The revised design included:

• a reorder point based on the 98th percentile of simulated lead-time demand for critical kits;
• a weekly order-up-to policy for route-delivered consumables;
• expiry-aware allocation using earliest-expiry-first rules;
• explicit treatment of open orders, allocations, and backorders in inventory position;
• supplier scorecards based on actual lead-time mean and variability;
• monthly parameter review for critical items and quarterly review for stable items.

Outcome

In this illustrative scenario, emergency orders declined while total inventory was reduced through better segmentation and central pooling of slow-moving stock. The improvement did not result from raising every service target. It resulted from matching policy, data, and uncertainty treatment to the operational characteristics of each item.

Learning points

The case demonstrates that replenishment design is a portfolio problem. One policy for all items is administratively convenient but economically weak. It also shows why stockout observations must be interpreted carefully: recorded sales may underestimate true demand when products were unavailable.

11. Challenges, limitations, and governance

Parameter instability

Mean demand and lead time are not permanent constants. Product launches, promotions, market shocks, labour disputes, and geopolitical events can make historical parameters obsolete. Rolling estimates, change-point detection, and planner review are useful safeguards.

Correlation and common-cause risk

Demand and lead time may rise together during peak seasons. Multiple suppliers may be exposed to the same disruption. Treating correlated risks as independent understates safety stock. Scenario analysis and simulation are preferable when common-cause risk is material.

Stockout-censored data

Observed demand during a stockout is not the same as unconstrained demand. Forecast and safety-stock parameters should use lost-sales estimates, backorders, market intelligence, or comparable items where appropriate.

Ethical and governance concerns

Inventory decisions affect access to medicines, food, emergency equipment, and essential services. Service targets should therefore reflect consequences for vulnerable users, not solely accounting cost. Automated systems should provide audit trails, explain parameter changes, preserve human override for exceptional risks, and avoid discriminatory allocation rules. In regulated industries, traceability, expiry control, validation, and documented approval of model changes are essential.

Best-practice checklist

• Define the service metric before calculating safety stock.
• Measure actual lead times and their distribution.
• Use inventory position, not on-hand stock alone.
• Match the statistical model to demand characteristics.
• Include the review interval in periodic-review protection time.
• Separate cycle stock, safety stock, and pipeline stock.
• Segment by criticality and supply risk, not only annual spend.
• Test parameters through simulation or back-testing.
• Monitor service and inventory jointly.
• Establish governance for overrides, exceptions, and model changes.

Chapter Summary

Replenishment under uncertainty is the controlled positioning of inventory against uncertain demand, lead time, and supply conditions. Continuous-review Q,r systems trigger a fixed order when inventory position reaches a reorder point; periodic-review P,S systems replenish at scheduled intervals to an order-up-to level. Periodic systems must protect demand during both the review interval and lead time.

Reorder points consist of expected protection-period demand plus safety stock. When demand and lead time vary independently, the variance of lead-time demand combines demand-variance and lead-time-variance components. Safety stock should be linked to a defined service measure and should not be treated as an arbitrary percentage of sales. Normal approximations are useful for stable, sufficiently aggregated demand, but empirical quantiles, probability models, and simulation are often better for intermittent or skewed demand.

Min–max systems offer practical simplicity, while base-stock systems closely match replenishment to demand events. No policy is universally best: selection depends on information availability, demand pattern, supplier performance, item criticality, order constraints, and shortage consequences. Effective implementation requires accurate data, explicit governance, segmentation, and continuous performance feedback. Technology and AI can improve estimation and exception detection, but managerial oversight remains necessary where uncertainty, regulation, or human consequences are significant.

Key Terms

Inventory position: On-hand plus on-order inventory minus backorders.

Continuous review: Monitoring inventory position continuously or after each transaction.

Periodic review: Reviewing inventory at predetermined time intervals.

Reorder point: Inventory-position level that triggers replenishment.

Order-up-to level: Target inventory position restored by a replenishment order.

Protection period: Time over which demand must be covered; usually review interval plus lead time.

Safety stock: Inventory held to protect against uncertainty.

Cycle stock: Inventory resulting from ordering in batches.

Cycle service level: Probability of avoiding a stockout during a specified cycle.

Fill rate: Fraction of demand supplied immediately from available stock.

Lead-time demand: Demand accumulated while replenishment is arriving.

Min–max policy: Policy that orders when inventory reaches a minimum and restores it to a maximum.

Base-stock policy: Policy that seeks to maintain inventory position at a target level.

Demand variability: Statistical variation in demand over time.

Lead-time variability: Statistical variation in replenishment duration.

Service factor: Standard-normal quantile associated with a target service level.

Stockout-censored demand: Demand hidden because inventory was unavailable.

Risk pooling: Reduction of aggregate uncertainty by combining demand across locations or products.

Review Questions

Short Answer Questions

1. Define inventory position and explain why it is preferable to on-hand inventory alone for replenishment decisions.
2. Distinguish cycle service level from fill rate.
3. Why does a periodic-review system protect demand over the review interval plus lead time?
4. State two assumptions underlying the normal approximation for lead-time demand.
5. Explain the difference between cycle stock and safety stock.

Long Answer Questions

1. Compare continuous-review, periodic-review, min–max, and base-stock policies in terms of data requirements, responsiveness, and inventory implications.
2. Derive and interpret the reorder-point expression when both demand and lead time are variable and independent.
3. Discuss why normal-based safety stock can be inappropriate for intermittent or highly skewed demand, and explain suitable alternatives.
4. Explain how service-level choice should incorporate shortage costs, holding costs, product criticality, and regulatory obligations.
5. Design a governance framework for maintaining replenishment parameters in a multi-site distribution network.

Analytical / Application Questions

1. Daily demand has mean 100 units and standard deviation 25 units. Lead time has mean 5 days and standard deviation 1.5 days. Assuming independence and a 95% cycle service level, calculate expected lead-time demand, the standard deviation of lead-time demand, safety stock, and the reorder point.
2. A warehouse reviews stock every 10 days and has a fixed 6-day lead time. Daily demand has mean 40 units and standard deviation 8 units. For a 97.5% service target, calculate the order-up-to level using the normal approximation, then explain how the result should change if demand is strongly seasonal.
3. A firm has frequent stockouts for a low-value but production-critical component and excess stock for a high-value slow-moving item. Recommend different replenishment policies and identify the data and service measures needed to justify your recommendations.

Multiple Choice Questions with Answers

Question 1

Which quantity should normally be compared with a reorder point?

A. On-hand inventory only

B. Inventory position

C. Annual demand

D. Average order cost

Correct Answer: B. Inventory position

Question 2

In a continuous-review Q,r system, what does Q represent?

A. The reorder point

B. The lead time

C. The fixed order quantity

D. The safety factor

Correct Answer: C. The fixed order quantity

Question 3

A periodic-review system with review interval P and lead time L must generally protect demand over which period?

A. L − P

B. P only

C. L only

D. P + L

Correct Answer: D. P + L

Question 4

If mean daily demand is 60 units and fixed lead time is 4 days, what is expected lead-time demand?

A. 15 units

B. 64 units

C. 240 units

D. 1,440 units

Correct Answer: C. 240 units

Question 5

Under the independent-variable assumptions, which term represents the effect of lead-time variability on lead-time-demand variance?

A. μ_Lσ_D²

B. μ_D²σ_L²

C. μ_Dμ_L

D. zσ_D

Correct Answer: B. μ_D²σ_L²

Question 6

What does safety stock primarily protect against?

A. Uncertainty in demand and supply timing

B. The accounting cost of ordering

C. Packaging standardization only

D. The fixed portion of annual demand

Correct Answer: A. Uncertainty in demand and supply timing

Question 7

A min–max policy normally places an order when inventory position reaches the minimum in order to restore inventory position to the:

A. Reorder cost

B. Average demand

C. Maximum level

D. Supplier lead time

Correct Answer: C. Maximum level

Question 8

Which demand pattern most clearly challenges a normal safety-stock approximation?

A. Stable, high-volume demand

B. Intermittent and highly lumpy demand

C. Independent demand with many observations

D. Constant demand with fixed lead time

Correct Answer: B. Intermittent and highly lumpy demand

Question 9

A base-stock system seeks to maintain which quantity near a target level?

A. Inventory position

B. Supplier capacity

C. Annual holding cost

D. Demand variance

Correct Answer: A. Inventory position

Question 10

If daily demand has standard deviation 10 units and the fixed protection period is 9 days, what is the protection-period standard deviation under independent demand?

A. 10 units

B. 30 units

C. 90 units

D. 100 units

Correct Answer: B. 30 units