Inventory and Revenue Optimization

Introduction

Inventory decisions determine whether an organization can meet customer demand economically and reliably. Too little inventory causes stockouts, lost sales, production interruptions, and reputational damage. Too much inventory ties up working capital, increases storage and insurance costs, creates obsolescence risk, and may conceal quality or process problems. Revenue optimization extends this problem by recognizing that inventory and pricing are interdependent: a firm must decide not only how much to stock, but also which customers to serve, at what price, and through which allocation rule.

Operations research provides models for these decisions under both certainty and uncertainty. Deterministic models assume that relevant demand and lead-time parameters are known. Probabilistic models explicitly represent variation in demand, lead time, or both. Multi-period models coordinate decisions over time, while revenue-management models use prices, booking limits, and inventory protection to allocate scarce capacity.

This chapter develops the principal models, explains their assumptions, and connects them to practical decisions in manufacturing, retail, health care, logistics, airlines, hospitality, and digital commerce.

Learning Objectives

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

1. Explain the economic role and major components of inventory.
2. Formulate and interpret deterministic order-quantity and production models.
3. Calculate reorder points, safety stock, cycle service levels, and fill rates.
4. Analyze quantity discounts and identify when a discount is economically worthwhile.
5. Apply newsvendor logic to single-period and perishable inventory decisions.
6. Describe multi-period replenishment and dynamic lot-sizing approaches.
7. Explain how capacity, pricing, and inventory protection interact in revenue optimization.
8. Evaluate model assumptions, implementation risks, and ethical implications.

1. Inventory as an Operations Decision

Inventory is a stock of materials, components, work in process, finished goods, spare parts, or service capacity held for future use. Its economic functions include:

• Cycle stock: inventory created by ordering or producing in batches.
• Safety stock: inventory held against uncertainty in demand or replenishment.
• Pipeline inventory: goods currently in transit or otherwise committed but not yet available.
• Anticipation inventory: stock built before predictable seasonal demand, promotions, strikes, or capacity constraints.
• Decoupling inventory: stock that separates stages of a process so that one disruption does not immediately stop another stage.

The principal inventory costs are ordering or setup cost, holding cost, purchase cost, shortage cost, and disposal or obsolescence cost. A useful annual cost approximation is:

Here, D is annual demand in units per year, C is unit purchase cost, Q is order quantity in units, S is ordering cost per order, and H is annual holding cost per unit. The expression assumes constant demand, instantaneous replenishment, no shortages, and a constant unit price. The purchase term is constant with respect to Q when no quantity discount exists.

A sound inventory policy should be evaluated using total relevant cost rather than purchase price alone. Working-capital impact, service requirements, capacity, sustainability, and risk are also important.

2. Deterministic Inventory Models

2.1 Economic Order Quantity

The classical economic order quantity (EOQ) balances ordering cost against cycle-stock holding cost. Average cycle stock is Q/2 because inventory declines linearly from Q to zero between replenishments.

The optimal number of orders per year is:

The time between orders, in years, is 1/N^*, or in working days, approximately:

where W is the number of working days per year. At the optimum, annual ordering cost equals annual cycle-stock holding cost:

Assumptions. EOQ assumes known and constant demand, constant lead time, instantaneous replenishment, no quantity discounts, no shortages, and independent ordering decisions. These assumptions are often unrealistic, but the model remains valuable as a benchmark.

Worked example. Annual demand is 12,000 units, ordering cost is $75 per order, and annual holding cost is $3 per unit.

Orders per year are approximately 12,000/775 = 15.5. Annual ordering cost is approximately 15.5 × $75 = $1,162.50, and annual cycle-stock holding cost is approximately (775/2) × $3 = $1,162.50. Their equality verifies the calculation, apart from rounding.

2.2 Reorder Point with Deterministic Lead Time

If demand is constant at d units per day and lead time is L days, the reorder point is:

The reorder point is a timing rule, not an order-size rule. When the inventory position reaches the reorder point, an order of size Q is placed. Inventory position equals on-hand inventory plus on-order inventory minus backorders.

2.3 Economic Production Quantity

When replenishment is gradual because the firm produces internally, the economic production quantity (EPQ) applies. Let p be the production rate and d the demand rate, with p > d. Maximum inventory is less than the production lot because demand continues during production.

The factor (1 − d/p) adjusts for simultaneous production and consumption. If p approaches d, replenishment becomes slow and the optimal production lot increases. The model is applicable only when production rate exceeds demand rate and production is sufficiently stable.

3. Probabilistic Inventory Models

Demand and lead time are rarely constant. A probabilistic policy separates two decisions:

1. How much to order, often represented by Q.
2. When to order, represented by the reorder point.

Safety stock protects against uncertainty during the replenishment interval. The cycle service level is the probability that no stockout occurs during a replenishment cycle. It differs from the fill rate, which is the proportion of total demand immediately satisfied from available stock.

3.1 Safety Stock under Variable Demand

Assume independent, normally distributed daily demand with mean μ_d and standard deviation σ_d, and constant lead time L days. Lead-time demand has mean:

and standard deviation:

For a target cycle service level associated with standard-normal quantile z, safety stock is:

The reorder point is:

Units are consistent: daily demand multiplied by days gives units, and the standard deviation is also in units.

Worked example. Daily demand has mean 100 units and standard deviation 20 units. Lead time is 4 days. For a 95% cycle service level, use z ≈ 1.645. Then:

Thus, an order is triggered when inventory position falls to approximately 466 units.

3.2 Variable Demand and Variable Lead Time

If daily demand and lead time are independent, with mean demand μ_d, demand standard deviation σ_d, mean lead time μ_L, and lead-time standard deviation σ_L, the variance of lead-time demand is approximately:

Therefore:

This approximation assumes independence and uses a demand-rate representation. If demand and lead time are correlated, a covariance term is required:

where Cov(L, D̄) represents covariance between lead time and the relevant demand-rate measure. Positive correlation increases required protection; ignoring it can materially understate stockout risk.

3.3 Service-Level Selection

A higher service level increases safety stock nonlinearly because the normal quantile rises rapidly in the upper tail. Service targets should therefore be differentiated by product importance, margin, substitutability, customer promise, and shortage consequences. A 99% target may be justified for a critical medical component but uneconomic for a low-margin, easily substituted item.

The expected shortage cost should be compared with the marginal holding cost of additional safety stock. Forecast accuracy, lead-time reliability, supplier flexibility, and postponement may improve service at lower cost than simply adding inventory.

4. Single-Period Inventory and the Newsvendor Model

The newsvendor model applies when inventory has one principal selling opportunity and leftover stock has a different value from a stockout. Examples include fresh food, newspapers, fashion items, event seats, and seasonal products.

Let c be unit cost, p selling price, v salvage value per unsold unit, and D random demand. Under the usual conditions v < c < p, the underage cost is:

and the overage cost is:

The critical fractile is:

where F is the cumulative distribution function of demand. The optimal order quantity is the demand quantile corresponding to this probability.

Worked example. A seasonal item costs $20, sells for $50, and has salvage value of $5. Then:

The firm should stock the 66.7th percentile of demand. If demand is normally distributed with mean 1,000 and standard deviation 200, the corresponding standard-normal quantile is approximately z = 0.43, giving:

The result reflects the asymmetry of the two errors: a stockout costs $30 in lost contribution, while an unsold unit costs $15 after salvage.

5. Quantity Discounts

Suppliers may offer all-units discounts, in which every unit receives the lower price once the order reaches a threshold. Discounts can reduce purchase cost but increase cycle-stock holding cost and possibly obsolescence risk.

For each price bracket, calculate the EOQ using that bracket’s holding cost. If holding cost is a percentage i of unit price C, then:

The candidate is feasible only if it lies within the relevant bracket. If not, use the smallest quantity that qualifies for the bracket. Compare the total annual cost of all feasible candidates:

Illustrative example. Annual demand is 10,000 units, ordering cost is $100, and holding rate is 20% of price. The supplier offers $10 per unit below 1,000 units and $9.50 per unit at 1,000 units or more.

At $10, H = 0.20 × 10 = $2, so:

At $9.50, H = $1.90, and the unconstrained EOQ is approximately 1,026 units, which is feasible. Compare approximately:

• At 1,000 units and $10: purchase cost $100,000; ordering cost $1,000; holding cost $1,000; total $102,000.
• At 1,026 units and $9.50: purchase cost $95,000; ordering cost about $975; holding cost about $975; total approximately $96,950.

The discount is economically attractive in this example. In practice, the analysis must also include storage capacity, cash-flow constraints, shelf life, supplier reliability, and the possibility that the discount encourages unnecessary consumption.

6. Multi-Period Replenishment

A multi-period problem involves decisions across several time periods, often with time-varying demand, capacity, prices, or costs. A static EOQ may perform poorly when demand is seasonal or when inventory carried from one period affects future decisions.

6.1 Periodic-Review Policies

Under a periodic-review policy, inventory is checked every R periods and replenished to an order-up-to level S. A common target is:

where R is review interval, L is lead time, and μ_R+L and σ_R+L are the mean and standard deviation of demand over the protection period R + L. The policy protects against demand until the next review plus supplier lead time. It usually requires more safety stock than continuous review because uncertainty accumulates over a longer period.

6.2 Wagner–Whitin Dynamic Lot Sizing

For deterministic, time-varying demand with zero lead time, fixed ordering cost, and no quantity discount, dynamic lot sizing can determine when to order and how much to cover. The Wagner–Whitin principle states that an optimal solution can be found in which an order is placed only when inventory is zero immediately before receipt, assuming nonnegative holding costs and no speculative benefit from early purchasing.

A dynamic-programming formulation is:

Here, C_t is minimum cost through period t, S is setup cost, h is holding cost per unit per period, and d_k is demand in period k. The summation is the holding cost incurred when an order in period j covers demand through period t. The recurrence compares all possible last order periods. For large or uncertain problems, mixed-integer programming, rolling-horizon optimization, or stochastic dynamic programming may be more appropriate.

7. Revenue Optimization and Inventory Allocation

Revenue optimization concerns the allocation of limited capacity among demand classes with different prices, arrival patterns, and cancellation behavior. Inventory may mean physical units, hotel rooms, aircraft seats, advertising impressions, appointment slots, or data-center capacity.

7.1 Capacity and Price Trade-Off

Suppose a firm has capacity K and two customer segments. Segment 1 pays a high price but arrives later; segment 2 pays a lower price and arrives earlier. Accepting every low-price request may fill capacity before high-price demand appears. The opportunity cost of using one unit for a low-price customer is the contribution that could have been earned from the best displaced alternative.

A simple protection-level rule reserves y units for the high-price segment. The firm accepts low-price demand only while remaining capacity exceeds the protection level. Under uncertain high-price demand, the protection level can be derived using a critical-fractile logic analogous to the newsvendor model. If the high-price contribution is r_H, low-price contribution is r_L, and high-price demand is random, reserve capacity should reflect the probability that the marginal unit is needed by the high-price segment.

7.2 Dynamic Pricing

Prices influence demand and therefore the value of inventory. A price reduction may increase volume but consume capacity that could be sold later at a higher price. A price increase may protect capacity but reduce utilization. Optimization models commonly use demand functions such as:

where p is price and a, b are estimated parameters. A linear demand model is convenient but not universally valid; demand may be nonlinear, price-sensitive only within a range, or affected by competitor prices and stock visibility.

Revenue optimization should consider contribution rather than revenue alone:

A high-revenue product can be inferior if it consumes disproportionate capacity or has high variable cost. Inventory availability, substitution, cancellation, and customer fairness must be included in the decision design.

Figure: Inventory and revenue decisions should be integrated through demand, uncertainty, capacity, and performance evaluation.

8. Applied Case: Perishable Medical Kits

Background

This illustrative case concerns a regional hospital that uses sterile procedure kits. Kits have a short shelf life, demand varies by day, and emergency stockouts can delay treatment. The purchasing department historically ordered monthly using average demand, while the operating theatre experienced both expiry and occasional shortages.

Problem

Average daily demand was 38 kits, with a standard deviation of 10. Supplier lead time averaged 3 days with a standard deviation of 1 day. Demand and lead time were treated as independent. Management wanted approximately 97.5% cycle service for emergency procedures, but storage capacity was limited to 220 kits.

Using the variable-demand and variable-lead-time approximation:

For a 97.5% target, z ≈ 1.96, so:

The calculated reorder point exceeded storage capacity. Simply increasing inventory was infeasible.

Analysis and Solution

The hospital adopted a coordinated response:

1. It negotiated a supplier service agreement reducing lead-time variation.
2. It introduced twice-weekly review rather than monthly review.
3. It classified emergency kits separately from routine kits.
4. It used expiry-first issuing and monitored lot age.
5. It established a lower routine service target while preserving emergency capacity.
6. It created a supplier-held buffer for infrequently used specialist kits.

The solution recognized that safety stock is not the only lever. Reducing uncertainty, shortening review cycles, postponing final kit assembly, and segmenting service requirements can all improve effective availability.

Outcome and Learning Points

In an implementation pilot, expiry declined while emergency availability improved. The main lesson is managerial rather than computational: a mathematically correct reorder point may be operationally impossible. Inventory policy must be integrated with supplier contracts, storage capacity, product criticality, review frequency, and clinical governance. The hospital also had to ensure that cost reduction never overrode patient safety.

9. Implementation, Technology, and Governance

Inventory models depend on data quality. Important data controls include accurate item masters, units of measure, lead-time records, order quantities, supplier performance, demand history, returns, substitutions, and stock adjustments. Forecast error should be measured out of sample rather than inferred from fit to historical data.

Enterprise resource planning systems, warehouse-management systems, radio-frequency identification, Internet-of-Things sensors, and cloud analytics can provide near-real-time inventory visibility. Machine-learning forecasts may improve performance when demand is nonlinear or affected by many variables, but they do not remove the need for inventory-policy logic. A highly accurate forecast can still produce poor decisions if service costs, capacity, lead-time uncertainty, or substitution are ignored.

AI-supported replenishment should therefore include explainability, human override, audit trails, drift monitoring, and controls against discriminatory allocation. Pricing algorithms require special care where customers may face different prices, essential goods are scarce, or algorithmic coordination could reduce competition. Hospitals and public agencies should document service priorities and ensure that optimization does not systematically disadvantage vulnerable groups.

10. Limitations and Best Practices

No single inventory model is universally optimal. Important limitations include:

• Normality may be inappropriate for intermittent, lumpy, or highly seasonal demand.
• Independent demand and lead time may be unrealistic during disruptions.
• Average holding cost may omit obsolescence, damage, carbon, and financing effects.
• Service level definitions may be confused or manipulated.
• EOQ may ignore supplier minimums, transport economies, and capacity constraints.
• Revenue models may fail when customers strategically delay purchases or when prices affect long-run loyalty.
• Historical data may reflect previous stockouts and therefore understate true demand.

Best practice is to segment inventory by value, criticality, demand pattern, and supply risk; estimate parameters from clean and sufficiently current data; validate policies through simulation or pilot testing; review assumptions periodically; and evaluate total cost and service jointly. Sensitivity analysis should examine demand growth, lead-time deterioration, price changes, capacity limits, and forecast bias. Robust or scenario-based optimization is often preferable to a solution that is optimal only under one fragile forecast.

Chapter Summary

Inventory optimization balances ordering, holding, shortage, purchase, and obsolescence costs while meeting an appropriate service objective. EOQ provides a deterministic benchmark by balancing ordering and cycle-stock holding costs. EPQ adapts the logic to gradual internal production. Reorder points determine when replenishment should occur, while safety stock protects against demand and lead-time uncertainty.

Probabilistic models distinguish cycle service level from fill rate and use the distribution of lead-time demand to set protection. The newsvendor model addresses single-period decisions by balancing underage and overage costs through a critical fractile. Quantity-discount analysis requires feasible EOQ candidates and total-cost comparison. Multi-period methods, including periodic review and dynamic lot sizing, coordinate decisions over time.

Revenue optimization extends inventory reasoning to scarce capacity and differentiated prices. Protection levels, booking limits, and dynamic pricing should be based on contribution, opportunity cost, demand uncertainty, and customer effects. In practice, model assumptions must be reconciled with storage, supplier reliability, data quality, technology, ethics, and governance. The strongest decisions combine quantitative analysis with operational redesign and managerial judgment.

Key Terms

Inventory: Stock held for future use or sale.

Cycle stock: Inventory resulting from batch ordering or production.

Safety stock: Inventory held to protect against uncertainty.

Pipeline inventory: Inventory in transit or otherwise committed but unavailable.

EOQ: Order quantity minimizing ordering and cycle-stock holding costs under classical assumptions.

EPQ: Production-lot model accounting for gradual replenishment.

Reorder point: Inventory-position level that triggers replenishment.

Lead-time demand: Demand occurring during replenishment lead time.

Service level: A specified measure of the probability or extent of meeting demand.

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

Fill rate: Proportion of demand satisfied immediately from available stock.

Safety stock: Protection quantity above expected demand during the protection period.

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

Critical fractile: Target demand percentile determined by underage and overage costs.

Quantity discount: Lower unit price offered when order quantity reaches a threshold.

Periodic review: Inventory policy in which stock is assessed at fixed intervals.

Dynamic lot sizing: Multi-period method for determining order timing and quantities.

Revenue optimization: Allocation of scarce capacity across prices and customer segments.

Protection level: Capacity reserved for higher-value future demand.

Opportunity cost: Value of the best alternative displaced by a decision.

Review Questions

Short Answer Questions

1. What are the principal assumptions of the classical EOQ model?
2. Distinguish cycle service level from fill rate.
3. Why does safety stock increase with the square root of lead time when daily demand is independent?
4. What is the purpose of the critical fractile in the newsvendor model?
5. Why can a quantity discount be uneconomic even when the unit purchase price is lower?

Long Answer Questions

1. Derive the EOQ expression and explain the economic meaning of each term.
2. Explain how variable demand and variable lead time affect the reorder point.
3. Compare continuous-review and periodic-review inventory policies.
4. Discuss how capacity and pricing interact in revenue optimization.
5. Evaluate the limitations of using normal-distribution safety-stock calculations in practice.

Analytical / Application Questions

1. Daily demand has mean 80 units and standard deviation 15 units. Lead time is 5 days and constant. Calculate safety stock and reorder point for a 95% cycle service level using z = 1.645.
2. A product costs $12, sells for $30, and has salvage value of $4. Calculate the newsvendor critical fractile and explain what it means operationally.
3. A supplier offers a discount at 2,000 units, but warehouse capacity is 1,500 units. Explain how the firm should evaluate the offer and identify at least three alternatives to accepting the discount.

Multiple Choice Questions with Answers

Question 1

Under the classical EOQ assumptions, which two costs are balanced at the optimum?

A. Purchase cost and shortage cost

B. Ordering cost and cycle-stock holding cost

C. Salvage cost and transportation cost

D. Labor cost and quality cost

Correct Answer: B. Ordering cost and cycle-stock holding cost

Question 2

Annual demand is 10,000 units, ordering cost is $50, and annual holding cost is $2 per unit. What is the approximate EOQ?

A. 224 units

B. 500 units

C. 707 units

D. 1,000 units

Correct Answer: C. 707 units

Question 3

Daily demand has mean 100 units and standard deviation 20 units. With a constant four-day lead time, what is the standard deviation of lead-time demand?

A. 20 units

B. 40 units

C. 80 units

D. 400 units

Correct Answer: B. 40 units

Question 4

Which expression represents a reorder point when demand is variable but lead time is constant?

A. ROP = Q/2

B. ROP = d/L

C. ROP = Lμ_d + zσ_d√L

D. ROP = S + H

Correct Answer: C. ROP = Lμ_d + zσ_d√L

Question 5

A product sells for $50, costs $20, and has salvage value of $5. What is its newsvendor critical fractile?

A. 0.333

B. 0.500

C. 0.667

D. 0.875

Correct Answer: C. 0.667

Question 6

Which statement correctly distinguishes fill rate from cycle service level?

A. Fill rate measures the percentage of demand immediately satisfied; cycle service level measures the probability of no stockout in a cycle.

B. Fill rate is always lower than the unit purchase price.

C. Cycle service level measures annual ordering cost.

D. They are identical measures with different names.

Correct Answer: A. Fill rate measures the percentage of demand immediately satisfied; cycle service level measures the probability of no stockout in a cycle.

Question 7

In an all-units quantity-discount analysis, an EOQ calculated for a price bracket is usable only when it:

A. Exceeds annual demand

B. Falls within the bracket or is adjusted to the bracket threshold

C. Produces the highest holding cost

D. Ignores the purchase-price term

Correct Answer: B. Falls within the bracket or is adjusted to the bracket threshold

Question 8

In a periodic-review system, the protection period generally includes:

A. Only the review interval

B. Only supplier lead time

C. The review interval plus supplier lead time

D. The annual planning horizon only

Correct Answer: C. The review interval plus supplier lead time

Question 9

In revenue optimization, a protection level is primarily used to:

A. Reserve capacity for potentially higher-value future demand

B. Eliminate all demand uncertainty

C. Guarantee that every low-price request is accepted

D. Set the supplier's production rate

Correct Answer: A. Reserve capacity for potentially higher-value future demand

Question 10

If high-price demand and lead time are positively correlated, ignoring the correlation will generally:

A. Overstate required safety stock

B. Understate required protection against stockout

C. Have no possible effect

D. Make EOQ equal to zero

Correct Answer: B. Understate required protection against stockout