Inventory, Backlog and Service-Level Decisions

Introduction and context

Aggregate production planning (APP) converts forecasts, capacity information and strategic priorities into medium-term decisions about production rates, workforce, subcontracting, inventory and customer service. A plan that balances production and demand only through output and workforce changes is incomplete. In most real systems, demand and supply do not coincide period by period. Inventory absorbs timing differences; backlogs defer fulfilment; lost sales represent demand that disappears when the customer will not wait; and service-level targets impose explicit constraints on the plan.

These choices have economic and strategic consequences. Inventory improves availability but consumes working capital, requires storage and may become obsolete. Backorders preserve the sale but create customer dissatisfaction, expediting costs and reputational risk. Lost sales avoid backlog carrying costs but reduce revenue and customer lifetime value. Service-level targets protect customer experience but may require excess capacity and safety stock. The central managerial problem is therefore not simply to minimize inventory. It is to choose an economically and strategically appropriate combination of availability, responsiveness and resource commitment.

This chapter integrates these decisions into aggregate plans. It distinguishes cycle stock from safety stock, explains the relationship between safety stock and service measures, formulates inventory–backlog balances, evaluates working-capital effects, and presents a practical planning case.

Learning objectives

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

1. Distinguish cycle stock, anticipation inventory, safety stock, pipeline inventory, backorders and lost sales.
2. Formulate an aggregate inventory balance that includes production, demand, backlog and disposal or substitution effects.
3. Explain cycle service level, fill rate and related service measures.
4. Calculate safety stock under commonly used demand and lead-time assumptions.
5. Evaluate the trade-offs among inventory, backlog, lost sales, working capital and customer experience.
6. Incorporate service targets into aggregate planning models and governance processes.
7. Critically assess the limitations of standard inventory formulas in volatile and multi-echelon environments.

1. The role of inventory and unmet demand in aggregate planning

A basic period-by-period balance for a single product or product family is:

where:

• I_t = ending on-hand inventory in period t, measured in units;
• P_t = production or receipts during period t, in units;
• D_t = demand fulfilled from available supply during period t, in units.

This simple equation assumes that all demand is fulfilled immediately and that there is no backlog. In practice, demand may exceed available supply. A more realistic model separates on-hand inventory and backlog. Let B_t denote ending backlog, with all quantities non-negative. If demand is allocated first to available stock, the net inventory position is:

The balance can be written as:

The net position is useful for planning, but it does not reveal whether a positive value represents stock or whether a negative value represents backlog. The operational state is recovered as:

This formulation assumes that production becomes available in the stated period, inventory is measured after receipts and demand allocation, and backorders are carried forward rather than immediately converted to lost sales.

When demand is lost rather than backlogged, let [[MATH)L_t]] be lost sales in period t. Fulfilled demand is then F_t = D_t - L_t, and the balance becomes:

where 0 ≤ F_t ≤ D_t. Lost sales are not inventory in another form; they are demand that is no longer owed to the firm. A model must therefore specify the customer-behaviour assumption: customers may wait, cancel, substitute, or leave permanently.

Inventory as a buffer

Inventory performs several functions:

• Time buffering: production and demand need not occur simultaneously.
• Uncertainty buffering: stock protects against forecast error, supply disruption and variable lead time.
• Economic buffering: inventory permits economical batch production and purchasing.
• Seasonal buffering: production can precede predictable seasonal demand.
• Decoupling: inventories separate operations whose rates or schedules differ.

These benefits must be weighed against holding cost, insurance, handling, damage, obsolescence, shrinkage and the opportunity cost of capital.

Backlog and lost sales

A backorder is an unfulfilled customer requirement that remains valid and is expected to be supplied later. Backorders may be appropriate when products are highly customized, customers have contractual commitments, or the cost of waiting is lower than the cost of holding finished goods.

A lost sale occurs when the customer does not wait. It may result in immediate substitution, cancellation, purchase from a competitor or permanent loss of demand. Treating all unmet demand as backorders systematically understates the commercial impact of shortages. Conversely, treating all shortages as lost sales can overstate the operational penalty when customers are willing to wait.

2. Inventory categories relevant to aggregate plans

Cycle stock is commonly approximated as half the order or production lot when demand is steady and replenishment is instantaneous relative to the planning horizon. This is an approximation, not a universal law. Large minimum order quantities, irregular demand and multi-stage replenishment can make the average substantially different.

Safety stock is not intended to satisfy average demand. Average demand is covered by expected supply; safety stock covers uncertainty around that expectation. Confusing the two leads to inflated inventories and weak diagnosis of forecast or process problems.

3. Service-level concepts and their managerial meaning

Service level is not a single universally defined metric. At least three measures are important.

Cycle service level

The cycle service level (CSL) is the probability that no stockout occurs during a replenishment cycle or specified protection period. If there are n independent cycles and stockouts occur in s cycles, an empirical estimate is:

CSL focuses on whether a shortage occurs at all. It does not indicate how large the shortage is when one occurs.

Fill rate

The fill rate is the proportion of total demand supplied immediately from available stock:

Equivalently, if D is total demand and U is units short, then:

Fill rate is sensitive to shortage magnitude. A system can have a high CSL but a lower fill rate if infrequent stockouts are severe. Conversely, frequent but very small shortages can produce a reasonable fill rate.

Order fulfilment and on-time performance

An order-line service measure may record whether the complete order is delivered by the promised date. This measure is often more closely related to customer experience than a facility-level fill rate. Aggregate planning should therefore define the service promise at the level at which customers experience it: unit, order line, order, shipment or response time.

A target such as “95% service” is incomplete unless the metric, time window, population, exclusions and measurement method are specified. A 95% CSL and a 95% fill rate require different inventories and produce different customer outcomes.

4. Safety stock and protection-period calculations

4.1 Demand variability with fixed lead time

Assume demand per period is independent with mean μ_d units and standard deviation σ_d units, and replenishment lead time is fixed at L periods. Demand during lead time has mean:

and standard deviation:

The second equation assumes independent demand across periods with constant variance. If demand is correlated, the square-root relationship may be inaccurate.

For a one-sided target cycle service level α, the safety stock is:

where z_α is the standard-normal quantile satisfying P(Z ≤ z_α) = α. The reorder point is:

The units are consistent: both μ_L and SS are units, so ROP is units.

Worked example: fixed lead time

A component has average weekly demand of 500 units and a weekly standard deviation of 120 units. Lead time is four weeks. Management wants a 95% cycle service level, for which z_α ≈ 1.645. Then:

The interpretation is not that 395 units will satisfy average demand. The 395 units protect against demand variation during the four-week lead time under the stated assumptions.

4.2 Variable demand and variable lead time

If demand per period and lead time are both variable, and demand and lead time are independent, a commonly used approximation for the variance of demand during lead time is:

where E[L] is mean lead time, σ_L is the standard deviation of lead time, and σ_d is the standard deviation of demand per period. Thus:

This is an approximation based on random-sum demand and independence. If demand and lead time are correlated, or if demand is non-normal, simulation or an empirical quantile of observed lead-time demand is preferable.

For a non-normal demand distribution, the conceptual rule remains:

where Q_α is the α quantile. The normal formula is only a convenient special case.

4.3 From service targets to inventory decisions

A higher service target increases the required quantile and therefore safety stock, but the relationship is nonlinear. Moving from 90% to 95% service may require a moderate addition, whereas moving from 98% to 99.9% can require a disproportionately large buffer. The marginal benefit should be compared with the marginal cost of capital and the value of avoiding shortage.

Service targets should be segmented. A critical spare part, a high-margin product and a low-value predictable item should not automatically receive the same target. Segmentation may consider customer criticality, substitutability, margin, demand volatility, lead time, obsolescence risk and contractual penalties.

5. Integrating inventory, backlog and service into an aggregate plan

An aggregate plan usually uses periods such as weeks or months, while safety-stock calculations may use daily or replenishment-cycle data. The planner must reconcile these levels. A practical model includes:

• opening inventory and backlog;
• regular-time, overtime and subcontracted production;
• forecast demand by period;
• minimum and target ending inventory;
• backlog limits or shortage variables;
• service constraints or penalties;
• capacity and workforce restrictions.

A simplified objective function is:

where:

• c^I = inventory holding cost per unit per period;
• c^B = backlog cost per unit per period;
• c^L = economic penalty per lost unit;
• c^P = production cost per unit or period, as defined;
• c^W = workforce, overtime or capacity-related cost;
• W_t = workforce or capacity decision;
• Σ_t = summation across planning periods.

The objective is meaningful only when costs have compatible time bases and when shortage penalties reflect more than accounting cost. A lost-sale penalty may include contribution margin, contractual charges, expediting, goodwill and estimated customer lifetime value. Because some of these effects are uncertain, scenario analysis is often more defensible than a single precise number.

The inventory balance constraint for a backlog-permitted model is:

with I_t ≥ 0 and B_t ≥ 0. If lost sales are modelled explicitly instead of backlog, then:

where 0 ≤ L_t ≤ D_t and fulfilled demand is D_t - L_t. This equation is useful when demand is exogenous and lost sales do not return later. If unmet demand reduces future demand or generates future recovery orders, a richer customer-response model is required.

Target inventory versus safety stock

A monthly aggregate plan may use a target ending inventory equal to expected demand over a protection period plus safety stock. It should not simply add the same safety stock to every period without considering seasonality, lead-time position, production timing and demand correlation. In a seasonal environment, the correct buffer may be built before the peak, while the end-of-season target may be deliberately reduced to avoid obsolescence.

6. Working capital and customer experience

Inventory affects working capital through several channels. A simplified operating working-capital relationship is:

Inventory investment is often approximated as:

If average inventory is 10,000 units and unit cost is $40, inventory value is $400,000. This value is not necessarily recoverable: obsolete, damaged or slow-moving stock may require a write-down.

The annual carrying cost can be estimated as:

If the carrying rate is 22%, the annual cost in the example is $88,000. The carrying rate should include financing, storage, insurance, handling, shrinkage and obsolescence where appropriate. It should not be applied mechanically to sales price when the relevant investment is production cost.

Backlogs affect working capital differently. They may reduce finished-goods inventory, but they do not create a free benefit. They can increase expediting, customer-service labour, penalties, cancellations and receivables delay. A backlog can also conceal capacity or quality problems if it is treated as a normal planning variable without a limit.

Customer experience depends on more than the percentage of units shipped. Important dimensions include promised-date reliability, completeness, communication, recovery speed and perceived fairness among customers. An organization may deliberately reserve scarce supply for strategic or safety-critical customers, but the allocation rule should be transparent, governed and legally defensible.

7. Economic trade-offs and decision frameworks

Cost-based optimization

Cost-based models are useful when costs and demand are sufficiently measurable. They compare holding, production, backlog and lost-sales costs under constraints. Their limitation is that customer trust and future demand are difficult to price accurately.

Service-constrained optimization

An alternative is to minimize total cost subject to service requirements, such as a minimum fill rate by customer segment. This avoids selecting a low-cost plan that violates the commercial promise. However, constraints must be measurable and achievable; an unrealistic target merely forces infeasibility or hidden workarounds.

Robust and scenario planning

Forecasts, lead times and customer reactions are uncertain. A robust plan tests high-demand, supply-disruption, low-demand and recovery scenarios. Decision-makers can compare:

• expected cost;
• worst-case backlog;
• probability of missing service;
• inventory exposure;
• recovery time;
• capacity flexibility.

A plan that is slightly more expensive in the expected case may be superior if it sharply reduces severe service failures.

Postponement and risk pooling

Postponement delays final customization until demand is clearer, reducing the number of finished-goods variants that require safety stock. Risk pooling aggregates demand across locations or products, often reducing relative variability, although transport time and allocation complexity may increase. These are structural responses to uncertainty and may be more effective than simply increasing every item’s safety stock.

8. Detailed illustrative case: Meridian Home Appliances

Background

Meridian Home Appliances produces a family of countertop water purifiers. Demand is seasonal, components are sourced internationally, and customers expect delivery within seven days. The firm historically measured “service” as monthly production shipped, even when orders were delayed. Inventory was rising while customer complaints increased.

Problem

The aggregate plan for the next four months showed the following opening conditions:

The organization proposed adding 500 finished units to every month’s plan. The supply manager questioned whether this would address the real problem.

Analysis

For the component, assuming independent weekly demand and fixed three-week lead time:

At 95% CSL:

Therefore, the component reorder point is approximately:

This result applies to component availability, not automatically to finished-goods service. The firm’s 300-unit finished-goods backlog also indicates that production timing and order-priority rules are relevant. Adding finished goods without correcting component replenishment, promise-date logic and demand segmentation could increase inventory while leaving shortages unresolved.

The annual carrying cost of one additional finished unit is approximately:

An extra 2,000 units would therefore create approximately $57,600 of annual carrying cost before storage and obsolescence effects. This is not necessarily excessive, but it must be compared with avoided expediting, lost contribution and customer dissatisfaction.

Approach and solution

Meridian adopted four measures:

1. It set separate service targets: 98% order-line fill for contracted institutional customers and 95% for standard consumer orders.
2. It calculated component safety stock from actual weekly demand variability and supplier lead-time performance rather than applying a fixed finished-goods increment.
3. It introduced a backlog ageing limit. Orders older than ten days triggered escalation, customer communication and capacity review.
4. It used postponement: common purifier modules were stocked, while final configuration and packaging were completed after customer-segment demand became clearer.

The plan also changed the performance dashboard. It reported fill rate, promised-date adherence, backlog age, inventory value, obsolete inventory and forecast bias together. A single monthly shipment measure was discontinued.

Outcome

In the illustrative scenario, finished-goods inventory initially increased modestly for priority products, but total variant inventory declined because of postponement. Component availability improved, aged backlog fell, and customer communication made remaining delays more predictable. Working capital was controlled through explicit inventory targets rather than blanket buffers.

Learning points

The case demonstrates five principles. First, service must be defined at the customer-experienced level. Second, safety stock must protect a specified uncertainty over a specified protection period. Third, backlog is a strategic choice only when customers will wait and the organization can recover reliably. Fourth, postponement can reduce uncertainty without sacrificing responsiveness. Fifth, inventory, backlog and service should be governed as one system rather than by separate departmental targets.

9. Implementation, technology and governance

A practical implementation begins with data discipline. Demand history should distinguish true demand from constrained shipments, cancellations and substitutions. Lead-time data should include actual receipt dates, not only supplier promises. Inventory records should identify available, allocated, quarantined, damaged and obsolete stock.

Advanced planning systems can calculate safety stock, simulate scenarios and optimize production across facilities. Machine-learning forecasts may improve demand estimates, but a more accurate forecast does not remove supply variability, structural breaks or judgment requirements. AI-generated recommendations should therefore expose assumptions, confidence ranges and the data used. Human planners remain accountable for service commitments, exception decisions and customer prioritization.

Governance should specify:

• who owns service definitions;
• which customers or products receive differentiated targets;
• how shortage allocation is performed;
• when safety stocks are reviewed;
• how obsolete inventory is authorized for disposition;
• how model assumptions are validated;
• how exceptions and overrides are recorded.

Ethical concerns arise when scarce products are allocated solely by profitability, especially in health, safety or essential-service contexts. Contractual commitments, nondiscrimination, transparency and regulatory obligations may constrain economically attractive allocation rules. Data privacy also matters when customer-level behaviour is used to estimate willingness to wait or cancellation probability.

10. Challenges and limitations

Standard formulas are valuable approximations, but managers should recognize their limits.

1. Non-stationary demand: product launches, promotions and structural changes invalidate historical averages.
2. Correlation: demand may be correlated across periods, locations or products; the square-root rule may then understate risk.
3. Intermittent demand: normal-distribution assumptions are often unsuitable for slow-moving items.
4. Multiple echelons: independently setting safety stock at plant, warehouse and retailer can create excessive total inventory.
5. Capacity interaction: inventory targets cannot be separated from production bottlenecks and recovery capacity.
6. Behavioural response: customers may cancel, substitute or return, so backlog is not always conserved demand.
7. Measurement gaming: a high fill rate can coexist with poor on-time delivery if definitions exclude late or cancelled orders.
8. Cost ambiguity: customer lifetime value, trust and strategic accounts are difficult to represent with one shortage cost.

The appropriate response is not to abandon quantitative models, but to combine them with empirical validation, scenario analysis, segmentation and managerial review.

Best-practice checklist

• Define service at the level customers experience it.
• Separate expected demand coverage from uncertainty protection.
• Use actual lead-time and demand variability data.
• Distinguish backorders from lost sales in both data and models.
• Segment service targets by business consequence and risk.
• Link inventory targets to working-capital limits.
• Test plans under demand, supply and capacity scenarios.
• Monitor backlog age, not only backlog quantity.
• Review safety stocks after structural changes, not merely on a calendar.
• Audit allocation rules for contractual, ethical and regulatory compliance.
• Use technology to expose assumptions rather than conceal them.

Chapter Summary

Inventory, backlog and lost sales are alternative mechanisms for dealing with temporal mismatch between supply and demand. Inventory protects availability but consumes working capital and creates obsolescence risk. Backlog preserves an order when the customer is willing to wait, while lost sales represent demand that the firm no longer owes and may never recover.

The net inventory balance links opening position, production and demand. Safety stock protects against uncertainty and should be calculated for a defined protection period under clearly stated assumptions. Cycle service level measures the probability of avoiding a stockout; fill rate measures the proportion of demand supplied immediately. These metrics are related but not interchangeable.

Aggregate plans should combine production, inventory, backlog, lost-sales assumptions, capacity and service constraints. Economic evaluation must include holding cost, shortage consequences and working-capital exposure, while recognizing that customer trust and future demand are difficult to price. Segmentation, postponement, risk pooling, scenario analysis and robust governance can improve service without indiscriminate inventory growth. Finally, formulas are decision aids, not substitutes for data quality, behavioural understanding and managerial judgment.

Key Terms

Inventory: Stock of materials, components, work-in-process or finished goods held for future use or sale.

Cycle stock: Inventory expected to cover demand between replenishments.

Safety stock: Additional inventory held to protect against uncertainty.

Anticipation stock: Inventory built before predictable seasonal or event-driven demand.

Pipeline inventory: Inventory currently in transit or moving through a production process.

Backorder: Customer demand that remains open and is expected to be fulfilled later.

Lost sale: Demand that is not fulfilled because the customer cancels, substitutes or leaves.

Service level: A defined measure of product availability or fulfilment performance.

Cycle service level: Probability of no stockout during a specified replenishment cycle.

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

Lead time: Time between initiating replenishment and receiving usable supply.

Protection period: Time over which demand must be covered by available inventory.

Reorder point: Inventory position at which replenishment is initiated.

Safety factor: Standard-normal quantile associated with a target service probability.

Inventory position: On-hand inventory plus relevant on-order supply minus outstanding demand or commitments.

Working capital: Funds committed to operating assets net of operating liabilities.

Postponement: Delaying final configuration or differentiation until demand is clearer.

Risk pooling: Aggregating demand or inventory locations to reduce relative uncertainty.

Inventory carrying rate: Annual percentage cost of holding inventory value.

Shortage allocation: Rule for deciding which demands receive scarce supply.

Review Questions

Short Answer Questions

1. Distinguish safety stock from cycle stock.
2. Why are cycle service level and fill rate not equivalent?
3. What is the difference between a backorder and a lost sale?
4. State the assumptions behind σ_L = σ_d√L.
5. Identify three components of inventory carrying cost.

Long Answer Questions

1. Explain how inventory, backlog and lost sales should be represented in an aggregate production plan.
2. Discuss the effects of service-level targets on working capital and customer experience.
3. Derive and interpret the fixed-lead-time safety-stock and reorder-point formulas.
4. Compare cost-minimizing and service-constrained aggregate planning approaches.
5. Evaluate the use of postponement and risk pooling as alternatives to increasing safety stock.

Analytical / Application Questions

1. Weekly demand has mean 800 units and standard deviation 150 units. Lead time is five weeks, demand is independent, and the target CSL is 95% with z = 1.645. Calculate lead-time demand, safety stock and reorder point.
2. A firm has 4,000 average units of inventory at a unit cost of $75 and a 20% annual carrying rate. Calculate annual inventory carrying cost and explain two limitations of this estimate.
3. Design a service and shortage policy for a firm serving emergency-maintenance customers and routine consumers. Specify service metrics, allocation priorities, backlog rules and inventory implications.

Multiple Choice Questions with Answers

Question 1

Which measure is the probability that no stockout occurs during a specified replenishment cycle?

A. Fill rate

B. Cycle service level

C. Inventory turnover

D. Order penetration rate

Correct Answer: B. Cycle service level

Question 2

Average weekly demand is 500 units, weekly standard deviation is 120 units, and fixed lead time is four weeks. What is the lead-time demand standard deviation under independent demand?

A. 120 units

B. 240 units

C. 480 units

D. 2,000 units

Correct Answer: B. 240 units

Question 3

In the standard fixed-lead-time model, safety stock primarily protects against:

A. Average demand

B. Production cost

C. Demand uncertainty during the protection period

D. Accounts payable

Correct Answer: C. Demand uncertainty during the protection period

Question 4

A backorder differs from a lost sale because a backorder:

A. Is never recorded

B. Remains an outstanding demand expected to be fulfilled

C. Always has zero customer cost

D. Reduces future demand automatically

Correct Answer: B. Remains an outstanding demand expected to be fulfilled

Question 5

If average inventory value is $400,000 and the annual carrying rate is 22%, what is the estimated annual carrying cost?

A. $8,800

B. $88,000

C. $182,000

D. $400,220

Correct Answer: B. $88,000

Question 6

Which statement about fill rate is correct?

A. It measures the probability of no stockout in a cycle

B. It measures the proportion of demand supplied immediately

C. It is identical to the reorder point

D. It excludes units short from total demand

Correct Answer: B. It measures the proportion of demand supplied immediately

Question 7

For independent demand with fixed lead time L, which expression gives the standard deviation of lead-time demand?

A. σ_d/L

B. Lσ_d

C. σ_d√L

D. σ_d²L

Correct Answer: C. σ_d√L

Question 8

Which practice delays final product differentiation until demand information is clearer?

A. Postponement

B. Obsolescence

C. Backlog ageing

D. Forward buying

Correct Answer: A. Postponement

Question 9

Why can a high cycle service level coexist with a relatively low fill rate?

A. Because cycle service level measures shortage magnitude

B. Because infrequent stockouts can involve large quantities

C. Because fill rate ignores demand

D. Because safety stock is always zero

Correct Answer: B. Because infrequent stockouts can involve large quantities

Question 10

Which is a key limitation of applying normal-demand safety-stock formulas without review?

A. They require no demand data

B. They are exact for all intermittent and correlated demand

C. They may understate risk when demand or lead time is non-normal or correlated

D. They eliminate the need for service definitions

Correct Answer: C. They may understate risk when demand or lead time is non-normal or correlated