Order-Picking and Replenishment Design

Introduction and Context

Order picking is the warehouse activity in which products are retrieved from storage to satisfy customer, production, retail, or service orders. It is commonly the most labour-intensive and operationally sensitive warehouse process. In many facilities, picking accounts for a substantial share of operating cost because it combines travel, search, handling, checking, and exception resolution. A well-designed picking system therefore seeks more than high hourly output: it must deliver the correct stock, in the correct quantity and condition, within the required service window, at an economically acceptable cost.

Replenishment is the supporting process that moves inventory from reserve storage, receiving, or another supply point to the forward-pick location. Picking and replenishment are interdependent. If forward locations are too small, replenishment frequency and congestion increase. If they are too large, storage capacity and capital tied up in stock increase. If replenishment is poorly timed, pickers encounter stockouts even though inventory exists elsewhere in the building.

The design problem is consequently a system problem. Order profiles, product characteristics, storage media, labour availability, information systems, material-handling equipment, packing capacity, dispatch schedules, and customer-service promises must be considered together.

Learning Objectives

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

1. Distinguish discrete, batch, zone, wave, cluster, and goods-to-person picking systems.
2. Select a picking method according to demand, order, product, service, and facility characteristics.
3. Design forward-pick and replenishment policies using demand and capacity information.
4. Estimate travel, labour, capacity, and service implications using appropriate metrics.
5. Coordinate picking with packing, consolidation, dispatch, and exception management.
6. Evaluate automation, warehouse-control systems, and AI-enabled decision support critically.
7. Diagnose trade-offs through an integrated warehouse case.

1. The Order-Picking System as an Operating Design

A picking system has five linked elements:

• Order demand: the number of orders, lines, units, priorities, destinations, and cut-off times.
• Stock presentation: where and how items are positioned for access.
• Picker or machine movement: the route and transport method between locations.
• Verification and handling: scanning, counting, weighing, labelling, and damage control.
• Downstream flow: packing, consolidation, staging, loading, and dispatch.

The unit of work is important. An order is a customer or internal request. An order line is one product reference and quantity within that order. A piece, unit, or case is the physical quantity picked. A facility with many small orders and few lines per order behaves very differently from one with fewer orders containing many lines.

Useful basic measures include:

These measures require consistent definitions. For example, accuracy may be measured at line, unit, order, or shipment level; these are not interchangeable. A single wrong unit can make an order inaccurate even when most of its lines are correct.

1.1 Sources of picking time

Picking time can be divided conceptually into:

• Travel time: movement between the depot, storage locations, and drop-off point.
• Search and identification time: locating the correct slot and product.
• Extraction time: reaching, lifting, opening, or separating stock.
• Confirmation time: scanning, counting, or entering data.
• Waiting and delay: congestion, blocked aisles, equipment unavailability, or system response.

The design objective is not simply to minimize travel. Excessive batching can reduce travel but increase sorting and order-identification errors. High-density storage can reduce space but increase extraction time. Automation can reduce walking while increasing capital cost, software dependence, and maintenance requirements.

2. Principal Picking Strategies

2.1 Discrete, or single-order, picking

In discrete picking, one picker completes one order before beginning another. The method is simple: order identity is always known, and no post-pick sorting is required. It is appropriate when orders are urgent, highly variable, large, or difficult to combine safely.

Its principal disadvantages are repeated travel and low utilisation when orders contain few lines. Discrete picking is often suitable for high-priority e-commerce orders, spare parts, customized orders, or operations with modest volume and high order complexity.

2.2 Batch picking

In batch picking, a picker collects the same or similar items for several orders during one route. The items are subsequently sorted into individual orders. If a route serves b orders and each order would otherwise require similar travel, batching can substantially reduce repeated movement.

Batching is effective when many orders contain common fast-moving items and when sorting capacity is available. It is less effective when orders have little overlap, when quantities are difficult to separate, or when order priorities differ greatly. Batch size should be determined by travel savings, cart capacity, sorter capacity, order cut-off times, and error risk—not by a fixed number alone.

2.3 Zone picking

The warehouse is divided into zones, and each picker works within an assigned zone. An order may pass sequentially through zones, or picked totes may be transferred to a consolidation point. Zone picking reduces walking distances and allows workers to develop product and equipment familiarity.

Two forms are common:

• Progressive zone picking: the order container moves from zone to zone.
• Parallel zone picking: zones pick simultaneously and the outputs are later consolidated.

Zone boundaries should consider demand density, travel distance, ergonomic conditions, equipment type, and workload balance. Poorly balanced zones create queues at the busiest zone even when total labour appears sufficient.

2.4 Wave picking

Wave picking releases work during defined time windows, often linked to carrier departures, route schedules, customer priorities, or replenishment cycles. A wave may combine discrete, batch, or zone methods. Its value lies in synchronisation: work is released in a controlled pattern so that picking, packing, staging, and loading are coordinated.

The weakness of rigid waves is that they can create peaks and idle periods. A wave that is too large overloads downstream processes; a wave that is too small increases administrative and travel overhead. Modern systems often combine planned waves with continuous release for urgent or highly predictable work.

2.5 Cluster picking

Cluster picking uses a cart, rack, or workstation with separate compartments for several orders. The picker selects an item once and places it directly into the correct order compartment. It combines the travel advantage of batching with immediate order separation.

Cluster size is constrained by cart capacity, order volume, product dimensions, visual access, and the likelihood of misplacement. The method is particularly useful for small-item retail and e-commerce orders with many overlapping stock-keeping units (SKUs).

2.6 Goods-to-person picking

In goods-to-person (GTP) systems, an automated storage and retrieval system, shuttle, carousel, vertical lift, or mobile robot presents inventory to a stationary operator. Travel is transferred from the person to equipment. The operator can achieve high productivity and improved ergonomics, but the system requires capital, software integration, maintenance capability, and sufficient demand stability.

GTP is not automatically superior to person-to-goods picking. Its economic and operational suitability depends on throughput, SKU dimensions, inventory policies, presentation rate, peak demand, equipment availability, and the cost of failure. A hybrid facility may use GTP for small, fast-moving items and conventional shelving or pallet picking for bulky or slow-moving items.

3. Selecting a Picking Method

A structured selection process begins with an order-profile analysis. Important variables include:

A useful concentration measure is the cumulative share of demand generated by the most active SKUs. ABC classification may identify fast movers, but velocity alone is not enough. Cube, weight, margin, seasonality, substitutability, hazard, and affinity with other items also matter.

The cost-to-serve perspective is valuable. A low-price item that causes disproportionate travel, replenishment, or packing effort may require a different storage or order policy from a high-value item. Conversely, a high-value item may justify secure storage and additional verification even if its volume is low.

4. Forward-Pick Design

The forward-pick area is the location from which routine orders are picked. Reserve stock supports replenishment. Forward locations should be positioned according to demand, affinity, ergonomic limits, and replenishment practicality.

4.1 Slotting principles

Effective slotting generally follows these principles:

1. Place high-frequency items where travel and access are easiest.
2. Position heavy items between safe lifting limits, usually low or waist height as appropriate.
3. Place items commonly ordered together near one another, while avoiding congestion.
4. Separate similar-looking or easily confused products when error risk is high.
5. Provide adequate presentation quantity and clear location identification.
6. Review slots when demand, packaging, seasonality, or order composition changes.

The shortest geometric route is not always the best route. A fast mover at the end of an aisle may be preferable if moving it closer creates cross-traffic or interferes with a high-volume zone. Slotting is therefore a network and flow decision, not merely a distance-minimization exercise.

4.2 Estimating forward capacity

Let d be average daily demand in units, L be the replenishment lead time in days, and s be the desired safety stock in units. A basic forward-location capacity requirement is:

Here Q_f is required forward capacity in units. This relationship assumes reasonably stable average demand and that replenishment lead time is expressed in the same time unit as demand. It does not account for case-pack constraints, slot dimensions, peak demand, or replenishment restrictions; these must be added in practical design.

If demand and replenishment lead time are uncertain, safety stock may be estimated using a service-level approach. For independent demand during lead time, a commonly used approximation is:

where SS is safety stock in units, z is the standard-normal service factor, and σ_LT is the standard deviation of demand during replenishment lead time. The formula assumes a suitable statistical demand model and a defined cycle-service target. If daily demand has standard deviation σ_d and lead time is a constant L days, then:

If both demand and lead time vary independently, a common approximation is:

where σ_L is the standard deviation of lead time in days. These formulas should not be applied mechanically when demand is intermittent, strongly seasonal, correlated with lead time, or constrained by minimum order quantities.

Worked example: forward stock

An item has average daily demand of 120 units. Replenishment normally takes 0.5 day, and the operation wants 30 units of safety stock. The minimum forward capacity is:

If stock is supplied in cases of 24 units, the physical capacity must be rounded upward to four cases, or 96 units. The slot must also accommodate packaging and practical access, so 96 units is a planning minimum rather than a guaranteed operational optimum.

5. Replenishment Design

Replenishment may be:

• Periodic: performed at scheduled times.
• Continuous or real-time: triggered when a threshold is reached.
• Top-off: forward locations are filled during low-activity periods.
• Emergency: immediate response to a stockout or priority order.
• Pre-wave: completed before a planned picking wave.

A robust design separates routine replenishment from emergency replenishment. Emergency work should be visible, prioritized, and investigated because repeated emergencies indicate a policy, slotting, inventory-record, or forecasting problem.

A simple reorder trigger is:

where ROP is the reorder point in units, d is average demand per time unit, L is replenishment lead time in the same time unit, and SS is safety stock. This is appropriate for a basic continuous-review policy with relatively stable conditions. In a forward pick face, the trigger may be based on the remaining quantity, open picks, and expected demand before the next replenishment opportunity.

Replenishment priorities should consider:

1. confirmed demand that cannot be completed from the forward face;
2. approaching dispatch cut-off times;
3. high-priority or regulated orders;
4. travel efficiency by combining nearby tasks;
5. equipment and aisle safety;
6. reserve-stock availability and inventory accuracy.

Replenishment should not be released merely because a location is empty if the item has no near-term demand. Conversely, waiting until an empty face causes a picker to stop is usually too late. WMS rules should use projected demand and open work where data quality permits.

Replenishment workload

If item i requires r_i replenishment trips during a planning period and each trip consumes t_i minutes, total replenishment labour is:

This is a workload estimate, not a guaranteed staffing requirement. Allowance must be made for travel variability, congestion, equipment exchange, breaks, supervision, and exceptions. The same reserve location may also serve multiple forward faces, so route design and task consolidation can materially change actual time.

6. Coordinating Picking, Packing and Consolidation

Picking does not create service value until the correct shipment is packed and dispatched. The hand-off should therefore be designed as an integrated flow.

6.1 Packing interfaces

Packing capacity must match the arrival pattern from picking. If picking is released in large waves but packing stations process continuously, accumulation occurs. If packing is overstaffed relative to picking, labour is idle. Required packing capacity can be estimated as:

where C_p is the number of packing stations required, N_p is the expected number of parcels or orders in the planning period, t_p is average packing time per parcel or order in hours, H is available hours per station, and u is the target utilisation expressed as a decimal less than 1. The utilisation allowance protects against variability and does not represent productive time.

Packing design should include carton selection, dunnage, weighing, documentation, labelling, quality checks, and destination-specific requirements. A packing station must have sufficient presentation space without allowing work-in-process to obscure order identity.

6.2 Consolidation

An order may be picked in several zones, from multiple temperature areas, or through different automation systems. Consolidation can occur by order, route, customer, carrier, or delivery time. Common controls include tote IDs, barcodes, RFID, sortation, scan-to-light, and system status messages.

A consolidation buffer should be sized for expected flow variability rather than average flow alone. Too little buffer causes blocking and missed cut-offs; too much hides problems and consumes space. The design must also handle short picks, substitutions, damaged items, and late-arriving lines.

7. Performance Measurement and Control

A balanced dashboard should include productivity, quality, flow, cost, safety, and service measures.

Productivity must not be improved by shifting work downstream. For example, a picker who works rapidly but creates mispicks may reduce apparent pick time while increasing packing inspection, returns, and customer-service cost. Balanced scorecards and process-level measures are therefore preferable to isolated individual rankings.

8. Technology, Automation and AI

Barcode scanning remains a practical foundation because it provides item and location verification at relatively low cost. RFID can reduce line-of-sight requirements but involves tag, reader, interference, and process-design considerations. Pick-to-light and voice systems can reduce visual search and improve hands-free operation, but their benefits depend on slot stability and accurate master data.

Warehouse-management systems allocate work, maintain inventory records, release waves, and record status. Warehouse-control systems coordinate conveyors, sorters, shuttles, robots, and sensors. Integration failures can create apparently mysterious shortages that are actually transaction or interface errors.

AI and advanced analytics can support:

• demand forecasting and forward-stock sizing;
• dynamic slotting based on changing order affinity;
• workload prediction and labour planning;
• route and batch formation;
• anomaly detection for mispicks or inventory discrepancies;
• predictive maintenance for automated equipment.

AI does not remove the need for governance. Models require reliable historical data, defined objectives, monitoring for drift, human override rules, cybersecurity controls, and auditability. A model trained on ordinary demand may perform poorly during promotions or disruptions. Decisions affecting worker evaluation should also be explainable and should not create unsafe work rates or discriminatory scheduling.

9. Illustrative Case: Northstar Health Supplies

Background

Northstar Health Supplies is an illustrative regional distributor serving hospitals and pharmacies. It carries 8,000 active SKUs, including small medical consumables, temperature-sensitive products, and bulky cartons. Daily demand is 2,400 orders with an average of 3.8 lines per order. Orders must be dispatched in three carrier windows. The existing system uses discrete picking from shelving.

Problem

Pickers walk long distances, the final dispatch wave is congested, and urgent hospital orders frequently require emergency replenishment. Management proposes either full automation or a larger labour force, but lacks a structured diagnosis.

Analysis

Order data shows that 18% of SKUs generate 72% of order lines. Many orders share the same fast-moving consumables. Temperature-controlled items require separate handling, while bulky products are unsuitable for small-item automation. The problem is therefore heterogeneous rather than a single-method problem.

A time study identifies travel as the largest component of small-item picking time. Inventory records are generally accurate, but forward locations are undersized for three high-volume items. Packing has spare capacity in the first two carrier windows but is overloaded in the last window.

Approach and solution

Northstar introduces a hybrid design:

1. Fast-moving small items move to a cluster-picking zone with scan verification.
2. Medium-volume items use progressive zone picking.
3. Bulky and temperature-controlled products remain in dedicated discrete zones.
4. A pre-wave replenishment run fills high-volume forward faces before each carrier window.
5. A continuous exception queue handles urgent hospital orders.
6. Packing labour is shifted toward the final dispatch window, while wave release is levelled earlier where service rules permit.
7. A GTP pilot is installed for a limited set of stable, high-density small items rather than the entire assortment.

Outcome

After a controlled pilot, travel per small-item order falls, emergency replenishment declines, and final-window packing congestion is reduced. The operation does not eliminate manual picking; instead, it assigns each product and order type to the method that best fits its characteristics. The pilot also reveals that scan compliance and location-master accuracy are prerequisites for automation benefits.

Learning points

The case demonstrates that a warehouse should not select technology before understanding demand structure. Hybrid systems can outperform a single universal method. Replenishment, packing, and dispatch must be designed as one flow. Finally, service improvement requires both physical redesign and disciplined operating controls.

10. Challenges, Ethics and Governance

Picking design affects worker safety, privacy, job quality, and fairness. Excessive rates, poorly designed reaches, repetitive handling, and inadequate breaks create ergonomic risk. Performance systems should distinguish controllable effort from congestion, equipment failure, training time, and difficult product mix.

Data collected from scanners, wearable devices, or voice systems should have a legitimate operational purpose, controlled access, defined retention, and transparent communication. Automated decisions should be reviewed for unintended discrimination. Hazardous, pharmaceutical, food, and temperature-controlled products require compliance with applicable regulations, traceability, segregation, and documented handling procedures.

Other limitations include demand volatility, inaccurate inventory records, supplier unreliability, seasonal peaks, system outages, and automation lock-in. Scenario analysis, manual fallback procedures, preventive maintenance, cyber controls, and regular process audits are essential.

Best-Practice Design Sequence

A practical sequence is:

1. Establish service promises, cut-offs, and safety constraints.
2. Analyse orders, lines, units, cube, weight, affinity, and variability.
3. Segment products and orders rather than assuming one method fits all.
4. Design slotting, forward capacity, reserve access, and replenishment triggers together.
5. Simulate or pilot alternative picking methods.
6. Balance zones, waves, packing stations, consolidation buffers, and dispatch doors.
7. Define data standards, exception processes, and human fallback procedures.
8. Measure total cost, accuracy, safety, and service—not productivity alone.
9. Review the design after promotions, assortment changes, automation changes, and demand shifts.

Chapter Summary

Order picking is a central warehouse design problem involving travel, access, verification, replenishment, packing, consolidation, and dispatch. Discrete picking provides simplicity and order control; batch picking reduces repeated travel; zone picking specialises labour; wave picking synchronises work; cluster picking combines travel reduction with immediate separation; and goods-to-person systems transfer movement from people to equipment. No method is universally best.

Effective design begins with order-profile analysis and continues through slotting, forward-capacity sizing, replenishment policy, downstream balancing, and performance control. Basic relationships such as forward capacity, reorder point, safety stock, and packing-station capacity are useful only when their assumptions and units are understood. Hybrid systems are often appropriate because product and order characteristics vary within the same facility. Technology and AI can improve visibility and decision quality, but they depend on accurate data, resilient processes, human oversight, ethical governance, and safe work design.

Key Terms

Order picking: Retrieving products to satisfy an order.

Order line: One product reference and quantity within an order.

Discrete picking: Completing one order before starting another.

Batch picking: Picking items for multiple orders in one route.

Zone picking: Assigning workers or equipment to defined warehouse areas.

Wave picking: Releasing work in planned time windows.

Cluster picking: Picking directly into separate containers for multiple orders.

Goods-to-person: Presenting inventory to a stationary operator.

Forward-pick area: Location used for routine order picking.

Reserve storage: Stock held to replenish forward locations.

Replenishment: Moving stock into a forward-pick location.

Slotting: Assigning products to storage and picking locations.

Order affinity: The tendency of products to be ordered together.

Pick accuracy: Correctness of picked lines, units, orders, or shipments.

Reorder point: Inventory position that triggers replenishment.

Safety stock: Buffer inventory held against uncertainty.

Consolidation: Combining picked outputs into complete orders or shipments.

Wave: A controlled release of warehouse work.

Warehouse-management system: Software managing inventory, work, and transactions.

Cost to serve: Total operational cost of fulfilling a particular customer or order type.

Review Questions

Short Answer Questions

1. What is the main operational difference between discrete and batch picking?
2. Why can a zone-picking system create a bottleneck even when total labour is sufficient?
3. Define forward-pick stock and reserve stock.
4. State two conditions under which goods-to-person picking may be appropriate.
5. Why should pick accuracy be measured at a clearly defined level?

Long Answer Questions

1. Compare discrete, batch, zone, wave, cluster, and goods-to-person picking in terms of travel, sorting, flexibility, technology, and risk.
2. Explain how forward-pick capacity and replenishment policy should be designed together.
3. Discuss how picking should be coordinated with packing, consolidation, and dispatch.
4. Evaluate the advantages and limitations of using AI for slotting and labour planning.
5. Develop a balanced performance-measurement framework for a high-volume e-commerce warehouse.

Analytical / Application Questions

1. A SKU has average daily demand of 160 units, a replenishment lead time of 0.75 day, and safety stock of 40 units. Calculate the minimum forward capacity before considering case-pack rounding. Explain the assumptions.
2. A warehouse has many small orders with high SKU overlap but limited sorting space. Which picking method or hybrid would you recommend, and what additional data would you require?
3. A facility has excellent pick productivity but frequent late dispatches and packing congestion. Diagnose possible system causes and propose a phased corrective plan.

Multiple Choice Questions with Answers

Question 1

Which picking method completes one customer order before the picker begins another?

A. Discrete picking

B. Batch picking

C. Cluster picking

D. Wave picking

Correct Answer: A. Discrete picking

Question 2

A forward location has average daily demand of 120 units, a replenishment lead time of 0.5 day, and 30 units of safety stock. What minimum capacity is required before case-pack rounding?

A. 60 units

B. 75 units

C. 90 units

D. 150 units

Correct Answer: C. 90 units

Question 3

What is the principal benefit of cluster picking?

A. It eliminates all replenishment activity

B. It combines travel reduction with immediate order separation

C. It requires no order identification

D. It is limited to pallet-sized products

Correct Answer: B. It combines travel reduction with immediate order separation

Question 4

In a progressive zone-picking system, what normally happens to the order container?

A. It moves through successive zones

B. It remains permanently in reserve storage

C. It is sent directly to the supplier

D. It is destroyed after each zone

Correct Answer: A. It moves through successive zones

Question 5

Which formula represents a basic reorder point when demand and lead time are expressed consistently?

A. ROP = d ÷ L − SS

B. ROP = dL + SS

C. ROP = SS ÷ dL

D. ROP = d + L − SS

Correct Answer: B. ROP = dL + SS

Question 6

Why can excessive batch size be harmful?

A. It always increases product damage

B. It may overload sorting and increase identification errors

C. It prevents all travel savings

D. It eliminates the need for packing

Correct Answer: B. It may overload sorting and increase identification errors

Question 7

Which situation most strongly supports goods-to-person picking?

A. Highly unstable bulky products with no repeat demand

B. Stable, dense small-item demand with sufficient throughput

C. A warehouse without reliable inventory records

D. A facility with no maintenance capability

Correct Answer: B. Stable, dense small-item demand with sufficient throughput

Question 8

What does pick accuracy measure?

A. The proportion of correctly picked work according to a defined unit of measurement

B. The distance travelled per picker

C. The number of packing stations installed

D. The quantity of reserve stock received

Correct Answer: A. The proportion of correctly picked work according to a defined unit of measurement

Question 9

If demand and replenishment lead time are both uncertain, what is a major design implication?

A. Safety stock may need to reflect uncertainty in both variables

B. Safety stock should always be zero

C. Forward locations should be eliminated

D. Wave picking becomes impossible

Correct Answer: A. Safety stock may need to reflect uncertainty in both variables

Question 10

Why should a warehouse use a hybrid picking design in many cases?

A. All products have identical handling requirements

B. Different products and order types have different operational characteristics

C. Hybrid systems require no information technology

D. Hybrid systems guarantee zero errors

Correct Answer: B. Different products and order types have different operational characteristics