Inventory under Uncertainty

Introduction

Inventory decisions are straightforward only when demand, replenishment time, and supply availability are known. In practice, a firm faces uncertain customer demand, variable supplier lead times, forecast error, promotions, shortages, transportation disruption, and changing product life cycles. Inventory control under uncertainty is therefore not simply a matter of finding an economic order quantity. It is a problem of balancing the cost of holding inventory against the cost and strategic consequences of stockouts.

For an entrepreneurial firm, this balance is particularly important. A young firm may have limited cash, weak bargaining power with suppliers, and little historical data, yet a stockout can damage reputation and delay growth. Excess inventory can be equally harmful when products become obsolete or cash is tied up in unsold goods. The central managerial task is to choose a policy that provides an economically appropriate probability of meeting demand while preserving financial and operational flexibility.

This chapter develops the principal tools for that task: safety stock, reorder points, service levels, the newsvendor model, and continuous- and periodic-review policies. It also examines uncertainty in both demand and lead time, implementation issues, and the limitations of statistical models.

Learning objectives

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

1. Distinguish cycle stock, safety stock, and inventory-position concepts.
2. Calculate demand during lead time when demand and lead time are uncertain.
3. Determine safety stock and reorder points for common service-level assumptions.
4. Explain cycle service level, fill rate, and their managerial interpretation.
5. Apply the single-period newsvendor model and interpret its critical ratio.
6. Compare continuous-review and periodic-review inventory policies.
7. Evaluate assumptions, data requirements, implementation risks, and ethical implications.

1. The logic of inventory under uncertainty

A replenishment decision must answer two questions: when should an order be placed? and how much should be ordered? Uncertainty affects both questions. If demand during replenishment is higher than expected, an order may arrive after available stock has been exhausted. Safety stock provides protection against this possibility.

Three categories are useful:

• Cycle stock is inventory used to satisfy expected demand between replenishments.
• Safety stock is additional inventory held to protect against uncertainty in demand, lead time, or both.
• Pipeline inventory is stock that has been ordered but has not yet arrived.

The inventory position is

This differs from on-hand inventory. A reorder decision based only on physical stock may place duplicate orders because it ignores goods already in transit.

A policy also requires a definition of service. A high service target is not automatically optimal. It may be justified for a critical medical component, but economically excessive for a low-margin, easily substituted product. Service targets should reflect contribution margin, customer consequence, substitution, contractual obligations, and strategic importance.

2. Demand during lead time

Let:

• D = demand per time period;
• μ_D = mean demand per period;
• σ_D = standard deviation of demand per period;
• L = lead time in periods;
• μ_L = mean lead time;
• σ_L = standard deviation of lead time;
• X_L = demand during lead time.

If lead time is constant at L and period demands are independent, then:

The square-root relationship is important: uncertainty grows with the square root of time, not in direct proportion to time, when independent demand observations are aggregated.

If demand is approximately normal, a reorder point for a specified cycle service level is:

where z is the standard-normal quantile corresponding to the desired probability of not stocking out during a replenishment cycle. The safety stock is therefore:

Common one-sided values are approximately 1.282 for 90%, 1.645 for 95%, and 2.326 for 99% cycle service.

Variable lead time

When lead time varies and demand per period is independent of lead time, a widely used approximation is:

and:

The first variance component reflects demand uncertainty during the average lead time; the second reflects uncertainty caused by the length of the lead time. Units are consistent: both terms inside the square root have units of units².

This formula assumes independent, identically distributed demand across periods, an approximately stable demand process, and independence between demand and lead time. If demand and lead time are correlated, an additional covariance term is needed. In general, for random lead time and demand rate, a more complete approximation is:

The final term is often omitted in simpler textbook treatments but can matter when both sources of variation are substantial. If demand and lead time are correlated, empirical simulation or a covariance-adjusted model is preferable.

3. Worked example: reorder point with uncertain demand and lead time

An online equipment firm sells an average of 80 units per week. Weekly demand has a standard deviation of 20 units. Supplier lead time averages 3 weeks with a standard deviation of 0.5 week. Management wants a 95% cycle service level. Assume demand and lead time are independent and approximately normal.

Expected lead-time demand is:

The standard deviation of lead-time demand is:

For a 95% cycle service level, z ≈ 1.645. Thus:

The reorder point is:

The firm should place a replenishment order when inventory position falls to approximately 327 units. The result is not a guarantee that every cycle will avoid a stockout; under the model, approximately 5% of replenishment cycles may experience demand exceeding supply before replenishment arrives.

4. Service levels and their interpretation

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

Cycle service level

The cycle service level (CSL) is the probability that no stockout occurs during a replenishment cycle. Under a continuous-review, normal-demand model:

A 95% CSL means that approximately 95 out of 100 replenishment cycles are expected to avoid a stockout, assuming the model is valid. It does not say that 95% of customer demand is filled.

Fill rate

The fill rate is the proportion of demand satisfied immediately from available stock. A system may have a high CSL but a lower fill rate if the few stockouts are severe. Conversely, a policy may have a lower CSL but a high fill rate if shortages are small and quickly replenished.

For normally distributed lead-time demand, expected shortage per cycle when the reorder point is ROP = μ_XL + zσ_XL can be expressed as:

where φ(z) is the standard-normal probability density and Φ(z) is the standard-normal cumulative distribution. If Q is the order quantity and expected demand per cycle is E[D_c], an approximate fill rate is:

This expression requires a cycle definition and assumes shortages are measured as units of demand not immediately filled. In real systems, backorders, lost sales, substitutions, and partial shipments must be treated consistently.

Selecting a service target

Service targets should be differentiated rather than imposed uniformly. Consider:

5. The newsvendor model

The newsvendor model addresses a single selling period or a product with one principal replenishment opportunity. It is appropriate for fashion items, event merchandise, fresh food, seasonal products, and products that become obsolete rapidly.

Let:

• Q = order quantity;
• D = random demand;
• C = unit purchase or production cost;
• P = selling price;
• V = salvage value per unsold unit;
• C_u = P − C = underage cost, assuming lost sales;
• C_o = C − V = overage cost.

The optimal quantity satisfies the critical-fractile rule:

Since C_u + C_o = P − V, the critical ratio can also be written:

The rule assumes a single decision, known cost and price, a specified demand distribution, and no meaningful replenishment after the selling period. If shortages are backordered rather than lost, or if multiple replenishment opportunities exist, the model must be modified.

Worked newsvendor example

A festival merchandise seller buys T-shirts for $12 and sells them for $30. Unsold shirts can be salvaged for $4. Demand is normally distributed with mean 500 and standard deviation 120.

The underage and overage costs are:

Therefore:

The standard-normal quantile for 0.6923 is approximately 0.50. Thus:

The result does not mean that demand will be 560 units. It means that ordering 560 units balances the expected marginal cost of one unit too few against the expected marginal cost of one unit too many under the stated assumptions.

6. Continuous-review policies

A continuous-review policy monitors inventory position after every transaction. The common (Q, R) policy orders a fixed quantity Q whenever inventory position reaches or falls below reorder point R. It is often called a fixed-order-quantity policy.

The reorder point protects against demand during lead time. The order quantity may be determined by an economic order quantity, supplier minimum, truckload size, production batch, or managerial constraint. The classical EOQ expression is:

where D is annual demand, S is ordering cost per order, and H is annual holding cost per unit. EOQ is a cycle-stock model; it does not by itself determine safety stock or service level. Its assumptions include relatively stable demand, constant ordering cost, constant holding cost, and no quantity discounts or capacity restrictions.

A continuous-review policy is attractive when transaction data are available in real time, items are important enough to justify monitoring, and stockout consequences are significant. Its disadvantages include information-system requirements, monitoring costs, and the possibility that frequent adjustments amplify forecast noise.

7. Periodic-review policies

Under a periodic-review policy, inventory is reviewed every T time units. The order quantity varies so that inventory position is raised to an order-up-to level S. Because the firm cannot respond until the next review, protection is required for the protection period:

For constant lead time and independent demand, the order-up-to level is:

The order at a review is:

If lead time is variable, its mean and variance should be incorporated into the protection-period demand. A practical approximation is to replace L with the total random protection period and calculate its mean and variance using the same logic used for lead-time demand.

Periodic review is useful when supplier visits, route schedules, administrative routines, or physical counting make continuous monitoring expensive. However, it generally requires more safety stock because demand can occur during the interval until the next review.

Comparison of policies

8. A unified implementation process

A sound inventory design should proceed in the following sequence.

First, define the decision context. Identify whether demand is lost, backordered, substituted, or delayed. Determine whether the item is replenishable, seasonal, perishable, or subject to obsolescence.

Second, establish data quality. Estimate demand mean, variation, seasonality, demand intermittency, lead-time distribution, supplier minimums, and actual stockout history. Do not calculate a precise safety stock from unreliable data.

Third, select the performance measure. Choose CSL, fill rate, on-time-in-full delivery, or a combination. A target must be linked to customer and financial consequences.

Fourth, model uncertainty. Use analytical formulas when assumptions are reasonable. Use empirical distributions, bootstrap methods, or discrete-event simulation when demand is intermittent, lead time is highly skewed, promotion effects are material, or correlations are important.

Fifth, segment items and suppliers. ABC analysis by annual consumption value can be combined with demand variability, criticality, and supply risk. A low-value item may still deserve high attention if its absence stops production.

Sixth, validate and monitor. Compare predicted service with actual service. Track forecast bias, forecast error, supplier lead-time adherence, inventory turns, aged stock, backorders, and emergency shipments. Re-estimate parameters when the operating environment changes.

9. Illustrative case: NovaBrew’s seasonal equipment launch

Background

NovaBrew, an illustrative specialty-equipment start-up, plans to sell a compact brewer during a twelve-week holiday season. The product has a short commercial window, and replenishment from an overseas supplier takes six weeks. A final order must be placed before the season begins, making the newsvendor model more appropriate than a normal continuous-replenishment model.

Problem

The marketing team forecasts normally distributed seasonal demand with mean 2,400 units and standard deviation 600. The purchase cost is $85, selling price is $160, and end-of-season salvage value is $40. The founders must choose a quantity while preserving cash for customer acquisition.

Analysis

Underage cost is:

Overage cost is:

The critical ratio is:

The corresponding standard-normal quantile is approximately 0.318. Therefore:

Approach and managerial evaluation

The calculated quantity is a baseline, not an automatic commitment. Management should test the result against supplier minimums, cash availability, forecast bias, salvage uncertainty, and the possibility that social-media promotion creates a demand distribution with a heavy upper tail. If the forecast is systematically optimistic, using an unadjusted normal distribution may produce excessive inventory. If a stockout would permanently lose customers, the economic underage cost may exceed the lost margin and should be revised.

NovaBrew could also negotiate a split shipment, a return option, or a post-season secondary-market agreement. Such contractual flexibility reduces effective overage cost and can justify a larger order. Conversely, a cash covenant may impose a hard upper limit below the unconstrained optimum.

Outcome

The board approves an initial commitment near 2,591 units, subject to a staged supplier contract and weekly demand monitoring. The policy is not to alter the original single-period calculation mechanically, but to reduce exposure through contractual and financial design.

Learning points

1. The critical ratio translates commercial consequences into a quantity decision.
2. Salvage value and return rights materially affect the order quantity.
3. A mathematical optimum is conditional on the demand distribution and cost assumptions.
4. Entrepreneurship requires combining inventory analytics with financing and contract strategy.

10. Limitations, governance, and technology

Model limitations

Normality may be inappropriate for intermittent, highly skewed, or bounded demand. Forecast errors may be correlated across products, and demand may be affected by price, availability, advertising, and competitor action. Lead time is often not independent of demand: congestion may increase lead time precisely when demand surges. In these cases, historical averages can understate risk.

Safety stock also cannot compensate indefinitely for poor supplier performance. If lead time is unstable because of inadequate supplier processes, management should consider supplier development, dual sourcing, revised incoterms, local inventory, or redesigned products. Holding more stock may conceal a structural problem.

Governance and ethics

Inventory decisions allocate scarce capital and can affect workers, customers, and communities. Managers should avoid deliberately creating artificial scarcity when customers depend on essential goods. Service differentiation must not discriminate unfairly among customers. Data used for demand prediction should be governed for privacy, security, and accuracy. Automated systems should retain human oversight, especially when recommendations affect critical medical, safety, or emergency supplies.

Technology and AI

Enterprise resource planning, warehouse-management systems, barcode scanning, RFID, and supplier portals improve inventory visibility. Machine-learning forecasting can identify nonlinear relationships, but it does not eliminate uncertainty. A model trained on periods of abundant availability may learn censored demand: observed sales equal actual demand only when stock was available. Promotion changes, new products, and structural breaks require human review. AI-generated reorder recommendations should therefore expose data provenance, confidence ranges, assumptions, and override controls rather than presenting false precision.

Best-practice checklist

• Use inventory position, not only on-hand stock, for reorder decisions.
• Separate cycle stock from safety stock in analysis and reporting.
• Model demand and lead-time variability explicitly.
• Match service metrics to customer and financial consequences.
• Use CSL cautiously; supplement it with fill rate and shortage magnitude.
• Use newsvendor logic for genuinely single-period or highly perishable decisions.
• Include review time in periodic-review protection stock.
• Segment items by value, variability, criticality, and supply risk.
• Validate model outputs against actual performance and stockout observations.
• Treat contracts, supplier development, and postponement as alternatives to simply holding more stock.
• Document assumptions and maintain human accountability for automated decisions.

Chapter Summary

Inventory under uncertainty requires decisions about both quantity and timing while demand and replenishment conditions are imperfectly known. Safety stock protects against variation in demand and lead time, while the reorder point combines expected lead-time demand with a selected protection quantity. When lead time is constant, demand variability grows approximately with the square root of lead time. When lead time is variable, both demand variance and lead-time variance contribute to the uncertainty of demand during replenishment.

Cycle service level measures the probability of avoiding a stockout during a cycle, whereas fill rate measures the proportion of demand supplied immediately. These metrics are related but not interchangeable. The newsvendor model handles single-period decisions by balancing underage and overage costs through the critical ratio. Continuous-review policies respond whenever inventory position reaches a reorder point; periodic-review policies replenish at scheduled intervals and must protect demand over the review period plus lead time.

No formula substitutes for sound data, appropriate segmentation, supplier management, and governance. Analytical models should be tested against actual performance and adapted when demand is intermittent, lead times are correlated, or the commercial environment changes.

Key Terms

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

Reorder point: Inventory-position threshold that triggers replenishment.

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

Cycle stock: Inventory used to satisfy expected demand between replenishments.

Pipeline inventory: Ordered inventory that has not yet arrived.

Lead-time demand: Demand occurring while replenishment is in transit.

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

Fill rate: Proportion of demand fulfilled immediately from available inventory.

Underage cost: Economic cost of ordering too little.

Overage cost: Economic cost of ordering too much.

Newsvendor model: Single-period inventory model balancing underage and overage costs.

Critical ratio: Target demand fractile equal to underage cost divided by total mismatch cost.

Continuous review: Inventory monitoring after each relevant transaction.

Periodic review: Inventory monitoring at predetermined time intervals.

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

Protection period: Review interval plus replenishment lead time.

Demand variability: Statistical dispersion in demand over time.

Lead-time variability: Statistical dispersion in supplier replenishment time.

Inventory segmentation: Differentiation of items according to value, risk, variability, or criticality.

Forecast bias: Systematic tendency for forecasts to overstate or understate actual demand.

Review Questions

Short Answer Questions

1. Distinguish cycle stock from safety stock.
2. Why is inventory position more useful than on-hand inventory for triggering replenishment?
3. Define cycle service level and fill rate.
4. What is the purpose of the critical ratio in the newsvendor model?
5. Why does a periodic-review system generally require protection for more than lead time alone?

Long Answer Questions

1. Explain how uncertainty in demand and lead time affects the reorder point. State the assumptions behind the standard approximation.
2. Compare continuous-review and periodic-review inventory policies, including their information requirements, advantages, and risks.
3. Derive and interpret the newsvendor critical-fractile rule using underage and overage costs.
4. Discuss why a high cycle service level does not necessarily imply a high fill rate.
5. Evaluate the role of supplier development, contracts, and postponement as alternatives to increasing safety stock.

Analytical / Application Questions

1. Weekly demand has mean 100 units and standard deviation 25 units. Lead time is constant at 4 weeks. Determine the reorder point for a 95% cycle service level, assuming independent normal demand.
2. A seasonal product sells for $50, costs $20, and has a salvage value of $5. Demand is normal with mean 1,000 and standard deviation 200. Calculate the critical ratio and approximate optimal order quantity.
3. A firm currently uses a weekly periodic review and has a two-week average lead time. Explain how its safety stock calculation would change if the review interval increased to two weeks, and identify at least two operational consequences.

Multiple Choice Questions with Answers

Question 1

Which expression defines inventory position?

A. On-hand inventory − on-order inventory

B. On-hand inventory + on-order inventory − backorders

C. On-hand inventory + backorders − demand

D. On-order inventory − safety stock

Correct Answer: B. On-hand inventory + on-order inventory − backorders

Question 2

Demand has a mean of 50 units per week, a standard deviation of 10 units, and constant lead time of 4 weeks. What is the standard deviation of lead-time demand under independent demand?

A. 10 units

B. 20 units

C. 40 units

D. 50 units

Correct Answer: B. 20 units

Question 3

For a 95% one-sided cycle service level under a normal-demand assumption, which z-value is approximately appropriate?

A. 0.50

B. 1.00

C. 1.645

D. 2.576

Correct Answer: C. 1.645

Question 4

Which statement best distinguishes cycle service level from fill rate?

A. Cycle service level measures average inventory; fill rate measures lead time.

B. Cycle service level measures probability of no stockout in a cycle; fill rate measures demand supplied immediately.

C. They are identical under every inventory policy.

D. Fill rate applies only to single-period products.

Correct Answer: B. Cycle service level measures probability of no stockout in a cycle; fill rate measures demand supplied immediately.

Question 5

A product sells for $30, costs $12, and has salvage value of $4. What is the newsvendor critical ratio?

A. 0.308

B. 0.500

C. 0.692

D. 0.818

Correct Answer: C. 0.692

Question 6

In a periodic-review system, the protection period normally equals:

A. Lead time minus review period

B. Review period plus lead time

C. Order quantity divided by demand

D. Safety stock plus cycle stock

Correct Answer: B. Review period plus lead time

Question 7

In a continuous-review (Q,R) policy, what normally happens when inventory position reaches R?

A. The firm places an order for Q units

B. The firm discards Q units

C. The firm changes the selling price to R

D. The firm stops monitoring inventory

Correct Answer: A. The firm places an order for Q units

Question 8

Which factor increases the reorder point when all other conditions remain unchanged?

A. A lower service target

B. A shorter lead time

C. Greater lead-time variability

D. A lower mean demand

Correct Answer: C. Greater lead-time variability

Question 9

The classical EOQ formula primarily determines:

A. Safety stock needed for a selected service level

B. Cycle-stock order quantity under specified ordering and holding costs

C. The probability distribution of lead time

D. The newsvendor critical ratio

Correct Answer: B. Cycle-stock order quantity under specified ordering and holding costs

Question 10

Why can machine-learning forecasts underestimate true demand during stockouts?

A. They always assume zero lead time

B. Observed sales may be censored because sales cannot exceed available inventory

C. They cannot use historical data

D. They automatically treat salvage value as zero

Correct Answer: B. Observed sales may be censored because sales cannot exceed available inventory