5. Inventory Under Uncertainty

Introduction and Context

Inventory decisions are rarely made under conditions of certainty. Demand fluctuates, suppliers deliver early or late, transport times vary, orders may be rejected or partially fulfilled, and market conditions can change while replenishment is in progress. Consequently, an inventory policy based only on average demand and average lead time may appear efficient on paper but perform poorly in practice.

Inventory under uncertainty concerns the design of policies that balance two competing consequences. Too little inventory creates stockouts, lost sales, production interruptions, emergency purchases, and dissatisfied customers. Too much inventory increases capital tied up in stock, storage costs, obsolescence, insurance, handling, and markdown exposure. The central managerial problem is therefore not to eliminate uncertainty, which is usually impossible, but to decide how much uncertainty the supply chain should absorb and at what cost.

This chapter develops the principal analytical models used to make that decision. It explains safety stock, reorder points, target service levels, continuous-review and periodic-review systems, base-stock policies, the newsvendor model, and distinctions between lost sales and backorders. The emphasis is on interpretation: a formula is useful only when its assumptions, data requirements, and operational consequences are understood.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Explain the sources and consequences of demand and lead-time uncertainty.

2. Distinguish cycle stock, safety stock, reorder point, service level, and fill rate.

3. Calculate safety stock and reorder points under common uncertainty assumptions.

4. Select between continuous-review, periodic-review, and base-stock policies.

5. Apply the newsvendor model to single-period inventory decisions.

6. Explain the operational difference between lost sales and backorders.

7. Interpret the managerial and financial implications of target service levels.

8. Evaluate the limitations of inventory models and improve decisions using analytics and technology.

The Nature of Uncertainty in Inventory Systems

Uncertainty may arise from demand, replenishment, or both. Demand uncertainty includes variation in quantity, timing, product mix, customer location, and order composition. Lead-time uncertainty includes variation in supplier processing, production, transportation, customs clearance, receiving, and put-away. In many supply chains, these sources interact: a promotion increases demand precisely when suppliers are congested, or a transport disruption extends lead time while demand remains high.

A useful distinction is between independent variability and systematic uncertainty. Independent variability is random fluctuation around a relatively stable pattern. Systematic uncertainty results from identifiable factors such as seasonality, promotions, economic cycles, product launches, or policy changes. Statistical safety stock is more appropriate for the first type. The second requires forecasting, scenario planning, event management, and deliberate adjustment of inventory parameters.

Inventory policy also depends on the consequences of shortage. A stockout of a low-value office supply may be inconvenient, whereas a stockout of a critical medical component can stop operations or endanger service delivery. The appropriate service level is therefore an economic and strategic decision, not simply a statistical target.

Important Terminology

Cycle stock is the inventory expected to be consumed between replenishment deliveries. For a stable system with order quantity Q, average cycle stock is approximately Q/2.

Safety stock is inventory held to protect against uncertainty in demand, lead time, or both.

Reorder point is the inventory position at which a replenishment order is released.

Inventory position equals on-hand inventory plus on-order inventory minus backorders or other committed quantities.

Lead-time demand is demand expected during the replenishment lead time.

Service level is a general expression of the probability or extent to which demand is satisfied. It must be defined precisely before calculation.

Cycle service level (CSL) is the probability that no stockout occurs during a replenishment cycle.

Fill rate is the proportion of total demand immediately supplied from available inventory.

Backorder is customer demand recorded for later fulfilment when stock is unavailable.

Lost sale occurs when unsatisfied demand is not retained and the customer purchases elsewhere or abandons the purchase.

From Average Demand to Protection Decisions

The basic reorder-point logic is:

Reorder point = Expected demand during lead time + Safety stock

If average demand per period is d and lead time is L periods, expected lead-time demand is dL. Under stable demand and lead time, the reorder point may be set close to dL. Under uncertainty, the safety-stock term protects against deviations from that expectation.

A common mistake is to add safety stock to average demand without identifying the source of variability. The correct calculation depends on whether demand varies, lead time varies, or both vary. It also depends on the selected service measure and the probability distribution assumed for the uncertainty.

Demand Variability with Constant Lead Time

Suppose demand per period has standard deviation sigma_d and lead time is constant at L periods. If demand observations are independent, the standard deviation of demand during lead time is:

sigma_LT = sigma_d multiplied by the square root of L.

For a target cycle service level associated with standard-normal value z:

Safety stock = z sigma_LT

Thus:

ROP = dL + z sigma_d square root of L.

For example, daily demand is 100 units, daily standard deviation is 20 units, and lead time is four days. At a 95 percent cycle service level, z is approximately 1.645. Expected lead-time demand is 400 units and safety stock is 1.645 × 20 × 2 = 65.8 units. The reorder point is therefore approximately 466 units.

The calculation assumes independent daily demand and a stable distribution. If demand is autocorrelated, seasonal, intermittent, or affected by promotions, the square-root relationship may not hold.

Variable Lead Time with Constant Demand

If demand per period is constant at d but lead time has standard deviation sigma_L, then the standard deviation of lead-time demand is approximately:

sigma_LT = d sigma_L.

The reorder point becomes:

ROP = d times average lead time + z d sigma_L.

This form is particularly useful when demand is predictable but supplier or transportation performance is inconsistent.

Simultaneous Demand and Lead-Time Variability

If demand and lead time are independent, and average demand is d, demand standard deviation is sigma_d, average lead time is L, and lead-time standard deviation is sigma_L, a widely used approximation is:

Variance of lead-time demand = L sigma_d squared + d squared sigma_L squared.

Therefore:

Safety stock = z square root of (L sigma_d squared + d squared sigma_L squared).

The first term captures demand variation during the average lead time; the second captures variation caused by changes in lead time. If demand and lead time are correlated, a covariance term should be included. Positive correlation increases risk and makes the simple independent formula understate required protection.

In practice, many organizations estimate the empirical distribution of lead-time demand directly from historical demand and receipt data. This can be preferable to forcing separate normal distributions onto variables that are skewed or intermittent.

Service Levels and the Economics of Protection

Service level is not one single metric. Two measures are especially important.

The cycle service level is the probability that inventory will not run out during a replenishment cycle. It treats a cycle with one unit short as unsuccessful, regardless of the quantity short.

The fill rate measures the fraction of demand fulfilled immediately. A system can have a high cycle service level but a lower fill rate if occasional stockouts are severe. Conversely, a system may experience some stockout cycles but satisfy most unit demand.

The target service level should reflect the marginal cost of additional protection. Raising a target from 90 to 95 percent may require a modest increase in safety stock, while raising it from 98 to 99.9 percent may require disproportionately more inventory. A rational policy compares:

• the carrying and obsolescence cost of additional inventory;
• the expected cost of lost margin, expediting, disruption, or customer dissatisfaction;
• the strategic value of availability;
• the capacity and cash-flow consequences of the policy.

ABC classification alone does not determine service level. A low-cost item may be operationally critical, while a high-cost item may have easily available substitutes. A more mature approach combines financial value, demand variability, criticality, substitutability, lead-time risk, and customer promise.

Continuous-Review Inventory Policies

In a continuous-review system, inventory position is monitored continuously and an order is placed whenever the reorder point is reached. The familiar (Q, R) policy orders a fixed quantity Q when inventory position falls to R.

The order quantity may be based on the economic order quantity, supplier minimums, container sizes, production constraints, or a jointly optimized replenishment policy. The reorder point is based primarily on expected lead-time demand and the selected protection level.

Continuous review is valuable when stockouts are costly, transactions are captured promptly, and replenishment can be triggered economically. It is common for high-value, critical, or fast-moving items. Its limitations include system and monitoring costs, dependence on accurate transaction records, and possible overreaction to short-term demand signals.

The policy must use inventory position rather than on-hand inventory alone. If 300 units are on hand and 500 units are already on order, treating the position as 300 may create unnecessary replenishment. Conversely, ignoring backorders can produce a false impression of availability.

Periodic-Review Policies

Under periodic review, inventory is assessed at fixed intervals, such as weekly or monthly. At each review, the organization orders enough to raise inventory position to a target or order-up-to level S. The period between reviews is called the review period P.

Because the system cannot respond until the next review, protection is needed for the protection interval, which is approximately P + L. If demand is variable and lead time is constant:

S = d(P + L) + z sigma_d square root of (P + L).

The order quantity at a review is:

Order quantity = S − inventory position.

Periodic review is convenient when suppliers visit on a schedule, transportation is consolidated, or administrative simplicity is important. However, it generally requires more safety stock than continuous review because the system must protect against demand during the review period as well as lead time.

A periodic system may be inappropriate for an item with highly erratic demand and severe shortage consequences unless review frequency is short and information is reliable.

Base-Stock and Order-Up-to Logic

A base-stock policy attempts to maintain inventory position near a specified level. Each unit of demand triggers a replenishment order, at least conceptually, so that the inventory position returns to the base-stock level. The policy is particularly relevant in continuous replenishment, e-commerce fulfilment, spare-parts systems, and multi-echelon inventory coordination.

The base-stock level represents expected demand over the relevant replenishment exposure plus protection against uncertainty. In a single-stage system, that exposure is normally lead time. In a multi-echelon system, the base-stock level must reflect the echelon lead time and the position of downstream inventory. Poorly chosen local base stocks can cause the total supply chain to hold excessive inventory even when each facility appears individually protected.

Base-stock policies are responsive but can generate frequent small orders. They work best when ordering and transaction costs are low, inventory records are accurate, and suppliers can respond flexibly.

The Newsvendor Model for Single-Period Decisions

Continuous replenishment models are unsuitable when inventory cannot be carried meaningfully into the next selling period. Examples include seasonal fashion, fresh food, newspapers, event merchandise, and products with severe end-of-season obsolescence. The newsvendor model determines an order quantity for a single period with uncertain demand.

Let:

• Cu = underage cost, the cost of ordering one unit too few;
• Co = overage cost, the cost of ordering one unit too many.

The critical fractile is:

Critical ratio = Cu / (Cu + Co).

The optimal order quantity is the demand percentile corresponding to this ratio. If the cost of a shortage is high relative to the cost of excess inventory, the critical ratio and order quantity are high. If excess stock is expensive, the quantity is lower.

For example, a seasonal product sells for 80 monetary units, costs 50, and can be salvaged for 30. The underage cost is the lost contribution margin, 80 − 50 = 30. The overage cost is 50 − 30 = 20. The critical ratio is 30/(30+20) = 0.60. The firm should order approximately the 60th percentile of the demand distribution.

The newsvendor model is not simply a method for maximizing expected sales. It explicitly balances the economics of shortage and excess. Its quality depends on realistic estimates of salvage value, lost margin, demand distribution, and the timing of the decision.

Lost Sales and Backorders

When inventory is unavailable, demand can be treated in two fundamentally different ways. In a backorder system, the customer waits and the demand remains in the system. In a lost-sales system, the demand disappears from the firm’s order book.

Backorders increase future replenishment requirements and may create a queue of unmet demand. They are common in industrial supply, configured equipment, and customer relationships where switching suppliers is difficult. Lost sales are more common in retail and highly competitive markets where customers can substitute or abandon the purchase.

The distinction affects both forecasting and inventory calculations. In a backorder environment, observed sales during a stockout understate true demand because demand is recorded later. In a lost-sales environment, observed sales also understate demand, but the missing demand may never be observed directly. Forecasting systems must therefore use censored-demand methods, customer surveys, substitution information, or lost-sales estimation.

The shortage cost also differs. A backorder may generate administrative and expediting costs, while a lost sale may involve lost contribution margin, customer lifetime value, and reputational damage. Treating all shortages as equivalent can produce systematically incorrect service targets.

A Practical Decision Process

A sound inventory policy can be developed through the following sequence:

1. Define the customer promise and the shortage consequence.

2. Establish the unit of analysis: SKU, location, echelon, or product family.

3. Measure demand using sales, orders, substitutions, backorders, and stockout information.

4. Measure actual lead time from order release to usable receipt, not merely quoted supplier lead time.

5. Select the service measure: cycle service level, fill rate, or another operational metric.

6. Estimate demand during the protection interval and its variability.

7. Select the policy structure: continuous review, periodic review, base-stock, or single-period.

8. Set parameters and simulate performance under realistic demand and lead-time scenarios.

9. Monitor bias, stockouts, excess inventory, supplier performance, and parameter drift.

10. Review the policy when products, suppliers, customer promises, or operating conditions change.

Illustrative Case: Regional Distribution of Diagnostic Kits

Background

A regional distributor supplies diagnostic kits to hospitals. The kits have moderate unit value but high service criticality because hospitals schedule testing around expected deliveries. The distributor operates a continuous-review system.

Situation and Problem

Average daily demand is 240 kits, with a daily standard deviation of 60. Supplier lead time averages five days with a standard deviation of 1.5 days. Demand and lead time are assumed independent for initial analysis. Management wants an approximate 97.5 percent cycle service level, corresponding to z ≈ 1.96.

Analysis

The standard deviation of lead-time demand is approximated by:

square root of [L sigma_d squared + d squared sigma_L squared]

= square root of [5(60 squared) + 240 squared(1.5 squared)]

= square root of [18,000 + 129,600]

≈ 384.2 kits.

Expected lead-time demand is 240 × 5 = 1,200 kits. Safety stock is 1.96 × 384.2 ≈ 753 kits. The approximate reorder point is therefore 1,953 kits.

This result is large because lead-time variation contributes substantially to risk. If the distributor reduced lead-time standard deviation from 1.5 days to 0.5 days, the standard deviation of lead-time demand would fall to approximately square root of [18,000 + 14,400] = 180 kits. Safety stock at the same service level would then be about 353 kits, a reduction of roughly 400 kits.

Approach and Strategy

The distributor should not rely only on carrying additional stock. It should negotiate delivery windows, use advance shipment notices, qualify a secondary supplier, and distinguish emergency hospital orders from routine demand. It should also validate whether stockouts create backorders or lost sales and measure fill rate separately from cycle service level.

Outcome

The analytical exercise shows that improving supplier reliability can release more working capital than simply reducing the service target. It also demonstrates why average lead time alone is insufficient for policy design.

Learning Points

The case illustrates four principles: variability is often more important than the average; service targets have financial consequences; operational improvements can substitute for inventory; and safety-stock formulas must be connected to actual business processes.

Challenges, Limitations, and Common Errors

Inventory models are simplifications. The normal distribution may be inappropriate for intermittent, lumpy, or highly skewed demand. Historical data may be contaminated by prior stockouts, promotions, exceptional disruptions, or changes in assortment. Forecast error may be correlated across products, and supplier lead times may have fat-tailed distributions in which rare delays are much worse than the standard deviation suggests.

Common errors include:

• using sales as demand without correcting for stockouts;
• applying the same service level to every SKU;
• using quoted rather than observed lead times;
• adding safety stock independently at every echelon;
• confusing cycle service level with fill rate;
• ignoring open orders and backorders in inventory position;
• treating forecast bias as random variability;
• failing to account for minimum order quantities and pack sizes;
• assuming that a higher service level is always economically superior.

A formula also cannot resolve poor master data, delayed transaction posting, inaccurate bills of material, or unreliable supplier confirmations. Governance and process discipline are prerequisites for analytical inventory control.

Technology, Analytics, and Digital Application

Enterprise resource planning and warehouse management systems provide the transaction foundation for inventory policy. Advanced planning systems can calculate dynamic safety stocks, incorporate forecast distributions, and distinguish demand by customer segment. Statistical forecasting can estimate seasonality and promotional effects, while machine-learning methods may improve prediction when sufficient explanatory data exist.

However, sophisticated algorithms do not automatically produce better decisions. A black-box recommendation may be difficult to audit, may learn from censored stockout data, or may overreact to temporary events. Practical governance requires parameter ownership, exception thresholds, version control, explainable inputs, and regular back-testing.

Digital twins and Monte Carlo simulation can test service and cost outcomes under alternative demand and lead-time scenarios. Supplier analytics can identify whether inventory is compensating for avoidable lead-time variability. RFID, barcode scanning, and real-time location data improve inventory-position accuracy, which is essential for continuous-review policies.

Current Developments and Best Practices

Contemporary inventory management increasingly emphasizes resilience and segmentation rather than a single enterprise-wide service target. Firms are using scenario-based safety stock, multi-echelon optimization, supplier risk data, and differentiated customer promises. Sustainability considerations also matter: excess inventory can create waste, returns, markdowns, and unnecessary storage energy, while shortages may trigger costly emergency transport.

Best practice is to combine statistical protection with operational action. Managers should reduce variability where possible, pool risk across locations when feasible, make shortage costs explicit, and review parameters after structural changes. Service targets should be visible to sales, procurement, finance, and operations rather than being set by one function in isolation.

Ethically, organizations should not promise unrealistic availability to customers merely to win contracts and then shift shortage costs to vulnerable suppliers or frontline workers. In healthcare and other critical sectors, allocation rules during shortages should be transparent, defensible, and consistent with regulatory and professional obligations. Data used for customer prioritization must also be governed carefully so that automated decisions do not create unjustified discrimination.

Chapter Summary

Inventory under uncertainty requires protection against variation in demand, lead time, or both. The reorder point combines expected demand during the protection interval with safety stock. Safety stock depends on variability, the chosen service definition, and the desired probability of avoiding shortage. Continuous-review policies respond when inventory position reaches a reorder point, while periodic-review policies replenish at review intervals and must protect against the longer review-plus-lead-time interval. Base-stock policies maintain a target inventory position and are useful where replenishment can be triggered frequently. The newsvendor model applies to single-period decisions and balances underage and overage costs through a critical fractile. Lost sales and backorders have different forecasting, cost, and service implications. Effective practice combines quantitative models with data quality, supplier management, segmentation, simulation, governance, and continuous review.

Key Terms

Safety stock: Inventory held to protect against demand or replenishment uncertainty.

Reorder point: Inventory position at which a replenishment order is released.

Lead-time demand: Demand expected during the time between order release and receipt.

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

Cycle stock: Inventory used to satisfy normal demand between replenishment deliveries.

Cycle service level: Probability of completing a replenishment cycle without a stockout.

Fill rate: Percentage of total demand supplied immediately from available inventory.

Protection interval: Time span for which inventory must provide coverage before replenishment becomes available.

Continuous review: Inventory monitoring system in which replenishment can be triggered at any time.

Periodic review: System in which inventory is assessed at predetermined intervals.

Base-stock policy: Policy that attempts to maintain inventory position at a specified target.

Order-up-to level: Target inventory position to which a periodic-review system replenishes.

Demand variability: Dispersion of demand around its expected value.

Lead-time variability: Dispersion of replenishment time around its expected value.

Newsvendor model: Single-period model balancing the costs of excess and insufficient inventory.

Underage cost: Economic consequence of ordering one unit less than needed.

Overage cost: Economic consequence of ordering one unit more than needed.

Backorder: Unfilled demand retained for later fulfilment.

Lost sale: Demand that is not fulfilled and does not remain as a future obligation.

Multi-echelon inventory: Inventory positioned at several linked stages of a supply network.

Review Questions

Short Answer Questions

1. What is the difference between cycle stock and safety stock?

2. Why must inventory position include on-order inventory?

3. Distinguish cycle service level from fill rate.

4. What is meant by the protection interval in a periodic-review system?

5. Define underage cost and overage cost in the newsvendor model.

Long Answer Questions

1. Explain how demand and lead-time variability jointly determine safety stock and the reorder point.

2. Compare continuous-review, periodic-review, and base-stock policies in terms of responsiveness, inventory requirements, and data needs.

3. Explain why a high cycle service level does not necessarily imply a high fill rate.

4. Discuss how lost sales and backorders affect forecasting, shortage costs, and inventory policy.

5. Explain the assumptions and managerial meaning of the newsvendor critical-ratio model.

Analytical / Application Questions

1. Daily demand averages 150 units with a standard deviation of 30 units. Lead time averages six days with a standard deviation of two days. Assuming independence and a z-value of 1.645, estimate safety stock and reorder point.

2. A retailer reviews stock every seven days, has a lead time of four days, average daily demand of 80 units, and daily standard deviation of 15 units. Using a z-value of 2.05, calculate an approximate order-up-to level.

3. A seasonal item has an underage cost of 24 monetary units and an overage cost of 16 monetary units. Determine the critical ratio and explain what demand percentile should guide the order quantity.

22. Multiple Choice Questions with Answers

Question 1

Daily demand has a standard deviation of 20 units and lead time is constant at four days. Assuming independent daily demand, what is the standard deviation of lead-time demand?

A. 20 units

B. 40 units

C. 80 units

D. 160 units

Correct Answer: B. 40 units

Question 2

Which expression correctly represents a reorder point under uncertainty?

A. Average order quantity plus annual demand

B. Expected lead-time demand plus safety stock

C. Safety stock minus expected lead-time demand

D. Annual demand divided by lead time

Correct Answer: B. Expected lead-time demand plus safety stock

Question 3

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

A. Only the review period

B. Only the lead time

C. The review period plus the lead time

D. The economic order cycle minus the lead time

Correct Answer: C. The review period plus the lead time

Question 4

Which statement best distinguishes cycle service level from fill rate?

A. Cycle service level measures order quantity, while fill rate measures lead time

B. Cycle service level concerns the probability of no stockout in a cycle, while fill rate concerns the proportion of demand immediately supplied

C. They are two names for exactly the same measure

D. Fill rate applies only to periodic review

Correct Answer: B. Cycle service level concerns the probability of no stockout in a cycle, while fill rate concerns the proportion of demand immediately supplied

Question 5

If lead-time variability is high but demand is relatively stable, which action can reduce safety stock without lowering the target service level?

A. Increase forecast bias

B. Improve supplier and transportation reliability

C. Ignore on-order inventory

D. Increase the number of backorders

Correct Answer: B. Improve supplier and transportation reliability

Question 6

In the newsvendor model, the critical ratio is calculated as:

A. Overage cost divided by underage cost

B. Underage cost divided by the sum of underage and overage costs

C. Selling price divided by purchase cost

D. Demand mean divided by demand standard deviation

Correct Answer: B. Underage cost divided by the sum of underage and overage costs

Question 7

What is the principal operational difference between a backorder and a lost sale?

A. A backorder remains as future demand, while a lost sale does not

B. A lost sale always has zero economic cost

C. A backorder occurs only in periodic-review systems

D. A lost sale increases inventory position

Correct Answer: A. A backorder remains as future demand, while a lost sale does not

Question 8

A firm uses a continuous-review (Q, R) policy. What normally triggers replenishment?

A. The calendar reaches the end of the month

B. Inventory position falls to the reorder point

C. Demand becomes exactly equal to average demand

D. The warehouse reaches maximum capacity

Correct Answer: B. Inventory position falls to the reorder point

Question 9

Why can using observed sales during stockout periods produce a biased forecast?

A. Stockouts cause demand to be recorded perfectly

B. Observed sales may understate true demand because some demand was censored or lost

C. Lead time becomes zero during stockouts

D. Safety stock is automatically counted as sales

Correct Answer: B. Observed sales may understate true demand because some demand was censored or lost

Question 10

A product has underage cost 30 and overage cost 20. What is its newsvendor critical ratio?

A. 0.40

B. 0.50

C. 0.60

D. 1.50

Correct Answer: C. 0.60