Int. J. Logistics Systems and Management, Vol. 3, No.

4, 2007 451

Order-picking methods: improving order-picking

efficiency

Goran Dukic* and Cedomir Oluic

Faculty of Mechanical Engineering and
Naval Architecture
Ivana Lucica 1
University of Zagreb
10000 Zagreb, Croatia
E-mail: goran.dukic@fsb.hr
E-mail: cedomir.oluic@fsb.hr
*Corresponding author

Abstract: Order-picking, or the process of retrieving items from storage

locations in response to a specific customer request, is the most laborious
and the most costly activity in a typical warehouse. This is especially true in
the case of the conventional warehouses, with almost 90% of total time
spent on order-picking activities and 55% of all operating costs attributed to
order-picking. As 50% of total order-picking time is spent on travelling,
organisational changes and application of various order-picking methods to
reduce travel distances could lead to significant improvements. In this paper,
three groups of order-picking methods are analysed: routing, storage and
order batching. In order to determine the potential order-picker travel distance
savings using a particular method or a combination of various methods,
an extensive simulation is done. The presented results showed up to 80%
possible reduction of travel distances using the appropriate combination of
order-picking methods.

Keywords: warehousing; order-picking; routing; storage; order batching.

Reference to this paper should be made as follows: Dukic, G. and Oluic, C.

(2007) ‘Order-picking methods: improving order-picking efficiency’, Int. J.
Logistics Systems and Management, Vol. 3, No. 4, pp.451–460.

Biographical notes: Goran Dukic is a Higher Assistant at the Faculty of

Mechanical Engineering and Naval Architecture, University of Zagreb, Croatia,
from where he obtained a BSc in Mechanical Engineering, and an MSc
and a PhD in Industrial Engineering. His teaching areas are engineering
logistics and supply chain management. His research interests focus on material
handling system design, warehouse system design, modelling, simulation and
optimisation of logistics systems, reverse logistics and information technology
for logistics systems and supply chain management. He is a member of
the International Centre for Innovation and Industrial Logistics (ICIILs)
scientific committee.

Cedomir Oluic is a Full Professor at the Faculty of Mechanical Engineering and

Naval Architecture, University of Zagreb, Croatia, from where he obtained a
BSc in Mechanical Engineering, and an MSc and a PhD in Industrial
Engineering. His teaching areas are industrial engineering, operations research,
organisational theory, engineering logistics and supply chain management, with

Copyright © 2007 Inderscience Enterprises Ltd.

452 G. Dukic and C. Oluic

research interests focused on logistics systems and processes design (internal

transport, material handling, warehousing, etc.) and the application of
operational research methods in material flow design.

1 Introduction

It is well-known that logistic costs have an important influence on the final success of
any company. In Western countries, these costs are estimated at about 10%–15% of GDP.
Besides transportation, warehousing is one of the largest cost drivers in supply chains.
The order-picking process, defined as the process of retrieving items from storage
locations in response to a specific customer request, is the most laborious and the
most costly activity in a typical warehouse, with up to 55% of warehouse total operating
costs (Tompkins et al., 1996). Therefore, it is very important to put some efforts on
reducing order-picking costs, i.e., to improve order-picking efficiency. One way to
improve the order-picking process is to redesign it, using new equipment, layout or/and
automatisation and informatisation of a process. However, these require usually
large investments. Hopefully, there are some other methods aimed at improving the
order-picking process, which are not so costly.
The fact that about 50% of total order-picking time is spent on travelling (Tompkins
et al., 1996) gives a potential to improve order-picking efficiency by reducing travelling
distances. There are several methods that could be used for that purpose. Routing
methods determine the sequences and routes of travelling, trying to minimise total travel
distances. Storage methods or assigning items to storage locations based on some rules,
could also reduce travel distances compared to random assignment. Order batching
methods or grouping two or more customer orders in one picking order, are also very
efficient in reducing total travel distances. All methods mentioned are well-known and
proven in improving order-picking efficiency. However, the performances depend greatly
on the layout and size of the warehouse, the size and characteristics of orders and the
order-picker capacity. The performance of a particular method also depends on the other
methods used, therefore it is important to understand their mutual interactions. The
purpose of this paper is to identify the performances of routing, storage and order
batching methods in combinations, depending on the given situation (layout, orders and
order-picker capacity).
The analysis is restricted to conventional warehouses with so-called basic warehouse
layout. These are rectangular warehouses with parallel aisles, a central depot
(pick-up/delivery point), and two possibilities for changing aisles, at the front and at the
rear of the warehouse. The picking aisles (main aisles) are wide enough to allow two-way
travel, but picking can be done from both sides of the aisle without a significant change
in position. The location of a depot (pick-up/delivery point) is at the front corner of the
warehouse. This is consistent with observations of several similar works presented in
literature (Chew and Tang, 1999; Petersen, 1997; Roodbergen and Petersen, 1999).
 Order-picking methods: improving order-picking efficiency 453

2 Routing methods

There are several routing methods (policies) developed and used in practice. Most of
them are heuristic methods, with performances depending greatly on the particular
operating conditions of the system under study due to their definitions. The simplest
routing heuristic is the S-shape policy. When this method is used, the order picker enters
every aisle where an item has to be picked and traverses the entire aisle. Aisles where
nothing has to be picked are skipped. An exception is made for the last aisle visited, in
case the number of aisles to be visited is odd. In that case a return travel is performed in
the last aisle visited. Another very simple routing heuristic method is the Return policy.
The order-picker enters and leaves aisles containing item(s) to be picked from the front
aisle. The midpoint routing policy, also one simple heuristic, looks like a return method
on two halves of a warehouse. Only the first and last aisle visited are traversed entirely.
Similar to the Midpoint routing policy, with the Largest Gap policy all aisles that contain
even one item to be picked are also left at the same side as they were entered (first and
last aisles visited are traversed entirely). The gap represents the separation between any
two adjacent picks, between the first pick in the aisle and the front aisle, or between the
last pick in the aisle and the back aisle. If the largest gap is between two adjacent picks,
the picker performs a return route from both ends of the aisle. Otherwise, a return route
from either the front or the back aisle is used. The largest gap is therefore the portion of
the aisle that the order picker does not traverse. This policy is a slightly more complex
routing heuristic method than the first three mentioned. The resulting route is somehow
similar, but definitely at least equal to or better than the route defined by the Midpoint
policy in all possible situations. Two relatively new routing methods developed are
Composite policy and Combined policy. The composite routing heuristic combines
features of the S-shape and Return heuristics, minimising the travel distance between the
farthest picks in two adjacent aisles for each aisle individually. Combined heuristics is
also a combination of S-shape and Return policies, but a small component of dynamic
programming gives it the possibility to look one aisle ahead. The decision about return or
traversal route in the aisle depends not only on minimised travel in that aisle, but also on
a better starting point for the next aisle. This, in turn, leads to a better overall result than
the Composite heuristic.
All routing policies described above have some restrictions in creating a route. An
optimal algorithm (Ratliff and Rosenthal, 1983), combining the graph theory and
dynamic programming, results in a shortest possible, thus optimal route. Examples of
routes created by the mentioned routing heuristics and an optimal algorithm are given
in Figure 1.
Even with an optimal algorithm developed, the majority of order-picking operations
use heuristic routing methods (Petersen, 1997). The reason for that is that heuristic
methods may provide near optimal solutions and avoid the confusion inherent in
optimal solutions (Hall, 1993). It is true that a specific heuristic method could, in
some situations, result in a near optimal route, but in some other situations it could
perform badly. Therefore, it is important to know in what situations some heuristics are
good or bad. Even more important, which are better than others and how much better in
particular situations.
454 G. Dukic and C. Oluic

Figure 1 Examples of routes by routing heuristics and optimal algorithm

Depot Depot Depot

S-shape Return Midpoint

Depot Depot Depot

Largest Gap Composite/Combined Optimal

3 Storage methods

Storage methods assign items to warehouse storage locations, based on popularity,

demand, size, hazard, etc. In order-picking systems, storage methods are usually based on
the rule of assigning the frequently accessed items to the locations near the depot (Choe
and Sharp, 1992). Volume-based storage policy assigns items to storage locations based
on the expected picking volume. The most effective storage method in reducing travel
distances seems to be the Cube-per-order index (COI)-based storage policy, assigning
items with the lowest ratio of the item’s required storage space to the item’s order
frequency, to the locations nearest to the depot. With the assumption that each type of
item is dedicated to only one location, volume-based storage and COI-based storage are
the same.
There are several different types (patterns) of storage used in practice, as shown in
Figure 2. Items with higher volume (or smaller COI) are stored in darker locations. They
are all proven in reducing the total travel distances in order-picking compared to random
storage assignment, but the performance of a particular storage type greatly depends on
the routing method implemented. The question is, which type of storage suits the best
particular routing method.

Figure 2 The types of volume-based (COI-based) storage

a) diagonal type b) within aisle type c) along front aisle type d) along front and rear
aisle type
 Order-picking methods: improving order-picking efficiency 455

4 Order batching methods

Order batching is grouping two or more customer orders in one picking order; therefore
the items from several customer orders are picked in one route (as opposed to a single
order-picking, where each order is picked in one route). That generally reduces the travel
distances per order.
There are several order batching methods (algorithms) developed and used in
practice, which could be divided into three main groups: simple, seed and savings
algorithms. First-Come First-Serve (FCFS) is the most obvious of the simple order
batching algorithms. This algorithm adds orders to a group in the sequence they arrive. If
the picker is full (capacity reached), a new group is started. Seed algorithms consist of
two steps. First, the initial order is selected based on some seed selection rule. Second,
the remaining orders are added to a group based on some seed order addition rule, up to
the picker’s capacity. Savings algorithms are based on travel savings that can be obtained
by combining two particular orders in one route as compared to the situation where both
orders are collected individually. For an overview of many different seed and savings
algorithms, readers are referred to De Koster et al. (1999).
Some seed selection rules (for example ‘select the order with the longest travel
time/distance’), some seed order addition rules (for example ‘select that order with the
property that the sum of the distances between every item of the seed and the closest item
in the order is minimised’ with distance measured by travel time/distance), and especially
savings in savings algorithms are based on the calculated distances. This implies that
applied routing and storage methods may have some influence on order batching
performances, opening questions like: which order batching algorithms is the best to use
in a given situation according to the routing method and storage type chosen.

5 The analysis of the routing, storage and order batching methods

The analysis of order-picking methods and their interactions is done by extensive

simulation. The first part of an analysis was a comparison of the routing methods and
their interactions with storage methods. There were 48 different situations examined
(six pick list sizes, two warehouse sizes and four warehouse layouts). In order to
examine routing methods’ performances and to explain the potential travel savings
using volume-based storage and their influence on routing methods, an analysis of
routing policies with random storage is presented first. The graph shown in Figure 3,
representing the performances of routing methods depending on pick list size for just one
examined layout, illustrates general conclusions. Two routing policies, the Midpoint
heuristic and the Composite heuristic, are excluded from the presentation because they
are very similar but slightly worse than the two other analysed policies, Largest Gap and
Combined heuristic.
The return routing policy was outperformed by all other routing policies in all
simulated situations, while difference increases as pick list size increases (but note that
this policy could be beneficial in some other aspects; for instance, in total space required
as there is no need for a rear aisle). The S-shape routing policy is just a few percent over
optimal in the case of large number of picks, but does not perform well with small pick
list sizes. On the contrary, the Largest Gap policy performs well with small pick sizes
(about 5% over the optimal policy), while it is not so good in the case of large number of
456 G. Dukic and C. Oluic

picks. The Combined policy is, in general, the best heuristic policy. It is slightly
outperformed only by the Largest Gap policy in situations with very small number of
picks in a large warehouse (long aisles). For a small number of picks per tour, the best
heuristic policy for a given situation is only 5%–10% over the optimal policy, depending
on the shape of the warehouse. The difference between optimal and best heuristics in the
case of a large number of picks per tour is negligible.

Figure 3 Comparison of routing methods in interaction with pick list size with random

190
Travel distance (m)

S-shape
Return
150
Largest Gap
Combined
110
Optimal
Number of aisles: 6
70 Length of an aisle: 12 m
0 10 20 30 40 50 Width of front and
rear aisle: 3 m
Pick list size Distance between two
adjacent aisles:4 m

To compare the routing methods with volume-based storage, the best type of storage for a
particular routing method had to be determined. The simulation analysis included four
different volume-based ABC curves, denoted as 50/20, 60/20, 70/20, 80/20 (the first
number indicates the percentage of total activity corresponding to items indicated in
percentage by the second number), in all 48 mentioned situations. A thorough
explanation is given in Dukic and Oluic (2003), with results summarised in Table 1.

Table 1 The best types of storage for particular routing method

Region
Routing method Number of picks ABC curve Preferred type of storage
S-shape All All Within aisle
Return Large Less skewed Along front aisle
Small More skewed Diagonal
Largest Gap All All Within aisle
Combined Larger All Within aisle
Few All Diagonal
Optimal Larger All Within aisle
Few More skewed Diagonal

The simulation showed that all volume-based storage types provide travel savings over
random storage. The large savings (45%–55%) are possible in the case of a small number
of picks and more skewed ABC curves, while for less skewed curves and large pick lists,
the advantage of using volume-based storage is diminished (only few to 15%, depending
on the routing policy). The performances of routing methods with preferred type of
volume-based storage are illustrated for one examined situation in Figure 4.
 Order-picking methods: improving order-picking efficiency 457

Figure 4 Comparison of routing methods in interaction with pick list size with volume-based

120 S-shape
Travel distance (m)

Return
100 Largest Gap
80 Combined
Optimal
60
Number of aisles: 6
Length of an aisle: 12 m
40 Width of front and
0 10 20 30 40 50 rear aisle: 3 m
Distance between two
Pick list size adjacent aisles: 4 m
ABC curve: 80/20

The Return routing policy is no longer inferior to other heuristics in all situations. With a
small number of picks and more skewed curves, it could outperform the S-shape and the
Largest Gap policies. Also, the decision factor between using S-shape or Largest Gap
routing policy is no longer only the average number of picks per (visited) aisle, but also
the skewness of the ABC curve. For an 80/20 ABC a curve, the Largest Gap routing
performs as good as the Combined heuristic for even very large number of picks.
Generally, the Combined policy is still the best routing heuristic. The right selected
simple heuristic with the appropriate type of volume-based storage results in a travel
distances that are only 4%–8% over the optimal route. Combined heuristics is even
better, with only 1%–5% over the optimal travel, depending on the pick list size and the
skewness of the ABC curve.
The analysis of order batching methods was also done by simulation. Because
of a very large number of different algorithms, the analysis included only nine seed
algorithms, basic Clarke and Wright’s and modified Clarke and Wright’s savings
algorithms, Small-Large savings algorithm (Elsayed and Unal, 1989) and FCFS
algorithm as a reference, based on experience and some previous research (De Koster
et al., 1999). The simulation was performed for 16 different combinations of warehouse
size and layout, order picker capacity and order size. Each order batching algorithm
was analysed in combination with four routing policies (S-shape, Largest Gap, Combined
and Optimal). Because of an assumed larger number of picks with order batching, the
Return routing policy was excluded from an analysis regarding their bad performance in
such situation. The storage was random, with later extension on the preferred type of
volume-based storage.
The results showed that among seed algorithms, those with ‘order with the longest
travel distance’ selected as a seed order are generally the best. For heuristic routing
policies, the best addition rule was ‘select the order which, together with a seed order,
results in greatest saving’, while for optimal algorithm it was the rule ‘select the order
with the property that the sum of the distances between every item of the seed and the
closest item in the order is minimised’. The best order batching algorithm was modified
Clarke and Wright savings algorithm, although outperforming the best seed algorithms on
average by only 2,5%. The performances of other savings algorithms are similar to
performances of best seed algorithms. The best order batching algorithm outperformed
simple FCFS algorithm 4%–21% (on average 12%), inversely proportional to the number
458 G. Dukic and C. Oluic

of orders in a group (defined by capacity and order size). That justifies in some situations
the use of advanced order batching algorithms in achieving order-picking efficiency, in
spite of their complexity.
The relative performances of routing methods with order batching stay the same as
for single order-picking with larger orders, as it was expected. The differences between
heuristic routing methods and optimal algorithm are decreased. The simple S-shape
routing method is outperformed by optimal algorithm just by a few percent, therefore the
complex optimal algorithm is not justified in situations with more customer orders
batched in one large group.
The potential savings using order batching in comparison with single order-picking
(picking by order) depend mostly on the number of customer orders per group
(order-picker capacity/average order size). They ranged from cca. 40% to 70% in
conducted simulations, as illustrated for one situation in Figure 5.

Figure 5 Travel distance savings using order batching in comparison with single order-picking

80
S-shape
Decrease of travel distance per

Largest Gap
70
Combined
Optimal
60
order (%)

Number of aisles: 8
Length of an aisle: 36 m
50 Width of front and
rear aisle: 3 m
Distance between two
40 adjacent aisles: 4 m
Order size: 10 picks
Order batching algorithm:
Clarke and Wright modified
30
0 1 2 3 4 5 6 7
Number of orders per group

An additional analysis of order batching algorithms with volume-based storage resulted

in two major observations. First, volume-based storage did not have any influence on the
ranking of order batching algorithms. Second, the potential savings using order batching
methods stay the same as with random storage. The savings are cumulative, therefore the
best combination of order-picking methods is: optimal routing algorithm – within aisle
type of volume-based storage – Clarke and Wright modified order batching algorithm,
with potential savings in travel distances up to 80% compared to a situation with routing
by simple heuristics, random storage and single order-picking.
 Order-picking methods: improving order-picking efficiency 459

6 Conclusion

The analysis of order-picking methods in conventional warehouses showed great

potential in reducing travel distances, therefore justifying their usage in improving
order-picking efficiency. The best combination of methods is determined, while the
relative effects of various methods are different and depend on particular situation.
Order batching on average showed the greatest potential in reducing order-picking
travel distances. This is true particularly when small order sizes are common (larger
number of orders per group). In this paper it is assumed that sorting of items is done
during the picking process (so-called sort-while-pick systems). On the contrary, the need
for additional sorting operations should be also taken into consideration.
The volume-based storage requires significally less travel than random storage does.
Even for the large order sizes (where order batching is not appropriate), volume-based
storage could reduce travel distances for more than 30% with more skewed ABC curves.
In such cases, the within-aisle type is the best type of storage. Note that such method may
also increase picker congestion within aisles containing the most popular items, which
was not taken into consideration in this paper.
Using more complex routing heuristics or optimal algorithm does offer reduction in
picker travel compared to simple heuristics (and especially compared to when not using
any routing method), but this reduction is insignificant for practical purposes while two
other methods (order batching and volume-based storage) are implemented. The
simple heuristics are more consistent in forming routes, while more complex routing
methods could cause confusion, which in turn increase errors and picking time (Petersen
and Aase, 2004).
Therefore, the implication on the managerial decision-making process is that
changing just one or two of three considered decision policies, depending on their unique
situation, could result in significant improvements in order-picking efficiency.

References
Chew, E.P. and Tang, L.C. (1999) ‘Travel time analysis for general item location assignment in a
rectangular warehouse’, European Journal of Operational Research, Vol. 112, pp.582–597.
Choe, K.I. and Sharp, G.P. (1992) ‘Small parts order picking: design and operation’, Technical
Report, Georgia Tech Research Corporation, Atlanta.
De Koster, M.B.M., Van Der Poort, E.S. and Wolters, M. (1999) ‘Efficient orderbatching methods
in warehouses’, International Journal of Production Research, Vol. 37, No. 7, pp.1479–1504.
Dukic, G. and Oluic, C. (2003) ‘Routing and storage policies in order-picking: the best
combinations’, in L. Barros, et al. (Eds.) Proceedings of International Conference on
Industrial Logistics ICIL 2003, Vaasa, Finland, pp.140–149.
Elsayed, E.A. and Unal, I.O. (1989) ‘Order batching algorithms and travel time-estimation for
automated storage/retrieval systems’, International Journal of Production Research, Vol. 27,
pp.1097–1114.
Hall, R.W. (1993) ‘Distance approximations for routing manual pickers in a warehouse’, IIE
Transactions, Vol. 25, No. 4, pp.76–87.
Petersen, C.G. (1997) ‘An evaluation of order picking routing policies’, International Journal of
Operations and Production Management, Vol. 17, No. 11, pp.1098–1111.
460 G. Dukic and C. Oluic

Petersen, C.G. and Aase, G. (2004) ‘A comparison of picking, storage and routing policies
in manual order picking’, International Journal of Production Economics, Vol. 92, No. 1,
pp.11–19.
Ratliff, H.D. and Rosenthal, A.S. (1983) ‘Order-picking in a rectangular warehouse: a solvable case
of the travelling salesman problem’, Operations Research, Vol. 31, No. 3, pp.507–521.
Roodbergen, K.J. and Petersen, C.G. (1999) ‘How to improve order picking efficiency with routing
and storage policies’, in G. Forger, et al. (Eds.) Progress in Material Handling Practice,
Material Handling Institute, Charlotte, North Carolina, pp.107–124.
Tompkins, J.A., White, J.A., Bozer, Y.A., Frazelle, E.H., Tanchoco, J.M.A. and Trevino, J. (1996)
Facilities Planning, 2nd ed., New York: John Wiley & Sons.