Supply chain managment

ARTICLE IN PRESS

0925-5273/$ - se

doi:10.1016/j.ijp

� Correspondi

E-mail addre

(K. van Donsel

Int. J. Production Economics 121 (2009) 620–632

www.elsevier.com/locate/ijpe

Logistics drivers for shelf stacking in grocery retail stores: Potential for efficiency improvement

Susan van Zelst, Karel van Donselaar � , Tom van Woensel,

Rob Broekmeulen, Jan Fransoo

Department of Technology Management, Technische Universiteit Eindhoven, P.O. Box 513, 5600 MB Eindhoven, The Netherlands

Received 17 March 2006; accepted 28 June 2006

Available online 30 August 2006

Abstract

In retail stores, handling of products typically forms the largest share of the operational costs. The handling activities are

mainly driven by the shelf stacking process. While the impact of the handling costs on the profitability of a store is

substantial, there are no models available of the different drivers influencing store handling. In this paper, a study of the

shelf stacking process is presented. First, a conceptual model based on warehouse operations is derived. It is shown that

stacking costs are non-linear with the number of consumer units stacked. Secondly, by means of a motion and time study,

data has been collected for dry groceries in four stores of two different European retail companies. Using regression, the

developed model clearly demonstrates the impact of the most important drivers for stacking efficiency: case pack (CP) size,

number of CPs stacked simultaneously, the filling regime and the working place of the employees. Efficiency gains of

8–49% by changing the driver parameter value are identified. Based on the presented insights both retail companies have

already decided to structurally change their current operations.

r 2006 Elsevier B.V. All rights reserved.

Keywords: Retail operations; Handling; Store; Shelf stacking; Motion and time study; Model

1. Introduction

In times of severe competition, many retailers recognize the importance of controlling costs. With known supply chain costs, the information needed to most effectively structure the supply chain can be provided. Moreover, different opportunities needed to simultaneously reduce costs and increase perfor- mance can be identified. In Fig. 1 (see Broekmeulen

e front matter r 2006 Elsevier B.V. All rights reserved

e.2006.06.010

ng author.

ss: [email protected]

aar).

et al., 2006) the operational logistical costs made in the part of the supply chain that includes the retailer’s warehouse and the store are presented. Since we focus on operational costs, total shelf space and the assortment are assumed to be known. Furthermore, since this cost analysis is based on non-perishables (dry groceries), obsolescence costs are negligible. As a result, the inventory costs in the cost pie in Fig. 1 only consist of the inventory holding costs. It can be seen that the majority of the operational costs are handling costs. In another empirical study by Saghir and Jönson (2001) the same trend was observed: they found that 75% of

.

ARTICLE IN PRESS

Transportation 22%

Handling in warehouse 28%

Inventory in store 7% Inventory in

warehouse 5%

Handling in store 38%

Fig. 1. Operational logistical costs in the retail supply chain for

non-perishables.

S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632 621

the handling time in the retail chain occur in the store.

1 In their paper efficiency improvements

through the integration and development of new packaging systems are described. In this article, the potential to improve store-handling efficiency is discussed by identifying the drivers for shelf stacking (i.e. given the packages and the inventory replenishment rules). Handling costs in the stores in the two retail chains investigated in this paper are equal to around 50 million euro (or 60 million dollar) per year, indicating that efficiency gains can lead to substantial cost savings as well.

Since handling costs are significantly higher than the inventory costs, it is worthwhile to assess the drivers of the handling costs. This shows the need for a model which adequately describes the handling process and its related costs in the store. Today, no complete models are available to estimate handling costs. Consequently, no realistic estimation can be made about the effect of potential improvements in order to reduce handling time and the related costs. Assortment planning and shelf space allocation are important issues in retail, which we would expect to be based on a trade-off between shelf space, inventory costs and handling costs. Yet, even recent studies in this area either ignore the handling costs or simply model them as a linear function of the number of consumer units (CUs) sold (i.e. without intercept). Almost all recent contributions in the literature on retail operations are mainly focussed on the inventory aspect (see e.g. Shah and Avittathur, 2006; Van Donselaar et al., 2006a, b; Hwang et al., 2005; Fleisch and Tellkamp, 2005;

1 The fact that the ratio ‘handling costs in the store vs. handling

costs in the warehouse’ presented in Fig. 1 is below the ratio 75:25

is due to the large difference in salaries: in stores, replenishment is

often done by (young) part-timers.

Ketzenberg et al., 2002; Wee and Dada, 2005; Gaur et al., 2005). So, although it is a major part of the profit equation, little literature is available on handling costs in retail stores. The scarce literature which is available, originates from the 1960s. Moreover, only a few papers focused on minimizing operating costs and reducing both inventories and handling costs (see Chain Store Age, 1963, 1965). SLIM (Store Labor and Inventory Management) was the most widely promoted system in the mid- 1960s for minimizing store handling expense by reducing backroom inventories and the double handling of goods (Chain Store Age, 1965). Today, no models for handling activities in retail stores are available which are tested on empirical data and research in the area is lacking as well.

The main contributions of this paper are twofold. First, while the body of literature that studies the

efficiency of handling in warehouse operations is substantial, the literature on handling related to store operations is scarce. Most of the literature on retail stores focuses primarily on the demand side, and less on the cost side. When there is a focus on cost, most attention is devoted to inventory holding costs. Little research is available on the assessment of handling costs in retail stores. This paper is a new entry into this almost unexplored domain. Starting from handling models used in warehousing literature, a conceptual model for shelf stacking activities in the retail store is derived.

Secondly, the conceptual model is validated using data collected at two retail companies in the grocery sector. Evaluating the conceptual model using regression analysis on the empirical data not only shows there is a relationship between the stacking time and its drivers, but also quantifies the effect of the different important drivers for shelf stacking time in the stores: (1) the number of case packs (CPs) per order line; (2) the CP size (3) the way the shelves are stacked (i.e. filling regime); (4) the worker doing the actual stacking. Using these empirical results, the efficiency gains are quantified. Moreover, based on the obtained results and insights both retail companies have decided to adjust their current operational processes.

The organization of the paper closely follows the methodology presented by Mitroff et al. (1974). First the problem is conceptualized and the main variables to be studied are identified. Then the model is built and analysis is conducted based on the model. The model is then validated using the

ARTICLE IN PRESS

2 Note that the design of a order pick lane in a warehouse is

quite different from an aisle in a store (Broekmeulen, 1998), since

the locations and the storage space allocations of the SKUs

(slots) in an warehouse are optimized for the handling activities,

while the slot allocations in a planogram try to optimize the sales

(Corstjens and Doyle, 1981; Urban, 1998).

S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632622

real life data that were collected in the stores on the actual shelf stacking process. As such this approach fits the concept of model-based empirical research where empirical data are analyzed based on quantitative models, and results can thus be interpreted within a validated modeling context (Bertrand and Fransoo, 2001) In Section 2, the conceptual model for the shelf stacking activities in the store is derived from the warehouse handling models. In Section 3, the research methodology is described in detail. In Section 4, the analysis and the results of the model are discussed using the data collected at two retailers; Section 5 explains the implications for efficiency improvement in the stacking process of a store. Finally, Section 6 concludes and further research options are de- scribed.

2. Conceptual model

The following replenishment process is observed for the items on the shelves in the different stores: after unloading the truck, the store clerks move the deliveries to the shelves, unpack the CPs and put the CUs on the shelves. To promote First In First Out retrieval from the shelves by customers and to improve the display, the CUs on the shelves are sometimes rearranged, putting the oldest inventory in front (depending upon the specific product category). Although this shelf stacking process in the store is similar to the order picking process at a warehouse, literature on store handling operations is very scarce, while literature on handling in warehouses is abundant. To derive a conceptual model for shelf stacking in the stores, we first describe the handling activities in warehouses (Section 2.1), then the stacking activities in stores (Section 2.2) and finally, the derivation of a formula for the stacking time.

2.1. Handling activities in warehouses

Handling activities are explicitly modeled in warehouse models (Rouwenhorst et al., 2000). These models are very useful as they consider each article or stock-keeping unit (SKU) separately and they include handling costs explicitly as part of the objective function. Moreover, different decisions (e.g. routing policies, picking, etc.) and parameters (e.g. productivity of the workers) can be modeled explicitly in these warehouse models. Therefore, the activities needed for stacking SKUs

into the shelves of a grocery store are compared with order picking in a warehouse. As such, shelf stacking is seen as the reverse of order picking, i.e., instead of unloading a container with dif- ferent SKUs in the store, an order picker in the warehouse loads different SKUs in a container for shipment.

2

In this paper, a shipment to be stacked in the store is considered as the equivalent of a customer order which has to be picked in the warehouse. In this analogy, shelf stacking at a store, where the relatively large shipments are divided over several store clerks and these store clerks move the goods to the storage locations on the shelves, resembles zone picking in a warehouse. Zone picking is defined by Frazelle and Apple (1994) as an order picking method where a warehouse is divided into several pick zones, order pickers are assigned to a specific zone and only pick the items in that zone, orders are moved from one zone to the next (usually on conveyor systems) as they are picked (also known as ‘‘pick-and-pass’’). To reflect this resemblance, shelf stacking at the store is referred to as zone stacking in this paper. Zone stacking assumes that the incoming goods are already sorted at the supplier according to the different aisles of the stores. This separation along product characteristics is called family grouping.

An order pick cycle in zone picking is the process of loading a container that is part of a shipment to a customer. According to Tompkins et al. (2003), an order pick cycle consists of the fixed setup activities that are related to the start and end of the cycle, such as getting the instructions and transferring the loaded container to the dock boards, and variable activities related to the number of order lines. An order line is defined as an instruction to pick a requested number of units from a specific SKU in the zone. The following activities depend on the number of order lines: traveling (to, from, and between the storage locations), searching for the location, and reaching and bending to access the location (included in activity ‘‘Other’’ in Fig. 2). The actual pick activities such as extracting items from the location and packing the items for shipment depend on the number of

ARTICLE IN PRESS

50%

20%

15%

10%

5%

0% 20% 40% 60%

Travel

Search

Pick

Setup

Other

A c

ti v

it y

% of order picker's time

Fig. 2. Typical distribution of an order picker’s time.

S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632 623

requested units. Fig. 2, based on Tompkins et al. (2003), shows a typical distribution of an order picker’s time based on a single order picking strategy, where each order picker completes one order at a time.

Note that in zone picking, traveling and searching within a zone are less important than in single order picking, since a zone is relatively compact and the order picker is familiar with the locations in the zone. As a result, traveling and searching in zone stacking in a store may also be less dominant than suggested by Fig. 2.

The time required for accessing the location can still be relevant in a warehouse. However, in a store accessing the location is an important activity, because it includes maintenance of the location such as preparing the shelves and removing old inventory. If one wants to promote First In First Out (FIFO) retrieval by the customers of the store, the items have to be shifted or removed before one can stack the new items behind them. In a warehouse, (gravity) flow racks, which are replen- ished from the back, can easily maintain FIFO retrieval. Normally, a store has no space for these kinds of racks or not all types of products are suited for these racks. A more costly solution is assigning more slots to a SKU, such that one slot is the active picking location and the other slot holds the backup inventory.

2.2. Stacking activities in stores

As mentioned before, (shelf) stacking in the store is considered as the mirror activity of picking in the warehouse. The process of stacking defined in this paper starts with grabbing a full casepack from a rolling container within the store and

ends with the disposal of the waste of the empty casepack.

There are three different ways a shelf can be filled with items in the store. The basic filling regime Unit is putting the individual CUs on the shelf. The filling regime is referred to as Tray if the complete CP can be put directly on the shelf. In the filling regime Loose, the items can be dumped on the shelf without rearranging. This requires normally the availability of a type of crate on the shelf. While the picking time in a warehouse is related to the number of requested units, the filling time in the store is expected to be dependent on the number of CUs (if filling regime is Unit) or the number of CPs (if filling regime is Tray or Loose).

Another important activity in zone stacking is grabbing and unpacking the items. Unpacking is necessary when the store wants to present the individual items or when the CP does not have the same physical dimensions as the storage location. This activity resembles the replenishment operation in a warehouse. During replenishment in a ware- house, the same type of problems are encountered when one wants to put a full pallet in a slot that is less than a pallet or that has still items at the location.

The review by Rouwenhorst et al. (2000) indicates that most of the research in warehousing is related to automated storage systems (AS/RS systems) and little research has been done discussing conventional warehouses (e.g. with manual picking). Since stack- ing in the store is done by store clerks, one needs to take into account their work pace in the model too. Only few researchers have reported on the differ- ence in work pace in a warehouse environment (see e.g., Bartholdi and Eisenstein, 1996; Bartholdi et al., 2001).

2.3. The derivation of a formula for the shelf stacking

time in stores

Most SKUs follow the filling regime Unit, so it is expected that the time needed for filling depends on the number of CUs. Other activities like grab and unpack a CP or travel to and from the shelf location, depend on the number of CPs filled. Finally, preparing the shelves and searching are done only once for each SKU, independent of the number of CPs or CUs.

The insights of the stacking process indicate that the number of CU and the number of CP are expected to have an influence on the total stacking

ARTICLE IN PRESS S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632624

time (TST). The dependent variable is the TST expressed in seconds. The explanatory variables are hypothesized to have the following influence on the TST:

1.

The higher the number of CUs to be filled (CU), the higher the TST will be. An increase in the number of CU stacked increases the time needed to put the individual CUs on the shelf, resulting in a higher TST.

2.

The higher the number of CPs, the higher the TST will be. More CPs imply more time needed for activities like grab and unpack a case pack or travel to and from the shelf location, thus leading to higher TST.

The basic starting equation is then as follows:

TST ¼ aþbCUþwCP:

Rewriting this specification by dividing the TST by the number of CUs filled (CU) and making use of the fact that CU ¼ CP Q, where Q stands for the case pack size, results in the following revised model:

TST

CU ¼

a CU þbþw

CP

CU

) TST

CU ¼ bþa

1

CP Q þw

1

Q .

It is important to be aware of differences in working pace of store clerks when interpreting the data, i.e. not every employee works equally fast. Consequently, n�1 dummies for store clerks are added, denoted as DWi, (i ¼ 1, y, n�1, with n the number of store clerks considered) and DWi ¼ 1 if store clerk i is selected and zero otherwise. It is expected that the stacking regimes Tray and Loose will have a different effect than the filling regime Unit. Consequently, two extra dummies are added for the filling regime Tray (DT) and the filling regime Loose (DL) which leads to the following general model (with e the error term):

TST

CU ¼ bþa

1

CP Q þw

1

Q þdDT þ gDL

þ Xn

i¼1

Zi DWi þ �.

Do note that the resulting model for the TST per CU is non-linear in the number of CUs and in the case pack size. This is in contrast to most literature where it is assumed that handling activi- ties are a constant and linear rate in the number of

CUs. Ketzenberg et al. (2000) and Cachon (2001) describe models to optimize the replenishment decisions in the absence of a backroom and assuming handling costs that are linear with the number of CUs. This may be explained by the type of store they are considering. In general three basic store types can be distinguished: stores which receive crates composed of multiple SKUs, where each SKU is less than a case pack size (like dense retail outlets); stores which receive CPs and stack CUs; and stores which receive and stack CPs (like discounters). This research is based on the second type of stores, resulting in nonlinear shelf stacking cost, effectively focusing on a different type of store than those studied by Ketzenberg et al. (2000) and Cachon (2001).

3. Research methodology

3.1. Experimental design

The research in this paper focuses on the stacking process for which data is collected. The data is collected by means of a motion and time study, which is defined by Barnes (1968) as: ‘‘the systema- tic study of work systems with the purposes of (1) developing the preferred system and method— usually the one with the lowest cost; (2) standardiz- ing this system and method; (3) determining the time required by a qualified and properly trained person working at a normal pace to do a specific task or operation; and (4) assisting in training the worker in the preferred method’’. The two main parts in this definition are motion study (or methods design) and time study (or work measurement). The first part is for finding the preferred method of doing work, that is, the ideal method or the one nearest to it. The second part is for determining the standard time to perform a specific task. Besides determining a certain normal time required for a task, time studies are done to detect work method improvements. In such a way, one can analyze a given process to eliminate or reduce ineffective movements, and facilitate and speed up effective movements. Through a motion study, the work is performed more easily and the rate of output is increased.

In the experiment, for each SKU the time needed by a store clerk for stacking the items on the shelf is measured for a delivery (i.e. an order line in the store). In the zone stacking process, each order line consists of taking a case pack from a container,

ARTICLE IN PRESS S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632 625

unpacking the case pack and putting the CUs on the shelf at the assigned location. As timing an entire operation as one element is seldom satis- factory, the TST for an order line has been separated into different sub-activities. The division should be such that the elements are as short as can be accurately timed and that constant elements can be separated from the variable elements (Barnes, 1968). The TST is divided into the following sub- activities:

�

grab and unpack the case pack;

�

search for the assigned location on the shelf;

�

travel to the shelf;

�

check the shelf life of the inventory on the shelf;

�

prepare the location on shelf for stacking;

�

put the new inventory on the shelf;

�

put the old inventory back on the shelf; and

�

44%

21%

11%

10%

8%

3%

2%

Stack new inventory

Grab and unpack case pack

Dispose waste

Travel

Prepare the shelf

Search

Stack old inventory

A c

ti v

it y

dispose the waste.

One more sub-activity has been distinguished, which is not part of TST, called ‘other activities’ to which all time was registered spent by the store clerks for activities which were not directly related to a specific order-line (like helping a customer). Since Saghir and Jönson (2001) mention the lack of standards and definitions on the handling (sub-)activities in a store, Appendix A contains the definitions which have been used in this research project.

3.2. Data collection

Empirical data on the stacking process in two grocery retail companies is collected. In four stores (two of each retail company) employees were followed while stacking the shelves. During the data collection period, the stores were not allowed to change their current operations and were asked to let the most qualified and properly trained qualified personnel do the shelf stacking. Moreover, the days were carefully selected such that the period of measurement did not include any periods of expected demand peaks/drops (e.g., no holidays). The data were gathered for product groups, which meet the following criteria (criteria 1–4 are chosen to enable the investigation of the potential impact of the drivers which are included in the TST-equation in Section 2.3):

0% 20% 40% 60%

% of stacking process' time

1.

Fig. 3. Distribution of the time of the stacking process.

The product groups should contain both fast- and slow movers;

2.

The product groups preferably should contain SKUs from all three filling regimes (Tray, Loose, and Unit);

3.

The product groups should contain different case pack sizes;

4.

The product groups should contain SKUs for which sufficient shelf space is available to accommodate more than one case pack in a delivery (see also Broekmeulen et al., 2006).

5.

All selected product groups should contain items that are comparable in terms of the shelf stacking process and productivity. For this reason, we did not consider product groups such as soft drinks, beers as well as dairy products.

The store clerks are followed during the shelf stacking with a camcorder. Advantages of using a camcorder are that any short cyclical activities can be measured, the stacking process can easily be reviewed and different aggregation levels can be looked at. After the recording process, the TST per order line was registered using a computerized time registration tool, resulting in an extensive database.

4. Analysis and results

Fig. 3 shows the empirical distribution of the TST in the stores. In the zone stacking process at the stores, putting the items on the shelves (‘Stack new inventory’) is the most important activity (for descriptive statistics on the different variables, refer to Appendix B).

When comparing Figs. 2 and 3, a difference between the travel time for order picking in a warehouse and the travel time for zone stacking in a store is observed. The first reason is that the typical

ARTICLE IN PRESS S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632626

distribution of an order picker’s time as given in Fig. 2 is not based on the zone picking strategy but on the single order picking strategy in a warehouse. In the zone picking strategy, travel time is reduced at the expense of increased sorting, which is not included in Fig. 2. The second reason is that the data collection was restricted to the movements within the aisle, which can directly be attributed to the stacking process of an order line. The time needed to bring the container to the right aisle is not part of the travel time in our TST model and has therefore not been measured, i.e., only the time needed for traveling between the container and the shelf is registered.

The general model is analyzed using regression analysis. The effect of the work pace of a store clerk is compared with the median store clerk in the data set, which was store clerk 8. Consequently, 8 dummies DWi for the remaining store clerks are added. The results of the Ordinary Least Squares estimation are shown in Table 1. All relevant collinearity tests (e.g. correlation coefficients, var- iance inflation factors) performed indicated no problems with regards to multicollinearity for the estimated model. The F-statistic indicates that the model is valid.

Almost 40% of the TST per CU for each order line is explained with this model. Table 1 confirms the a priori expectations: the signs of all coefficients are as expected. Looking at the standardized

Table 1

Estimation results TST/CU model

Explanatory

variable

Coefficient t-statistic Standardized

coefficient

Constant 1.758 15.613**

1/(CP Q) 11.126 8.724** .275

1/Q 10.464 8.767** .273

DT �0.454 �4.074** �.080

DL �1.805 �5.262** �.097

DW1 �.562 �4.372** �.094

DW2 �.292 �0.996 �.018

DW3 2.800 11.145** .211

DW4 �.321 �2.129* �.046

DW5 .143 .971 .020

DW6 .328 3.199** .077

DW7 1.252 7.803** .160

DW9 �.091 �.308 �.006

R2a .379

N 1922

*Significant at 5% and; **significant at 1%.

coefficients one can see that most of the explanatory power comes from the variables 1/(CP Q) and 1/Q. The filling regimes Tray and Loose are faster than the filling regime Unit. The filling regime Tray reduces the TST per CU with 0.454 s per CU (1.805 s per CU for the filling regime Loose). The TST per CU is equal to 1.758 s per CU (see constant term in the table). Looking at the different store clerks, Store clerk 1 appears to be the fastest as he stacks on average 0.562 s faster per CU. Although some store clerk dummies are not significant the group of the dummies related to the store clerks is significant as a whole (as confirmed by an F-test; see Gujarati, 1995).

An alternative model specification where product specific characteristics were included to test the effect of product heterogeneity on the TST per CU did not improve the specification. Product hetero- geneity was tested by adding physical volume of an SKU (and its interactions with the other variables). Estimation results indicated that physical volume and the interaction variables were highly insignif- icant indicating that product heterogeneity did not have a proven influence based on this data set. Moreover, alternative specifications (e.g. multipli- cative models and logarithmic functions of the variables) have been tested extensively, but they did not improve the results. Analysis of the other activity revealed that worker 3 was significantly more disturbed by customers than the other workers explaining part of the reduced efficiency of worker 3 (i.e. due to the startup effect after an interruption).

The specification is used to quantify the effect of (1) increasing the case pack size; (2) increasing the number of case packs ordered; (3) changing the filling regime. For example, focusing on the median store clerk 8 (i.e. all dummy variables for the store clerks are thus zero) and looking only at the filling

0

5

10

15

20

25

0 2 4 6 8 10 12 14 16 18 20 22 24 Q

T S

T /C

U . Unit CP=1

Unit CP=2

Fig. 4. Influence of case pack size and number of case packs on

total stacking time per CU [s/CU].

ARTICLE IN PRESS S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632 627

regime Unit (i.e. DT ¼ 0 and DL ¼ 0), the effect of an increase in the number of CPs ordered and the CP size can be evaluated (Fig. 4). It can be seen that the TST in seconds per CU decreases if the CP size increases. CP sizes typically occur around the following three values: 6, 12, and 24 CUs. This analysis advocates using the largest possible CP size for a SKU if sufficient shelf space is available. Alternatively, it helps to recognize the impact on handling efficiency if for any other reason (e.g. reduced risk of obsolescence or perishability), it is decided to decrease the CP size.

Table 2 shows the effect of increasing the CP size from 6CU to 12CU and from 12CU to 24CU for the three different filling regimes and for each of the store clerks (Wi denotes DWi ¼ 1 for store clerk i and DWi ¼ 0 for all others). On average, when stacking in units the time gain is 28%, stacking in trays results in an efficiency gain of 31% and

Table 2

Potential gains on total stacking time per CU [s/CU] when

increasing the case pack size

Unit (%) Tray (%) Loose (%)

6–12CU 12–24CU 6–12CU 12–24CU 6–12CU 12–24CU

W1 37.53 30.03 41.45 35.40 60.19 75.58

W2 35.53 27.55 39.02 32.00 55.20 61.61

W3 22.06 14.15 23.36 15.24 28.33 19.76

W4 35.73 27.80 39.27 32.33 55.70 62.86

W5 32.72 24.31 35.66 27.71 48.70 47.47

W6 31.65 23.15 34.40 26.22 46.38 43.25

W7 27.23 18.71 29.23 20.66 37.46 29.94

W8 33.59 25.29 36.70 28.99 50.66 51.34

W9 34.17 25.95 37.39 29.86 51.99 54.15

Average 32.24 24.11 35.17 27.60 48.29 49.55

Table 3

Potential gains on total stacking time per CU [s/CU] when stacking tw

Unit (%) Tray (%)

Q ¼ 6 Q ¼ 12 Q ¼ 24 Q ¼ 6

W1 19.34 15.48 11.06 21.36

W2 18.31 14.20 9.80 20.11

W3 11.37 7.29 4.25 12.04

W4 18.41 14.33 9.92 20.24

W5 16.86 12.53 8.28 18.38

W6 16.31 11.93 7.76 17.73

W7 14.03 9.64 5.93 15.07

W8 17.31 13.03 8.72 18.91

W9 17.61 13.37 9.03 19.27

Average 16.62 12.42 8.31 18.12

stacking in Loose gives a 49% time reduction when the CP size is increased.

A second observation involves the number of CPs ordered: the more CPs per order line, the higher the time gains, suggesting that more CPs per order line is more efficient. Note that increased casepack sizes or higher number of casepacks per orderline will result in higher inventories resulting in a need for sufficient shelf space. Therefore a trade-off has to be made between the shelf stacking costs and inventory holding costs. In Broekmeulen et al. (2006) it is shown that for a large part of the assortment excess shelf space is available to enable higher inventories in the store without the need to allocate more facings to the items. Moreover, Fig. 1 showed that inventory holding costs are small compared to handling costs for non-perishables in retail stores.

Table 3 shows for each worker the gains that can be achieved when stacking two CPs rather than one CP for the same SKU. As can be seen from the table, significant gains can be realized when products are not ordered with only one CP at the time, but with 2 CPs per order line. Depending upon the fill regime the average gains are 12% (Unit), 14% (Tray) and 26% (Loose). The reason these gains are smaller than the gains from increased casepack sizes is visible in the equation for the total stacking time per CU at the end of paragraph 2: the casepack size Q influences two terms of this equation and CP only influences one of these terms. In other words: if the casepack size Q is changed, not only the number of orderlines change, but also the number of casepacks per year.

The last observation made is that the filling regime has a clear influence on the gains that can be

o case packs instead of one case pack

Loose (%)

Q ¼ 12 Q ¼ 24 Q ¼ 6 Q ¼ 12 Q ¼ 24

18.24 14.12 31.02 38.95 79.77

16.49 12.13 28.45 31.75 41.35

7.85 4.63 14.60 10.18 6.35

16.66 12.31 28.70 32.39 43.60

14.28 9.88 25.10 24.46 23.28

13.51 9.16 23.90 22.29 19.63

10.64 6.71 19.30 15.43 11.01

14.94 10.52 26.11 26.46 27.19

15.39 10.97 26.79 27.91 30.44

14.22 10.05 24.88 25.54 31.40

ARTICLE IN PRESS

Table 4

Potential gains on total stacking time per CU [s/CU] when

changing the filling regime from units to tray and tray to loose

U-T (%) T-L (%)

Q ¼ 6 Q ¼ 12 Q ¼ 24 Q ¼ 6 Q ¼ 12 Q ¼ 24

W1 9.47 15.16 21.66 31.13 53.16 82.30

W2 8.96 13.90 19.19 29.30 48.06 70.67

W3 5.57 7.14 8.32 17.54 22.89 27.00

W4 9.02 14.03 19.43 29.49 48.56 71.76

W5 8.26 12.27 16.21 26.78 41.62 57.57

W6 7.99 11.69 15.21 25.83 39.37 53.37

W7 6.87 9.44 11.61 21.95 31.02 39.10

W8 8.48 12.76 17.08 27.56 43.54 61.31

W9 8.62 13.10 17.69 28.08 44.85 63.95

Average 8.14 12.17 16.27 26.41 41.45 58.56

Table 5

Overview of the efficiency gains achieved

Increase case

pack size

(%)

Increase

number of

case packs

(%)

Filling

regime

Unit 28 12

Tray 31 14 12% (U-T) Loose 49 26 42% (T-L)

S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632628

achieved: The filling regimes Tray and Loose are more favorable. Moreover, store clerks that are faster than the median worker have larger gains when using the filling regime Loose. These results can partially be derived from the above tables; next to this, Table 4 shows the efficiency gains for different CP sizes if the filling regime changes from Unit to Tray (U-T) and from Tray to Loose (T-L). On average, the filling regime Tray is 12% faster than Unit and the filling regime Loose is on average 42% faster.

5. Discussion and managerial implications

Using the specification and the results for the empirical data obtained, important managerial in- sights can be obtained with regards to the effect of (1) increasing the CP size; (2) increasing the number of CPs per order line; (3) changing the filling regime. Table 5 summarizes the main findings from the previous section.

Based on this table, the first step, which contributes the most to an efficiency gain, is to set the filling regime to Tray or Loose for as many SKUs as possible. This strategy needs an additional amount of shelf space needed compared to Unit or Tray which might not be available in stores where usually shelf space in the breadth is scarce and expensive. Next to this logistical constraint, the marketing department might perceive the filling regimes Tray and/or Loose not suitable for the store format. Tang et al. (2001) analyze the different price formats a retail chain can follow: they consider the whole continuum from Every Day Low Price

(discount stores, e.g. Wal-Mart) to HI-LO or Promotional Pricing (e.g. Ahold formats such as Giant). Changing the filling regime to Tray or Loose might imply that the customers perceive the retail format as a discount store rather than a high-end service oriented store. As such, marketing consid- erations need to be taken into account too when the presentation of the assortment in the store is changed (Campo et al., 2000).

A second way of gaining efficiency in shelf stacking, is to increase all CP sizes to the largest possible size (e.g. from 6 CUs to 12 CUs or from 12 CUs to 24 CUs). In practice, some retailers have already recognized the possible gains especially with regards to their private label products. However, for the branded products the manufacturer decides upon the CP size. Studies show that also brand manufacturers (e.g. Procter and Gamble, Nestlé, etc.) are investigating the consequences of different CP sizes and different packages in the retail supply chain (Saghir and Jönson, 2001). Note that there is a trend observed to reduce the CP sizes (see e.g. Ketzenberg et al., 2000). The above results still apply and then can be used to see how much shelf stacking efficiency is lost in the store due to the reduced CP size.

The third and last option involves the store ordering policy: increase the number of CPs per order line as much as possible. Van Donselaar (1990) and Whybark and Yang (1996) showed that in a 2-echelon distribution system locating most of the inventory close to the customer is the best choice for companies that must fill customer demand from inventory. Putting more inventory on the shelves however implies that there should be enough space to accommodate for this extra inventory. However, shelf space is limited in the breadth, but Broekmeu- len et al. (2006) showed that there is a significant amount of unused space available in the back of the shelf (behind the products), which is called Excess

ARTICLE IN PRESS S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632 629

Shelf Space. Excess Shelf Space is defined as the retail space that is not required to carry out the current operations with respect to customer service and costs. The available space on the shelf is shown to be strongly influenced by the physical dimensions of the product, the CP size, and the shelf dimen- sions. This observation advocates stacking multiple CPs of one product at the same time instead of stacking one CP at multiple times for these products where enough Excess Shelf Space is available. This can be achieved by consolidating replenishment orders (see Van Donselaar et al., 2006a, b).

Finally, it has to be noted that the effect of the worker should not be neglected: fillers who work faster have higher efficiency gains than slower workers. For example, focus on the slowest worker (number 3), he is on average already 2.8 s per CU slower than the median worker who typically spends only 3–5 s on shelf stacking per CU. This result suggests that worker training is an important aspect in order to get the full benefit from the different actions that can be taken. Next to this, it was also observed that due to interruptions of customers in the store, the worker can also slow down in his stacking time. It might thus be worthwhile to evaluate whether it is worthwhile stacking after opening hours.

According to Saghir and Jönson (2001), every second reduction in the total handling time would represent a reduction of five million euro (or 6 million dollar) in the Swedish grocery industry. Since handling costs in the 2 retail chains studied in this paper are around 50 million euro (or 60 million dollar) per year, every reduction in handling time would lead to a substantial increase in yearly profits for these two companies. This indicates that a lot of costs can be reduced by following the above recommendations with regards to the efficiency in handling.

6. Conclusions

It is argued that when store-handling costs have an important share of the retail supply chain operations costs, it is important to know the cost drivers. A conceptual model for shelf stacking in stores was derived using the analogy based on order picking models for warehouses. It was shown that the presented model for the stacking time was, unlike reported in the literature so far, non-linear in CUs. By means of a motion and time study, data was collected in four grocery stores from two

different retail companies. Regression models re- vealed the impact of the most important drivers for shelf stacking efficiency, measured by the TST per CU. The main results of the model are: (1) increasing the CP size results in an average efficiency gain of 24–49%; (2) stacking multiple CPs of one product at the same time instead of stacking one CP at multiple times, results in an average gain between 8% and 31% in TST per CU; (3) the filling regime has a significant effect on the stacking time (12–42%); (4) Increased training, experience and/or motivation may help to improve the working pace of the employees. Based on the presented results both retail chains have decided to structurally change their current operations.

Future research involves extending the currently used reorder policies in the retail companies to take into account the handling efficiency with the replenishment. Usually, the underlying logic is based on a (R, s, nQ)-reorder policy with a dynamic reorder level s. The reorder level s is based on a demand forecast for the coming L+R days (L+R being the sum of the lead time and the review period). The above analysis shows the need for an adapted inventory replenishment rule taking into account the handling aspects. This implies that for the majority of the items the new replenishment logic should be: whenever a replenishment can no longer be postponed, order as many CPs as can be added to the existing inventory on the shelves. Future research is also needed to analyze the impact on handling of different types of packaging materi- al, different shelf maintenance strategies (such as ‘mirroring’) as well as different levels of inventory just before stacking the shelves. Moreover, it is expected that larger CPs also reduce the cost of packaging material and the costs of waste. Increased number of CPs per order line reduces the ordering and delivery frequency of a product, which also may lead to lower ordering costs in the store and to lower picking costs in the retailers’ warehouse.

Appendix A

The definitions which have been used in this research project are given in Table A1.

Appendix B

For descriptive statistics on the different variables see Table B1.

ARTICLE IN PRESS

Table A1

Sub-activity Starting/ending point of sub-activity

Grab/open case pack (G) Start The filler stands in front of the rolling container and reaches for a case pack

End The filler prepares to walk away from the rolling container (case pack is or is not

opened)

or Start The filler has arrived at the shelf location and starts opening the case pack

End The filler is ready with opening the case pack and an other sub-activity starts

Search (S) Start The filler starts with checking the product and he/she lookes for the right shelf location

End The filler sees the right shelf location and prepares to approach it (walk)

Walk (W) Start The filler prepares to walk away from the rolling container or walks after searching the

right shelf location

End The filler stands still in front of the shelves

and Start The filler prepares to walk away from the shelf location or waste disposal place, to the

rolling container

End The filler stands in front of the rolling container and reaches for a case pack

Prepare the shelves/check

‘best before’ date (P)

Start The filler reaches for the old inventory on the shelves and start to check the ‘best before’

date (if needed)

End The filler is ready with preparing the shelves. This means that old inventory is

straightened or is removed from the shelves

Fill new inventory (Fn) Start The filler reaches for the new inventory in the case pack

End The filler reaches for the old inventory or grabs the empty box or plastic

Fill old inventory (Fo) Start In case old inventory was removed from the shelves, the filler starts with putting old

inventory back on the shelves

End The filler is ready with putting old inventory back on the shelves en grabs the empty

box or plastic

Waste disposal (D) Start The filler holds an empty box (or plastic) and starts to flatten it (sometimes the box is

preserved for customers)

End The moment the filler prepares to leave the waste disposal place (a trolley or a place

near the rolling container)

Extra (E) Any activity not part of the first sub -activities, e.g. help a customer, customer is in the way, get or put

away crate, process inventory remainder, organize labels, general cleaning, discuss with a colleague, take

away waste, bring empty boxes for customers to check out area, get a new rolling container, take away

misplaced products, repair a broken product, remove cord from rolling container, take a product to the

kiosk, straighten separation plate

Nota bene: *Grabbing and opening the case pack are taken together, because the individual activities were difficult to separate; **Walking

does not include walking with the rolling container from the storage area to the right aisle or walking with the rolling container between

the aisles. But it does include (in exceptional cases) walking with the rolling container when the rolling container is moved to bring certain

case packs to the right shelf location (e.g. heavy products); ***It is possible that a filler performs multiple sub-activities at once, e.g.

walking while opening the case pack, searching or disposing waste. When this took place, the following reasoning was used: if the walking

time was significantly influenced by the attention focused on opening the case pack (or searching or waste disposal), the time for e.g.

opening the case pack was measured as sub-activity ‘‘G’’, and the remaining time as sub-activity ‘‘W’’. If the walking time was not

significantly influenced by one of these sub-activities, then the total time was measured as walking time (W).

S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632630

ARTICLE IN PRESS T a b le

B 1

C a te g o ry

N u m b er

o f S K U s

N u m b e r

o f o rd e r

li n es

C a se

p a c k si z e (C

U )

] C a se

p a ck s p er

o rd e r li n e

F il li n g re g im

e (] S K U )

T S T (s /C

U )

A v g .

M in

M a x

A v g .

M in

M a x

U T

L A v g .

M in

M a x

C o ff e e

1 1 8

2 1 0

1 2 .3 1

1 3 0

1 .5 3

1 9

9 1

2 1

6 4 .6 1

0 .2 5

2 5 .1 7

C h o c o la te

1 3 7

2 3 5

1 6 .8 9

6 3 3

1 .3 8

1 8

1 1 8

1 9

0 3 .0 4

0 .6 2

1 5 .4 2

C a n d y

2 2 6

3 6 7

1 5 .8 6

6 3 6

1 .1 6

1 3

2 1 9

7 0

3 .0 6

0 .6 5

9 .8 7

B a b y fo o d

8 3

1 0 5

1 0 .6 9

3 1 6

1 .2 0

1 5

2 6

5 7

0 4 .9 4

1 .2 0

1 6 .2 5

C a n n e d m e a t

4 0

4 7

1 2 .9 7

6 2 4

1 .7 7

1 5

4 0

0 0

3 .6 8

0 .6 0

7 .3 3

C a n n e d

v e g e ta b le s

8 9

1 2 2

1 1 .8 6

8 2 4

1 .8 0

1 6

4 6

4 3

0 3 .0 9

1 .1 7

7 .9 2

W in e

9 3

1 4 0

6 .8 4

6 1 5

1 .2 9

1 5

8 4

9 0

4 .7 7

1 .0 0

1 5 .6 7

C o ff e e m il k

3 8

6 3

1 6 .7 3

6 3 0

1 .5 4

1 5

3 1

7 0

3 .6 4

1 .4 3

7 .0 0

S u g a r

1 5

2 3

1 0 .1 7

5 2 0

1 .8 3

1 4

1 1

4 0

1 .5 0

2 .1 0

8 .1 0

C o o k ie s

1 3 0

1 8 5

1 5 .4 6

8 3 6

1 .2 6

1 8

1 2 5

2 3

3 .5 7

0 .6 9

9 .1 7

P e rs o n a l c a re

2 1 4

3 1 2

7 .6 9

3 2 4

1 .1 5

1 3

1 9 3

1 6

5 4 .6 4

0 .8 3

1 1 .5 0

C a n n e d fr u it

2 0

3 2

1 2 .3 7

6 2 4

1 .7 1

1 4

2 0

0 0

3 .5 1

1 .8 3

8 .5 0

S a n d w ic h

sp re a d

6 2

8 1

1 2 .6 0

6 3 0

1 .6 4

1 6

3 9

2 3

0 3 .2 4

0 .7 2

9 .8 3

T o ta l:

1 2 6 5

1 9 2 2

1 0 4 3

2 0 8

1 4

S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632 631

References

Barnes, R.M., 1968. Motion and Time Study Design and

Measurement of Work. Wiley, New York.

Bartholdi, J., Eisenstein, D., 1996. Bucket brigades: A self-

organizing order-picking system for a warehouse, working

paper.

Bartholdi, J., Eisenstein, D., Foley, R., 2001. Performance of

bucket brigades when work is stochastic. Operations

Research 49, 710–719.

Bertrand, J.W.M., Fransoo, J.C., 2001. Operations management

research methodologies using quantitative modeling. Inter-

national Journal of Operations and Production Manage-

ment 22, 241–264.

Broekmeulen, R.A.C.M., 1998. Layout and operations manage-

ment of distribution centers for perishables, Ph.D. Thesis,

Eindhoven University of Technology.

Broekmeulen, R., van Donselaar, K., Fransoo, J., van Woensel,

T., 2006. The opportunity of excess shelf space in grocery

retail store, under review.

Cachon, G., 2001. Managing a retailer’s shelf space, inventory,

and transportation. Manufacturing & Service Operations

Management 3, 211–229.

Campo, K., Gijsbrechts, E., Goossens, T., Verhetsel, A., 2000.

The impact of location factors on the attractiveness and

optimal space shares of product categories. International

Journal of Research in Marketing 17, 255–279.

Chain Store Age, 1963. Cifrino’s Space Yield Formula:

A Breakthrough for Measuring Product Profit 39.

Chain Store Age, 1965. Shelf allocation breakthrough 41, 77–88.

Corstjens, M., Doyle, P., 1981. A model for optimizing

retail space allocations. Management Science 27 (7),

822–833.

van Donselaar, K., 1990. Integral stock norms in divergent

systems with lot-sizes. European Journal of Operational

Research 45 (1), 70–85.

van Donselaar, K., van Woensel, T., Broekmeulen, R., Fransoo,

J., 2006a. Inventory control of perishables in supermarkets.

International Journal of Production Economics, in press.

van Donselaar, K., Gaur, V., van Woensel, T., Broekmeulen,

R., Fransoo, J., 2006b. An empirical study of ordering

behavior of retail stores, under review.

Fleisch, E., Tellkamp, C., 2005. Inventory inaccuracy and

supply chain performance: a simulation study of a retail

supply chain. International Journal of Production Econom-

ics 95 (3), 373–385.

Frazelle, E.H., Apple Jr., J.M., 1994. Warehouse Operations.

The Distribution Management Handbook, McGraw-Hill

Inc., New York.

Gaur, V., Fisher, M.L., Raman, A., 2005. An econometric

analysis of inventory turnover performance in retail services.

Management Science 51 (2), 181–194.

Gujarati, D., 1995. Basic Econometrics. McGraw-Hill Health

Professions Division, 849pp.

Hwang, H., Choi, B., Lee, M., 2005. A model for shelf space

allocation and inventory control considering location and

inventory level effects on demand. International Journal of

Production Economics 97 (2), 185–195.

Ketzenberg, M., Metters, R., Vargas, V., 2000. Inventory policy

for dense retail outlets. Journal of Operations Management

18, 303–316.

ARTICLE IN PRESS S. van Zelst et al. / Int. J. Production Economics 121 (2009) 620–632632

Ketzenberg, M., Metters, R., Vargas, V., 2002. Quantifying

the benefits of breaking bulk in retail operations.

International Journal of Production Economics 80 (3),

249–263.

Mitroff, I.I., Betz, F., Pondy, L.R., Sagasti, F., 1974. On

managing science in the systems age: two schemas for the

study of science as a whole systems phenomenon. Interfaces 4

(3), 46–58.

Rouwenhorst, B., Reuter, B., Stockrahm, V., van Houtum, G.J.,

Mantel, R.J., Zijm, W.H.M., 2000. Warehouse design and

control: framework and literature review. European Journal

of Operational Research 122, 515–533.

Saghir, M., Jönson, G., 2001. Packaging handling evaluation

methods in the grocery retail industry. Packaging Technology

and Science 14 (1), 21–29.

Shah, J., Avittathur, B., 2006. The retailer multi-item in

ventory problem with demand cannibalization and sub-

stitution. International Journal of Production Economics, in

press.

Tang, C.S., Bell, D.R., Ho, T.-H., 2001. Store choice and

shopping behavior, how price format works. California

Management Review 43 (2), 56–74.

Tompkins, J.A., White, J.A., Bozer, Y.A., Frazelle, E.H.,

Tanchoco, J.M.A., Trevino, J., 2003. Facilities Planning,

second ed. Wiley, New York.

Urban, T.L., 1998. An inventory-theoretic approach to product

assortment and shelf-space allocation. Journal of Retailing

74, 15–35.

Wee, K.E., Dada, M., 2005. Optimal policies for transshipping

inventory in a retail network. Management Science 51 (10),

1519–1533.

Whybark, D.C., Yang, S., 1996. Positioning inventory in

distribution systems. International Journal of Production

Economics 45, 271–278.