Spare Parts, Obsolescence and Multi-Echelon Inventory

Introduction

Spare parts inventory supports the maintenance, reliability and continuity of operations. Unlike ordinary production materials, spare parts may have highly intermittent demand, long replenishment times, uncertain failure rates, high technical specificity and severe consequences when unavailable. A low-cost bearing may stop an entire production line, while an expensive electronic module may remain unused for years and then become obsolete.

This chapter examines how operations managers can control slow-moving, intermittent-demand, repairable and critical spares. It also considers lifecycle inventory, excess and obsolete stock, postponement, pooling and the positioning of inventory across multiple locations. The central managerial challenge is to balance service risk against the financial and operational cost of holding stock.

Learning Objectives

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

1. Distinguish spare-parts categories according to demand, criticality, value and repairability.
2. Select suitable inventory policies for intermittent and slow-moving demand.
3. Explain lifecycle planning, obsolescence and excess-stock governance.
4. Evaluate repair, replacement, pooling and postponement alternatives.
5. Analyse inventory positioning in a multi-echelon network.
6. Apply service, cost and risk measures to spare-parts decisions.

1. The Distinctive Nature of Spare Parts

Spare parts are purchased or held to restore equipment to an operating condition. Their demand is usually dependent on equipment failures rather than directly on customer orders. This creates several important differences from conventional inventory:

• Intermittency: many periods may have zero demand, followed by an unpredictable demand event.
• Criticality: the operational consequence of a stockout may range from inconvenience to safety failure or major production loss.
• Long and uncertain replenishment: specialised parts may require weeks or months to manufacture.
• Technical specificity: a part may fit only one machine generation or software configuration.
• Repairability: some items can be refurbished, while others are disposable.
• Lifecycle exposure: demand often rises during installation, stabilises during maturity and declines when equipment is retired.
• High variety: thousands of low-demand items may support a relatively small number of asset types.

A useful classification combines four dimensions rather than relying on annual consumption value alone.

Criticality should be assessed through consequence analysis. Relevant consequences include lost output, safety exposure, environmental damage, contractual penalties, emergency procurement and reputational harm. A low-value item can therefore justify a high service target.

2. Demand and Stocking Policies for Intermittent Items

2.1 Measuring intermittent demand

For a time series with demand observations y_t over periods t = 1,2,...,T, the demand occurrence rate is:

where n⁺ is the number of periods with positive demand. The average demand per period, including zero-demand periods, is:

The average interval between positive-demand periods is approximately:

These measures describe frequency but not variability. A part with one unit demanded every three months differs materially from a part with twelve units demanded in one month and none thereafter, even if their average demand is similar.

2.2 Forecasting approaches

Simple moving averages often perform poorly when most observations are zero. Suitable approaches include:

• Croston-type methods, which separately estimate demand size and the interval between demands.
• Syntetos–Boylan adjustments, which reduce the bias associated with basic Croston forecasting.
• Intermittent-demand exponential methods, including variants that update demand occurrence and size separately.
• Failure-based models, when equipment population, age, operating hours and reliability data are available.
• Judgemental and engineering forecasts, especially for new assets, design changes and end-of-life events.

No forecasting method should be selected solely because it is mathematically sophisticated. The method must be validated using relevant measures, such as stockout frequency, service level, forecast bias and inventory cost. Forecast error metrics based only on percentage error can be misleading when actual demand is zero.

2.3 Reorder-point logic

For a conventional continuously reviewed item, a basic reorder point is:

where μ_L is expected demand during lead time and SS is safety stock. If demand and lead time are independent and approximately normally distributed, a common approximation is:

where μ_d is mean demand per period, σ_d is its standard deviation, L is mean lead time in periods, σ_L is lead-time standard deviation, and z is the standard-normal factor for the selected cycle-service probability. The expression assumes that demand and lead-time variation can be represented adequately by these moments and that independence is reasonable. If lead time is constant, it reduces to SS = zσ_d√L.

For intermittent spares, normality may be inappropriate, particularly when expected demand during lead time is less than one unit. A discrete model or simulation is preferable. If demand during lead time has probability distribution P(D_L = k), the reorder point for a cycle-service target α is the smallest integer r satisfying:

This formulation is more defensible for zero-inflated demand and whole-unit decisions.

2.4 Policy selection

Common policies include:

• Min–max policy: review stock periodically and order enough to restore the inventory position to a maximum level.
• Periodic review order-up-to policy: order at review time to reach a target covering review-period and lead-time exposure.
• One-for-one policy: often appropriate for critical repairables, where each withdrawal triggers replenishment or repair.
• Base-stock policy: hold a target number of units in the system, including units in repair or transit.
• No-stock or buy-on-demand policy: suitable when the consequence of waiting is low and supply is reliable.

The inventory position includes on-hand stock, on-order stock and, where relevant, recoverable items in the repair pipeline, less backorders. Policy design must specify whether defective items are immediately returned, quarantined, repaired or scrapped.

3. Criticality, Service and Risk

Inventory service is not a single concept. Cycle service level is the probability of no stockout during a replenishment cycle. Fill rate is the proportion of demand units supplied immediately. For a one-unit critical spare, cycle service may be more meaningful than fill rate because a single shortage can cause complete equipment downtime.

A decision should consider expected total cost:

where C_h is holding cost, C_o is ordering or procurement cost, C_p is repair, purchasing or production cost, and C_f is the expected failure or shortage cost. Failure cost may include downtime, emergency freight, overtime, safety consequences and contractual losses. These terms are not always independent; for example, higher inventory may reduce emergency procurement and downtime.

For high-criticality items, a purely statistical target may be inadequate. A risk-based process should identify failure modes, estimate consequence severity and determine whether redundancy, preventive replacement, local stock, supplier agreements or engineering redesign is the best control. Safety-related spares require compliance with applicable legislation, certification and change-control requirements.

4. Repairable and Recoverable Spares

A repairable system has a replenishment cycle that includes failure, removal, diagnosis, repair, testing and return to service. The manager must control both the physical stock and the repair pipeline.

If the average failure rate is λ units per period and the average repair turnaround time is τ periods, the expected number of units in repair is approximately:

This is a special application of Little’s Law, [[MATH:WIP = throughput × flow

time]], and assumes a stable system over the observation period. If demand is highly variable or repair capacity is constrained, a safety allowance is needed and simulation may be more suitable.

The effective availability of a repairable fleet depends on ready-for-use units, units in transit, units awaiting diagnosis, units under repair and units awaiting test or certification. A repairable policy should therefore define:

1. Return-material authorisation and identification.
2. Inspection and failure coding.
3. Repair-versus-replace criteria.
4. Approved repair procedures and test standards.
5. Turnaround-time targets.
6. Warranty and supplier recovery.
7. Scrapping and environmental disposal rules.

Repair decisions should compare the total cost and risk of repair with replacement. Relevant factors include repair cost, expected additional life, turnaround time, obsolescence risk, warranty, reliability after repair and availability of qualified technicians.

5. Lifecycle Inventory and Obsolescence

5.1 Lifecycle phases

Spare-parts planning should begin during asset design and procurement. A typical lifecycle contains:

1. Introduction: uncertainty is high; initial provisioning and engineering validation are important.
2. Growth and maturity: failure data improves; stocking parameters can be recalibrated.
3. Late life: supplier support, component availability and demand decline become central.
4. Withdrawal: final buys, substitutions, cannibalisation and controlled disposal are considered.

Lifecycle planning should link the asset bill of materials, approved substitutes, supplier support period, installed population, failure rates, maintenance strategy and planned retirement date.

5.2 Obsolescence

Obsolescence occurs when stock can no longer be economically, technically, legally or operationally used. Causes include product redesign, discontinued components, software incompatibility, regulatory change, supplier exit, changed standards and asset retirement.

The risk of obsolescence is not identical to slow movement. A slow-moving part may still be essential, while a fast-moving part may become obsolete after a design change. Governance should distinguish:

• Excess: quantity exceeds expected future requirement.
• Surplus: stock is not required at the current site but may be useful elsewhere.
• Obsolete: no credible future use exists under approved operations.
• Quarantined: usability is uncertain pending inspection, engineering or compliance review.

A practical review process includes demand and asset-population analysis, engineering confirmation, site-to-site search, supplier return negotiation, substitution analysis and documented disposal approval. Write-downs or provisions should follow the organisation’s accounting policy; operational classification and financial valuation should not be confused.

5.3 Economic treatment of end-of-life decisions

A final-buy decision should compare the expected cost of buying now with the cost and risk of future shortage. A simplified expected-cost comparison is:

This is not a universal valuation formula; it is a decision aid. P(shortage) is the estimated probability of future shortage, C_shortage is the resulting consequence, C_emergency is emergency procurement cost and C_substitute is the cost or risk of an approved substitute. Holding, capital and disposal costs must also be included in a complete analysis.

6. Postponement, Standardisation and Pooling

Postponement delays final configuration, customisation or commitment until demand information is clearer. In spare parts, examples include holding a common platform module and installing the required connector, firmware or accessory only when needed. Postponement reduces variety and obsolescence, but it requires technical modularity, qualified processes and adequate response time.

Standardisation replaces several similar parts with an approved common part. It may reduce stockholding and improve purchasing leverage, but compatibility, performance, certification and lifecycle effects must be verified before substitution.

Pooling aggregates demand across locations so that uncertainty is shared. If two sites each hold a separate safety stock, a shared stock may provide comparable protection with less total inventory, especially when demand is not perfectly correlated. However, pooling creates transportation time, allocation rules, ownership issues and the possibility that simultaneous demands exhaust the common stock.

Pooling is strongest when:

• demand is geographically dispersed and weakly correlated;
• transport is fast and reliable;
• the part is expensive or slow moving;
• technical interchangeability is high; and
• the shortage consequence is manageable during transfer.

Local stocking remains justified for safety-critical items, long-distance sites, poor transport links or equipment with no operational substitute.

7. Multi-Echelon Inventory Positioning

A multi-echelon network contains several inventory stages, such as a central distribution centre, regional depots, service vehicles and customer or plant locations. Decisions concern both how much inventory to hold and where to hold it.

A centralised system may reduce aggregate safety stock through risk pooling, while a decentralised system may reduce response time. The correct design balances inventory, transportation, replenishment lead time, service risk and operating complexity.

Important concepts include:

• Echelon stock: inventory at a stage plus all downstream inventory controlled by that stage.
• Echelon lead time: time required to replenish an echelon from its upstream source.
• Local lead time: time from the immediate supplying location to the using site.
• Installation base: the population of assets that can generate failures.
• lateral transshipment: movement between locations at the same echelon.

A practical positioning process is:

1. Map asset locations, suppliers, repair centres and transport links.
2. Segment parts by criticality, value, demand and interchangeability.
3. Estimate demand and lead-time distributions at each location.
4. Identify allowable response times.
5. Compare central, regional, local and pooled alternatives.
6. Model emergency transfers, repair loops and simultaneous failures.
7. Pilot the policy and monitor service and cost.

For a network with independent local demands, centralisation often reduces aggregate variability. Yet independence is a strong assumption. Weather events, common equipment design, epidemics, grid failures and supplier defects can create correlated demand. Models should therefore include common-cause scenarios rather than relying only on average historical demand.

Figure: A multi-echelon spare-parts network in which central pooling supports regional and local service points.

8. Technology, Data and Artificial Intelligence

An effective spare-parts system requires an accurate item master, equipment bill of materials, serial-number history, installed-base records, failure codes and repair status. Digital twins and connected-condition monitoring can improve failure prediction, but they do not eliminate uncertainty. Predictive maintenance may shift demand from unplanned failures toward planned replacement; it may also expose new data-quality and cybersecurity risks.

AI and machine learning can assist with failure prediction, anomaly detection, substitution search and prioritisation of scarce stock. Governance should require explainable decision rules for safety-critical recommendations, human approval for engineering substitutions, protection of maintenance data and monitoring for model drift. An algorithm trained on incomplete failure reporting may systematically understate risk.

9. Performance Measurement and Governance

Useful measures include:

Governance should assign ownership for technical approval, inventory policy, accounting treatment, supplier contracts, disposal and emergency overrides. Segregation of duties is important where scrapping, write-offs or high-value purchases are involved. Environmental obligations include responsible disposal of batteries, electronics, oils and hazardous materials.

10. Illustrative Case: Regional Pooling of Wind-Turbine Inverter Modules

Background

The following is an illustrative case. A renewable-energy operator maintains 120 turbines across three regions. Each turbine uses an interchangeable inverter module. The module is expensive, demand is intermittent and an unavailable module can stop a turbine. Failed modules can be repaired at a specialist centre.

Problem

Each regional depot held two serviceable modules. This produced six units of local stock, but one region experienced repeated shortages while another held unused inventory. Repair turnaround was variable, and the inventory system did not count units in repair as part of the replenishment pipeline.

Analysis

The operator classified the module as high value and high criticality. It mapped installed turbines, historical failures, transport times and repair status. Analysis showed that regional demand was not perfectly correlated and that emergency transfers between regions could usually occur within the required response window. However, severe storms could create correlated failures, so a regional-only historical average was unsafe.

Approach and solution

The operator created a pooled central reserve, retained one local module at each region, and introduced a one-for-one replenishment rule. Failed modules were shipped within 24 hours to the repair centre. The central inventory record included ready units, units in transit, units awaiting diagnosis and units under repair. A storm scenario was added to the annual capacity review, and an emergency allocation rule prioritised sites by safety, contractual and production consequences.

Outcome

In the first operating year, the operator reduced total serviceable stock by one unit without reducing the measured response target. Emergency transfers increased modestly, but emergency purchases declined. The principal improvement came not from a forecasting algorithm but from visibility of the repair pipeline, technical interchangeability and explicit prioritisation rules.

Learning points

1. Pooling is valuable only when transfer time is compatible with the service requirement.
2. Repair pipeline visibility is part of inventory control.
3. Historical averages should be supplemented by common-cause scenarios.
4. Criticality and consequence must determine policy, not unit price alone.

Challenges and Limitations

Spare-parts decisions remain difficult because failure data may be sparse, equipment populations may change and engineering knowledge may be held outside the inventory function. Models can produce false precision when parameters are uncertain. Centralisation can reduce stock while increasing transport dependence. Repair can be cheaper than replacement but may lower reliability if quality control is weak. Finally, optimisation based only on financial carrying cost can create unacceptable operational or safety exposure.

Best practice is therefore a controlled combination of statistical analysis, reliability engineering, maintenance expertise, supplier collaboration, scenario testing and periodic managerial review.

Chapter Summary

Spare-parts inventory differs from ordinary materials because demand is often intermittent, failure-driven and highly consequential. Effective management begins with classification by demand pattern, criticality, value and recoverability. Forecasting should match the data, and discrete or simulation-based methods are often preferable to normal approximations for very slow-moving items.

Repairable spares require control of the entire repair loop, not merely the warehouse balance. Lifecycle planning links installed assets, failure rates, supplier support and retirement decisions. Excess, surplus, obsolete and quarantined stock must be distinguished and governed differently. Postponement, standardisation and pooling can reduce variety and uncertainty, but only when technical compatibility and response times are adequate.

Multi-echelon inventory positioning balances local responsiveness against central risk pooling. Reliable data, scenario analysis, ethical governance, environmental controls and human oversight are essential, particularly when AI and predictive maintenance are introduced.

Key Terms

Spare part: An item held to restore equipment to operating condition.

Intermittent demand: Demand separated by periods with zero demand.

Criticality: The consequence of a part or equipment failure.

Repairable item: An item that can be restored for further use.

Installed base: The population of assets capable of generating spare-parts demand.

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

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

Reorder point: Inventory position at which replenishment is triggered.

Safety stock: Protection against demand or lead-time uncertainty.

Croston-type method: Forecasting approach that separately estimates demand size and occurrence interval.

Obsolescence: Loss of future usability or economic value.

Excess stock: Quantity above the expected future requirement.

Surplus stock: Stock unnecessary at one location but potentially useful elsewhere.

Postponement: Delaying final configuration or commitment until information improves.

Pooling: Combining demand or inventory across locations.

Echelon: A stage in a multi-level supply network.

Lateral transshipment: Transfer between locations at the same echelon.

Repair turnaround time: Time from receipt of a failed item to its return to service.

Common-cause demand: Demand generated by a shared event affecting multiple assets.

Review Questions

Short Answer Questions

1. Why can a low-value spare require a higher service target than a high-value noncritical item?
2. Distinguish excess, surplus and obsolete stock.
3. What information should be included in a repairable-item inventory record?
4. Explain the difference between cycle service level and fill rate.
5. State two conditions under which pooling is likely to be beneficial.

Long Answer Questions

1. Explain how lifecycle planning can reduce both shortage risk and obsolescence.
2. Compare min–max, one-for-one and no-stock policies for spare parts.
3. Discuss the advantages and limitations of centralising spare-parts inventory.
4. Explain why intermittent-demand forecasting requires methods different from ordinary high-volume demand forecasting.
5. Develop a governance framework for excess, obsolete and quarantined stock.

Analytical / Application Questions

1. A critical module has a mean demand of 0.2 units per month and a replenishment lead time of six months. Explain why a discrete lead-time demand model may be more appropriate than a normal safety-stock calculation, and identify the additional data required.
2. Two sites each hold two interchangeable units. Their emergency transfer time is two days, while the operational tolerance for an unavailable unit is one day. Should all stock be pooled? Justify your answer and propose a revised network design.
3. A repair centre processes an average of four failed units per week, with an average turnaround time of 2.5 weeks. Use Little’s Law to estimate average repair work-in-process and discuss why the result should not be treated as a guaranteed stock level.

Multiple Choice Questions with Answers

Question 1

Which characteristic most clearly distinguishes many spare parts from routine production materials?

A. Their demand is always seasonal.

B. Their demand is often intermittent and failure-driven.

C. They never require forecasting.

D. They are always low in value.

Correct Answer: B. Their demand is often intermittent and failure-driven.

Question 2

If demand occurs in 12 of 24 periods, the demand occurrence rate is:

A. 0.25

B. 0.50

C. 2.00

D. 12.00

Correct Answer: B. 0.50

Question 3

Which classification means that stock may be useful at another location but is not required at the current site?

A. Obsolete

B. Quarantined

C. Surplus

D. Scrapped

Correct Answer: C. Surplus

Question 4

A one-for-one policy is particularly suitable when:

A. Each withdrawal should trigger replenishment or repair.

B. Demand is perfectly smooth and high volume.

C. The item has no technical importance.

D. No replenishment source exists.

Correct Answer: A. Each withdrawal should trigger replenishment or repair.

Question 5

In the relationship WIP = throughput × flow time, WIP represents:

A. Warehouse inspection price

B. Work in process

C. Weighted inventory percentage

D. Waiting inventory provision

Correct Answer: B. Work in process

Question 6

Which condition generally strengthens the case for inventory pooling?

A. Highly correlated demand and slow transfers

B. Weakly correlated demand and reliable transfers

C. Non-interchangeable parts

D. A response time shorter than every transport option

Correct Answer: B. Weakly correlated demand and reliable transfers

Question 7

Postponement primarily seeks to:

A. Increase product variety before demand is known

B. Delay final configuration until information improves

C. Eliminate all repair activity

D. Move every item to local warehouses

Correct Answer: B. Delay final configuration until information improves

Question 8

For a repairable spare, the repair pipeline should include:

A. Only units already on the warehouse shelf

B. Only units purchased during the current year

C. Units awaiting diagnosis, repair, testing and return to service

D. Only units classified as obsolete

Correct Answer: C. Units awaiting diagnosis, repair, testing and return to service

Question 9

If a repair centre processes 4 units per week and average turnaround is 2.5 weeks, Little’s Law estimates average repair WIP of:

A. 1.6 units

B. 2.5 units

C. 6.5 units

D. 10 units

Correct Answer: D. 10 units

Question 10

Why should common-cause failure scenarios be included in multi-echelon planning?

A. They guarantee that demand is normally distributed.

B. They can cause several locations to need stock simultaneously.

C. They eliminate the need for safety stock.

D. They make all parts interchangeable.

Correct Answer: B. They can cause several locations to need stock simultaneously.