Cost-Effective Pallet Management Strategies
Cost-Effective Pallet Management Strategies
Cost-Effective Pallet Management Strategies
a r t i c l e i n f o a b s t r a c t
Article history: The existing industry strategies for managing pallets, (single-use expendable pallets, buy/
Received 23 September 2015 sell programs, and leased pallet pooling programs), are analyzed and compared using push
Received in revised form 20 May 2016 and pull inventory control policies. A two-stage integrated framework is developed that
Accepted 17 June 2016
combines cost relationship models with data gathered in industry with multi-echelon
Available online 9 July 2016
inventory performance measures. For the base case, the single-use expendable pallet
approach presents the least cost of all strategies, but the leased pallet pooling programs
Keywords:
outperform the buy/sell programs in terms of total cost. The intervals where each strategy
Pallet logistics
Single-use expendable pallets
is most attractive are shown with respect to pallet cost, salvage cost, dwell fees, effective
Buy/sell pallet programs issue fees, retention rates, and transportation costs.
Leased pallet pooling programs Ó 2016 Elsevier Ltd. All rights reserved.
Pallet fee structure
1. Introduction
Pallets play a central role in the handling and transportation of products through all echelons in the supply chain. Pallets
are used to support unit loads of raw materials that get sent from suppliers to manufacturers; in turn, manufacturers send
finished products on pallets to distributors; and finally, distributors fill requested orders to retailers on pallets. As the struc-
tural foundation of a unit load, pallets provide for efficient and standardized material handling and logistics throughout the
world. With an estimated 80% of United States (U.S.) trade carried on pallets (Raballand and Aldaz-Carroll, 2007), pallets are
undoubtedly the most commonly used unit load platform in the world. Approximately 450–500 million new pallets are
manufactured annually and join the approximately 2 billion pallets that are in circulation in the U.S. (Buehlmann et al.,
2009). In the European Union, some 280 million pallets are in circulation every year.
The pallet industry is vast, complex, and geographically dispersed. In the U.S., the wood pallet and container industry is
composed of 2666 establishments and accounts for $7 billion dollars (estimated receipts) in 2012 (The United States Census
Bureau, 2012; Rupert, 2015). There are three common industry strategies for managing pallets. In describing these systems,
we provide the theoretical term with the industry term in parenthesis.
1. Open-loop system with no salvage value (single-use expendable pallet system): This case involves a single-use expendable
pallet in which the chain of custody of the pallet transfers with the load. Rather than returning to the distributor or man-
ufacturer, the pallet is expendable and is disposed (likely to a landfill). The pallets are whitewood or limited-use pallets
that are chosen due to their use of inexpensive wood, which results in a low purchase price of a pallet.
⇑ Corresponding author at: Indian Institute of Management Ahmedabad, Vastrapur, Ahmedabad, Gujarat 380015, India.
E-mail address: debjit@iima.ac.in (D. Roy).
http://dx.doi.org/10.1016/j.tre.2016.06.005
1366-5545/Ó 2016 Elsevier Ltd. All rights reserved.
D. Roy et al. / Transportation Research Part E 93 (2016) 358–371 359
2. Open-loop system with salvage value (buy/sell program): In existing buy/sell programs, pallets are sold to the customers,
transported with the product throughout the supply chain. After use the pallets are repurchased by a local pallet program
or recycling facility. After inspection, the local pallet program either resells, refurbishes, or disposes of the pallets.
3. Closed-loop rental system (leased pallet pooling program): In leased pallet pooling programs, customers contract the use
of pallets from a pallet pooling provider (essentially renting them). These programs offer a variety of leasing agreements
to customers and place a predetermined number of pallets at user requested locations in the supply chain. After use, pal-
lets are collected at a downstream location in the supply chain (e.g. at a retailer) and then repositioned to a point
upstream in the supply chain (e.g., at the manufacturing echelon, or a consolidation DC). The reverse logistics associated
with this operation (i.e. backhaul of pallets and pre-position of pallets), as well as the refurbishing activities gain signif-
icant importance. Leased pallet pooling programs rely on a network of return depots that collect empty pallets from many
operations and perform inspection, sortation, repair, and backhauling. The return depots can be operated either by the
pallet pooling provider or by locally contracted pallet recyclers.
This work develops an integrated two-stage framework that is used to understand the trade-offs pallet users face when
selecting a pallet management strategy. Our analysis is conducted with respect to a single user/company who operates a
multi-echelon supply chain. For example, many retail and grocery companies, like Sam’s Club, Costco, Walmart, Publix,
own and operate distribution centers and retail stores, and are thus responsible for the pallets at both the distributor and
the retailer. Our work provides new knowledge into the trade-offs of the choice of a pallet supply chain management strat-
egy by comparing strategies based on the total pallet costs. Total pallet costs incorporate both the multi-echelon nature of
pallet management, as well as varying cost structure associated with purchasing, holding, ordering, and transporting pallets.
To analyze the cost trade-offs we develop queuing network models for both open and closed-loop pallet supply chains oper-
ating under two variants of inventory control policy: push and pull. These models provide inputs to the cost analysis model
for the pallet management selection problem.
This paper is organized as follows. The literature on pallet management strategies and related areas are reviewed in Sec-
tion 2. The contributions of our work are articulated in Section 3, which includes development of a two-stage framework. The
first stage are cost relationship models, which are presented in Section 4. The second stage includes multi-echelon inventory
performance estimates, which we analyze through the stochastic open-loop and the closed-loop models presented in Sec-
tion 5. Through numerical experiments and cost model sensitivity analysis, in Section 6 cost intervals where each strategy
is most attractive are presented with respect to pallet costs, salvage costs, dwell fees, effective issue fees, retention rates, and
transportation costs. The conclusions and directions for future research from this study are included in Section 7.
2. Literature review
Pallet supply chains differ from the traditional product supply chains in several dimensions. For instance, the traditional
product supply chains have a predictable forward and reverse material flow among its actors such as: suppliers, manufac-
turer, distributor, wholesaler, retailer, end customer, recycler (if applicable) (Savaskan et al., 2004; Atasu et al., 2008, 2013).
However, such a predictable flow is absent from the pallet flows, where the pallet can move from any actor in the supply
chain to a repair facility and sent back (after repair) to another actor. Product supply chains exhibit cyclic flow patterns,
whereas pallet supply chains exhibit network flow patterns. Further, pooled pallets are typically used for a large number
of cycles compared to a traditional product. In addition, the recovery/disposal of pallets are closed to 100% especially in pallet
pooling programs unlike products where the product usage is dependent on consumer usage and behavior.
Most of the literature with respect to pallet management is focused on pallet loading issues (Ram, 1992), regional differ-
ences of pallet systems (Guzman-Siller et al., 2010; Jin et al., 2008), specific military applications and systems (Harris and
Worrell, 2008; Peterson, 2005), repair-replacement models (Tornese et al., 2016; Lanzenauer et al., 1978), or pallet end-
of-life scenarios (Carrano et al., 2014; Buehlmann et al., 2009; Gasol et al., 2008; Corbiere-Nicollier et al., 2001). We focus
our review on the limited archival literature available on pallet supply chain management and logistics. This work can be
categorized into deterministic and stochastic pallet management models, depending on the certainty of pallet flows and
demands at each point of the chain. We also briefly comment on selected literature on empty container supply chain man-
agement that has similarities to pallet logistics models.
Deterministic management models: These works treat demand for pallets deterministically and assume that the num-
ber of pallets that flow from each echelon is known with certainty. Kroon and Vrijens (1995) analyze the design of a return
logistics system for returnable packaging (of which pallets are an example) that facilitates the transportation of returnable
packaging from recipient to sender. They develop a mixed-integer programming model to determine how many units of
returnable packaging should be available in the system; how many depots and where the depots should be located; how
should the distribution, collection, and relocation of returnable packaging be organized; and what should be the fee struc-
ture. Duhaime et al. (2001) develop a minimum-cost flow model to plan the distribution of empty returnable containers in
the Canada postal network to determine if the pool of containers was sufficient to meet demand and whether the containers
were located at the correct locations in the network. Comparison of their model with historical data confirms that the com-
pany has adequate containers to meet demand; however, the company needs to improve the management of its return con-
tainers. Bilbao et al. (2011) provide a framework for choosing the pallet management strategy and the material with which
360 D. Roy et al. / Transportation Research Part E 93 (2016) 358–371
the pallet is constructed that considers both cost and environmental factors associated with the pallet’s life cycle (i.e. man-
ufacturing, procurement, use and disposition). Using deterministic optimization models, Carrano et al. (2015) explored the
effectiveness of mixed pallet management strategies in minimizing carbon equivalent emissions under various loading, han-
dling, and end of life scenarios. An integer linear programming model is developed in Tornese et al. (2016) that considers the
benefit to a pallet supply chain of pre-emptive component repair.
Stochastic pallet management models: Gnoni et al. (2011) develop a discrete-event simulation model to evaluate dif-
ferent organizational scenarios for pallet management. The focus of their study is on the impact of empty pallet interchange
procedures (i.e., exchanging pallets directly or using a pallet voucher system). They compare three management strategies:
direct management (where all pallet activities are carried out by a firm), outsourcing management (where a firm purchases
the pallets but the management of the pallets are outsourced), and pallet pooling (where a third-party provider rents out and
manages the pallets). They find that a management strategy that interchanges empty pallets without a pallet voucher is most
effective when supply chain coordination of the echelons exists. Elia and Gnoni (2015) develop a discrete-event simulation
model to analyze the effect of both forward and reverse flows of pallets on the cost performance of closed loop pallet man-
agement systems. Their analysis consider both direct and postponed pallet interchange conditions.
Empty container supply chain management models: Both pallets and containers can be reusable and their reverse (or
empty) flows must be managed. Consequently, pallet supply chain management decisions have similarities to empty con-
tainer supply chain management literature. However, because pallets and containers are characteristically different (i.e., pal-
lets are transported in batches, whereas containers are single item shipments, pallets are cheaper and have shorter life cycles
than containers, and pallets require repair prior to reuse whereas containers typically do not), the management of their
reverse supply chains are fundamentally different. For examples of empty container management models, see Dejax and
Crainic (1987), Li et al. (2007), Dong and Song (2009), and Choong et al. (2002).
In spite of the ubiquity of pallets, pallet logistics have not been extensively studied and scientific literature is extremely
limited (as was discussed in Section 2). One of the reasons for the lack of study in pallet logistics might be the fact that the
pallet logistic networks are complex and industry data is generally not publicly available, greatly vary across different busi-
ness practices, and can be difficult to estimate in an exact manner. Also, in an industry where 99% of the establishments
(2365 out of 2666) (The United States Census Bureau, 2012) are small business, the bandwidth and resources to embark
in these studies and models is almost non-existent.
Therefore, the contribution of this work is in developing a two-stage approach framework to understand the trade-offs
users face when selecting a pallet management strategy. To do so, we develop cost relationship models that combine cost
input parameters gathered through observations and time studies at various pallet industries with multi-echelon inventory
performance estimates. The framework is illustrated in Fig. 1. The first stage involves developing cost relationship models for
the three strategies. These cost relationship models incorporate the fees associated with purchasing, holding, ordering, and
transporting pallets with multiple-echelon inventory performance estimates of the expected number of pallets at each of the
entities. Our framework is general and can be used to evaluate pallet strategies given any inventory policy. Given that each of
the three pallet management strategies will have different cost structures, the developed framework can be used to consider
specific cost values associated with each strategy, as well as customer characteristics (such as the type and demands of pro-
duct transported and the configuration of their supply chain). The pallet flow supply chain network is especially challenging
to analyze due to the stochastic nature of the dwell times of pallets at each echelon in the supply chain, the stochastic nature
of the demand requests for pallets, the likelihood that a pallet will be recovered at the end of the supply chain, and the uncer-
tain movement of pallets throughout the supply chain. Therefore, we develop both an open-loop and a closed-loop stochastic
model that are able to handle uncertainties associated with pallet flows at different echelons in the supply chain considering
two inventory control policies: (1) a push system, and (2) a pull system. In the push system, the loaded pallet batches are
dispatched from the upstream to the downstream facility (supplier-to-distributor or distributor-to-retailer) without consid-
ering the current inventory position at the downstream facility. In the pull system, the upstream facility checks the current
inventory position at the downstream facility before dispatching loaded pallet batches. In this work, we use stochastic multi-
Fig. 1. Two-stage approach framework to analyze and compare pallet management strategies.
D. Roy et al. / Transportation Research Part E 93 (2016) 358–371 361
echelon inventory network models to capture the dynamics among the supplier (of pallets loaded with products), a distrib-
utor of unit loads, a retailer, and a pallet repair depot (for the leased pallet pooling program). These models provide the sec-
ond stage system performance estimates. A separate open-loop model for a push and pull inventory policy is used for the
single-use expendable, as well as the buy/sell strategy. A buy/sell strategy is modeled as an open queuing network because
the number of pallets in the system is never constant. In contrast, the leased pallet pooling program is modeled as a closed-
loop model as the pallet pooling provider reuses the pallets in the supply chain. For push and pull inventory policies, the
analysis of both open and closed-loop is done through the use of a discrete event simulation.
In this section, we present cost relationship models for the three pallet management strategies. Before presenting the
relationship models, we describe the costs, from a pallet user perspective, associated with each scenario. The specific costs
described below are summarized in Table 1. This data is based on direct observations and time studies performed at eight U.
S. facilities involved in the various aspects of pallet manufacturing, refurbishing and disposal operations. These include lum-
ber mills, pallet manufacturers, pallet recyclers, and pallet pooling providers (both 3rd party-owned and 3rd party-managed
pooling companies). A documentation of standard practices, as well as interviews were conducted during these visits.
Table 1
Summary of user cost data for performance comparison.
Backlog costs (hb ): The backlog cost per pallet is assumed to be the same as the storage and operational holding
costs, $2.80/pallet position held over a year.
3. Closed-loop rental system (leased pallet pooling program):
Purchase cost (c): Given the retailer rents (and does not own) pallets, the purchase cost of the pallet is the respon-
sibility of the pooling provider and the user does not incur a purchase cost in the closed-loop rental system. Thus,
the cost to the pallet user is $0 per pallet.
Salvage value (s): $0/pallet as the pallet remains the property of the pallet pooling provider.
Lost pallet cost (l): Users are typically charged $22 per pallet for every pallet lost or otherwise unaccounted for. A
reusable (extended use) high-grade block pallet is assumed. We denote the probability that a pallet is lost after use
as pl , which is typically around 5–15%.
Effective issue fee (kissue ): This is a composite fixed cost per pallet that is made up of an initial issue fee, an admin-
istration fee, and a transport fee. Sometimes this fixed cost also includes a transfer fee assessed when the pallet is
picked up and which amount varies depending on the cross-docking distance to the next repositioning location.
The effective issue fee is typically in the neighborhood of $4–$6 per pallet issued.
Dwell fee (hdwell ): The dwell fee is assessed as a days-on-rent fee. This is about 3 cents/day or 0.03 ⁄ 365 = $10.95
pallet held over a year.
Storage and operational holding costs (ho ): This includes the storage and operational holding costs of average
inventory held at the distributor or retailer and is $2.80/pallet position held over a year (same as in the previous
case).
Backlog costs (hb ): The backlog cost per pallet is assumed to be same as the storage and operational holding costs,
$2.80/pallet position held over a year.
For the three strategies, we develop annual total cost functions that incorporates the baseline cost parameters in Table 1
with the expected inventory levels in a multi-echelon supply chain. In this work, we consider both a push and pull inventory
system and obtain expected inventory levels from the stochastic models developed in Sections 5.1 and 5.2. To simplify our
exposition, we scope our analysis to distributor and retailer facility only. The performance outputs include E½Id ; E½Idr t , and
E½Ir , which denote the expected number of pallets in inventory at the distributor, in transit, and at the retailer, respectively.
Other outputs include the expected number of backlogs at the distributor and the retailer denoted by E½Bd and E½Br , respec-
tively. Note, if the user is not responsible for the pallets at any echelon, our models can still be used by setting the expected
number of pallets in inventory at that echelon to zero. Also, the expected number of pallets in inventory at the supplier is not
considered in the cost model, but can be easily incorporated if the user also owns the pallets at the supplier echelon.
Descriptions and representative data for the input parameters are provided in Table 1. Inputs to the stochastic models
that are needed for the cost models include the expected demand rate of the distributor, kd , and its batch size b2 ; and the
expected demand rate at the retailer, kr in batch size b3 . Note, that given Eqs. (2)–(4) are provided on an annual basis, kd
and kr are required to be given in a rate per year. Outputs to be obtained from a multi-echelon supply chain are
E½Id ; E½Idr
t , and E½I r .
Eq. (1) provides the annual total cost for the single-use expendable pallet strategy (TC no-salv age ), which considers the pallet
purchasing costs, ordering costs in the form of transportation costs, holding costs due to storage operations and to capital
investment, and backlog costs. We obtain the expected inventory performance measures for both a push and pull inventory
strategy using the open-loop models in Section 5.1.
Eq. (2) provides the annual total cost for the buy/sell program (TC salv age ), which includes an effective purchasing cost (the
purchasing cost per new pallet minus the salvage value multiplied by the retention rate). We obtain the expected inventory
performance measures for both a push and pull inventory strategy using the open-loop models in Section 5.1.
Eq. (3) provides the user’s annual total costs for the leased pallet pooling program (TC rental ). The first term is the cost of lost
pallets, where pl is defined as the probability that a pallet is lost after use. The second term is an effective issue fee (per pal-
let) that is charged every time a pallet is dispatched to the distributor regardless of whether the pallet arrives new or from
the repair depot. The third term is the holding costs in the form of operational and storage holding costs occupying at the
distributor and retailer. The fourth term is the dwell fee charge based on the average inventory the user has in their posses-
sion. The fifth term is the backlog costs. As the user rents the pallets from a pooling provider, the purchasing and repair costs,
as well as the holding costs at the repair facility, are the responsibility of the pooling provider. We obtain the expected inven-
tory performance measures for both a push and pull inventory strategy using the closed-loop models in Section 5.2.
TC rental ¼ lpl ðkr b3 Þ þ kissue ðkr b2 Þ þ ho ðE½Id þ E½Ir Þ þ hdwell ðE½Id þ E½Idr
t þ E½Ir Þ þ hb ðE½Bd þ E½Br Þ ð3Þ
D. Roy et al. / Transportation Research Part E 93 (2016) 358–371 363
To obtain the expected inventory levels in a multi-echelon supply chain necessary for our cost relationship models, we
develop stochastic models that capture the system dynamics in a pallet supply chain. We are interested in performance mea-
sures such as the average carrying inventory and backlogs at the distributor and the retailer facility. The dynamics in the
open and the closed-loop system are complex to model in a deterministic setting due to interactions present among the enti-
ties such as distributor, supplier, retailer, and the repair depot. Hence, we model the pallet flow supply chain network as a
stochastic multi-echelon inventory network (with or without feedback loops) and adopt queuing network models to capture
the modeling elements. Further these models capture the effect of several design parameters that influence system perfor-
mance: the accumulated batch size for repair, the supplier’s, distributor’s, and retailer’s storage capacity, the first and the
second moments of the demand inter-arrival times at the supplier, distributor, and the retailer, the first and the second
moments of the pallet arrival times to the supplier, distributor, and retailer, the batch size of the distributor’s, retailer’s,
and customer’s demand, the travel times between supplier and distributor, and between distributor and retailer, the repair
process times, and the fraction of pallets undergoing repair.
In this paper, we consider that the pallet user’s supply chain has three basic facilities, a supplier, a distributor, and a retai-
ler. We demonstrate the supply chain model workings under two inventory control policy: (1) a push system, and (2) a pull
system. In the push system, the current inventory position of the distributor and retailer are not accounted for when ship-
ping loaded pallet units to their downstream facilities. Contrastingly, in the pull system, the inventory position of the dis-
tributor and retailer facilities are explicitly considered before shipping loaded pallet units to their facilities. We first
describe the push system and then discuss the pull system.
Fig. 2. Pallet flow model without feedback loop for (a) push system and (b) pull system.
364 D. Roy et al. / Transportation Research Part E 93 (2016) 358–371
In this section, we describe a pallet flow model in which a fraction of pallets return to the distributor facility after being
processed at a repair depot, which is modeled as a stochastic closed-loop model shown in Fig. 3. Similar to the open-loop
model, we have two variants of the underlying supply chain inventory control policy: a push and a pull policy.
D. Roy et al. / Transportation Research Part E 93 (2016) 358–371 365
Fig. 3. Pallet flow model with feedback loop for (a) push system and (b) pull system.
The supplier will receive pallet supply from both the external sources, as well as from the repair depot. The arrivals from
the external sources and the repair depot wait in a buffer to form the batches according to the distributor’s demand batch
size. Once the distributor’s demand meets the supply, the demand is satisfied and this demand is supplied to the distributor
1
facility with a transportation delay (ðlsd
t Þ ). Likewise, at the distributor’s facility, the retailer’s demand meet the supply and
1
this demand is supplied to the retailer facility with a transportation delay (ðldr
t Þ ). At the retailer facility, a fraction of pallets
(p1 ) waits for the batch formation after the goods on the pallet are sold.
We model a strategy where all pallets are sent to the repair depot for inspection irrespective of the pallet’s condition. This
strategy is commonly followed by pallet providers in North America because the expertise and infrastructure associated with
pallet inspection can be centralized at the repair depot, and also consolidates transportation costs of empty pallets. At the
repair depot, depending on the pallet condition, some pallets undergo repair whereas some do not. Thus, to model this vari-
ation in repair times we model the process time at the repair depot as a random variable. The pallets are repaired in fixed
batch sizes with mean repair time for each pallet as l1 rep . There may be more than one repair stations at the depot. The batch
of pallets reaches the supplier facility after a transportation delay (l1
t 2 ).
Due to complex batching and synchronization requirements, this closed-loop model is analyzed using a discrete-event
simulation model. In addition to the expected available inventory of pallets at the different entities, we also estimate the
amount of pallet inventory at the repair process and in between the repair process and the supplier’s facility. We distinguish
the repair process inventory (including the pallets waiting for batch formation for repair) from the remaining inventory in
the system. The repair process inventory, as well as the inventory between the repair process and the supplier’s facility
belong to the pallet provider whereas the other inventory belongs to the user.
366 D. Roy et al. / Transportation Research Part E 93 (2016) 358–371
Note that in the numerical section, we compare the cost model of the open-loop and closed-loop pallet management sys-
tem. To compare the performance between the open and the closed-loop systems, we ensure that the effective arrival rates
of pallets to the supplier node in both systems are equal. In the closed-loop case, the rate of pallet arrival at the supplier
facility is the sum of the rate of pallet arrival from the external sources and the repair depot. This relation is expressed using
Eq. (4).
b1 ks p1 b4 kr
kex ¼ ð5Þ
bex
For instance, if the external arrival rate (ks ) for the open-loop case is 1.2 batches/day, then the external arrival rate (kex ) for
the closed-loop case is calculated such that the effective pallet arrival rate at the supplier facility is the same as the external
arrival rate for the open-loop case. Thus, if 70% of pallets (p1 ) are sent to the repair depot and the retailer’s demand to the
distributor facility (kr ) is 0.96 batches/day, then the external arrival rate to the distributor facility for the closed-loop case is
0.528 batches/day assuming that the batch sizes b1 ; b2 , and bex are equal.
6. Numerical experiments
For our numerical experiments, we reference a base case, which uses the values included in Table 2. Based on the distrib-
utor receiving requests for one truck-load of loaded-pallets per day from the supplier, demand for pallets at the retailer was
set at kr = 0.095, 0.145, 0.195 pallet batches per hour, respectively for the high, medium, and low demand conditions. Note
that the pallets are received with unit-loads from the distributor to the retailer. However, the pallets are broken and shipped
from retailers to customers in cases or eaches. The other parameters are based on our discussions with pallet users and pro-
viders. The retailer’s demand for loaded-pallets at the distributor would vary based on the industry type (for example appa-
rel versus beverage). We use two sufficient conditions to obtain tractable results in our models: (1) impose limits on buffer
Bi1 ’s for all queues (except the repair center, which is a first-come-first-serve single server queue) using K i ’s, where i 2 fs; d; rg
and (2) assume that for each facility the demand rate is less than the pallet arrival rate. These conditions do not preclude the
formation of backlogs or holding pallet inventory. There will be states in the stochastic model where the demand is waiting
to be fulfilled (backlogs) and states where pallet inventory is waiting.
From a pallet user’s perspective, we explore trade-offs and insights associated with the different pallet management
strategies. To do so, we apply our framework that combines the cost relationship models with outputs from the stochastic
models for both push and pull inventory policies. This framework captures the inventory requirements, as well as the costs
associated with purchasing, holding, ordering, backlog, and transporting pallets. The total annual costs of the three strategies
Table 2
Input data for the base case.
Table 3
Annual total costs for the three strategies for both pull and push system using the cost parameters provided in Table 1.
for three customer demand cases are calculated and shown in Table 3. These costs are based on using the input costs param-
eters from Table 1, the base case input values from Table 2, and setting ps and pl to 0.90 and 0.07, respectively.
The least-to-most expensive pallet management strategy for the user is the single-use expendable pallet strategy, the
leased pallet pooling program, and the buy/sell program. However, this recommendation is sensitive to the input parame-
ters, which vary based on the user’s supply chain network and the fee structure. As the customer demand increases, the
expected inventory in the system increases, which results in the open-loop strategies’ total costs to increase due to increase
in holding costs. In particular, the buy/sell program is most expensive because of high effective purchase cost per pallet
($6.30 vs. $5.0). For a leased pallet pooling program, the holding costs also increase as the demand increases. However,
the total costs in the leased program is lower than the buy/sell program. We also observe that the push strategy, in general,
has a higher cost than the pull strategy.
Next, we are interested in the sensitivity of the annual total costs functions for the three strategies to changes in the cost
parameters. We conduct a one-factor-at-a-time sensitivity analysis because we are interested in the breakpoints of when the
different strategies are preferred. For parameters that impact the open-loop systems, we vary the cost of a no-salvage-value
pallet, the cost of a salvage pallet, the salvage value of the pallet, the percentage of salvaged pallets (retention rate), and
ktransport . For parameters that impact the rental system we vary the effective issue fee (kissue ), the dwell fee (kdwell ), and the lost
pallet percentage. We also vary the MARR, which impacts all strategies. We explore the impact of the cost parameters for the
cases in Table 1; however, because the behavior is shown to be robust against variation of the demand rate, only the behavior
for the case where kr = 7 pallets per hour with the pull inventory system is presented.
Fig. 4. Impact of the purchase cost of a single-use expendable pallet on the total annual costs; break-even points: Buy/sell = $6.40, leased pallet pooling =
$5.60.
368 D. Roy et al. / Transportation Research Part E 93 (2016) 358–371
Fig. 5. Impact of the pallet purchase cost for a buy/sell program pallet on the total annual costs and different salvage values; break-even points: s = $1.50:
Expendable = $7.65, leased pallet pooling = $8.30; s = $2: Expendable = $8.10, leased pallet pooling = $8.80; and s = $2.50: Expendable = $8.60, leased pallet
pooling = $9.20.
Fig. 6. Impact of the transportation drop fee (ktransport ) on the total annual costs.
strategies that support multiple-use pallets (i.e., leased pallet pooling and buy/sell programs). The pallet customers that ship
their products on single-use pallets have different pallet performance expectations (none extending beyond the single
loaded trip) and thus engage in negotiations (with aggressive competition from whitewood pallet manufacturers) on the
key driver parameter for this strategy that is the purchase cost. This is all reflected in the current pricing scheme. From a
pallet user perspective, this graph establishes a cost threshold at around $5.60 per single-use expendable pallet in which
other alternatives become attractive.
Fig. 5 depicts the behavior of the total cost function when the purchase cost of a buy/sell program pallet (with residual
salvage value) is varied. This models the existing buy/sell programs in industry. Three levels of salvage value ($1.50, $2.00,
and $2.50) were investigated for a moderate demand-to-pallet-supply ratio and 90% constant retention rate. From interviews
and observation of industry practices, the baseline assumption is that 10% of the pallets at any given time are lost due to
various attrition factors, thus making the 90% retention rate consistent with current industry practices. As expected, as
the salvage cost increases, a postponement of the breakeven points with the alternative strategies takes place. When the sal-
vage value is increased between $1.50 and $2.50 per pallet, the single-use expendable pallet strategy becomes attractive
around $7.65–$8.60, while the leased pallet pooling program becomes feasible around $8.30–$9.20, respectively. It should
be noted that the salvage value of a pallet in a buy/sell program is usually negotiated on a per-batch (e.g. per truck) basis
and greatly varies. It depends on many factors, including the mix of pallet sizes and conditions, local supply and demand
of pallets, the contractual conditions with the recyclers in the area, among others. Current pricing schemes for recycled pal-
lets are in the range of $1.50–$2.25 per pallet, which fall within the range shown in he graph. With respect to the purchase
cost, when comparing the two strategies that support multiple-use pallets (i.e., leased pallet pooling and buy/sell programs),
all the breakeven points for the three scenarios fall in the range between $7.65 and $9.20. Given the current industry pricing
scheme is between $8.00 and $10.50, this range also appears highly sensitive to changes in the pallet pricing schedule.
Fig. 6 depicts the impact of transportation fees related to the delivery of pallets to the user upon the total cost. This linear
relationship increases for the purchase-based strategies but remains constant for the leased pallet pooling program. The rea-
son is that pallet rental pricing schedules incorporate the transportation fees in one or more components of the effective
issue fee (transfer fee, initial issue fee or both) thus is not shown as a separate cost parameter. With respect to the purchase
based systems, practicality determines that most pallet users source their pallets from manufacturers located in the vicinity,
typically within 25 miles. This establishes a nominal fee around $50–$60 per batch delivered. If the distance were to increase
D. Roy et al. / Transportation Research Part E 93 (2016) 358–371 369
in a way that the transportation fee would approach $84, then the option of leased pallet pooling program would become
attractive.
Fig. 7. Impact of the effective issue fee (kissue ) on the total annual costs for the 80% demand-to-pallet-supply ratio; break-even point: Expendable = $4.35,
buy/sell = $5.65.
370 D. Roy et al. / Transportation Research Part E 93 (2016) 358–371
Fig. 8. Impact of the dwell fee (kdwell ) on the total annual costs for a medium demand-to-pallet-supply ratio. Break-even points: Buy/sell = $0.17.
$510,000
$490,000
Total Annual Costs
$470,000
$450,000
$430,000
$410,000
$390,000
$370,000
$350,000
0.0% 2.5% 5.0% 7.5% 10.0% 12.5% 15.0%
Lost Percentage for Leased Pool Program
Fig. 9. Impact of the lost percentage for the leased pallet pooling program (pl ) on the total annual costs; break-even point: Expendable = 4.10%, buy/
sell = 10.20%.
are recovered (based on the strength and location of the local recycling network) and establish an ‘‘NPD surcharge fee” per
pallet in the entire contracted pool. This fee typically falls in the range of $0–$1.50 per pallet (Ray et al., 2006). For the sce-
nario in which pallets are assessed a full lost pallet cost, this graph suggests that any pallet user that ships more than 10% of
their pallets to NPDs or that otherwise experiences an attrition of such magnitude should re-evaluate their pallet logistics
system and consider a buy-sell management option (or alternatively renegotiate their fee structure).
7. Conclusions
We develop a framework that combines cost relationship models with multi-echelon inventory system performance esti-
mates. We use the framework to analyze three common industry pallet management strategies: a single-use expendable pal-
let system, a buy/sell program system, and a leased pallet pooling program. To obtain the inventory performance estimates,
we develop stochastic models that capture the uncertainty associated with pallet arrival rates, inventory dwell times, and
repair processes in pallet supply chains under both push and pull inventory control policy.
We conduct an experimental design to provide insights into a user’s selection of a pallet strategy. The results using the
base case input cost values show that, while the single-use expendable pallet strategy presents the least cost of all strategies,
the leased pallet pooling programs outperform the buy/sell programs in terms of total cost. However, this recommendation
is sensitive to cost parameters that would vary based on the user’s supply chain network and fee structure. Also, all current
cost parameters used by industry fall within the breakeven points estimated by the theoretical model. Results show that the
operational range for the effective issue fee that results in the leased pallet pooling programs as the most competitive option
is between $4.35 and $5.65. Also, an increase in the purchase cost of expendable pallets beyond $5.60 would result in other
options becoming more attractive. When the salvage value of pallets under the buy/sell programs is set at $1.50 and the
retention rate is kept at 90%, the buy/sell option remain attractive for pallet purchase costs up to $7.65 (when compared
against the expendable pallet alternative) and up to $8.30 (when compared against the leased pallet pooling alternative).
Increasing the salvage value of buy/sell pallets from $1.50 to $2.50 increases the breakeven point at which buy/sell programs
cease being an attractive alternative by 11–12% regardless of the alternative. Other results show the impact and robustness
of various cost parameters on the total cost of the three strategies.
Future research directions include incorporating additional objectives beyond costs for selecting a pallet management
strategy. Other extensions are to develop models that consider the uncertainty associated with pallets that are shared among
multiple customer supply chains. For example, uncertainty on the location where a pallet will end up clearly exists and the
D. Roy et al. / Transportation Research Part E 93 (2016) 358–371 371
pallet pooling provider may only have limited knowledge of the mix of participating versus non-participating customer loca-
tions. Thus, of interest are examining strategies to handle the redistribution of pallets. Finally, the development of models to
understand policies that can handle the seasonality of the business can be pursued.
Acknowledgment
This project was funded by Material Industry of America (MHIA) through its Research Grant #11100576. The field data
collection needed to conduct this study was only possible thanks to the following graduate students: Abbey Burns, Andres
Morales, Darinee Niramarnkarn, Ricki Pavia and Shwe Sin Win.
References
Atasu, A., Guide, V.D.R., Van Wassenhove, L.N., 2008. Product reuse economics in closed-loop supply chain research. Prod. Oper. Manage. 17 (5), 483–496.
Atasu, A., Toktay, L.B., Van Wassenhove, L.N., 2013. How collection cost structure drives a manufacturer’s reverse channel choice. Prod. Oper. Manage. 22 (5),
1089–1102.
Bilbao, A.M., Carrano, A.L., Hewitt, M., Thorn, B.K., 2011. On the environmental impacts of pallet management operations. Manage. Res. Rev. 34 (11), 1222–
1236.
Buehlmann, U., Bumgardner, M., Fluharty, T., 2009. Ban on landfilling of wooden pallets in North Carolina: an assessment of recycling and industry capacity.
J. Clean. Prod. 17 (2), 271–275.
Buzacott, J.A., Shanthikumar, J.G., 1993. Stochastic Models of Manufacturing Systems. Prentice Hall, Englewood Cliffs, NJ.
Carrano, A.L., Pazour, J.A., Roy, D., Thorn, B.K., 2015. Selection of pallet management strategies based on carbon emissions impact. Int. J. Prod. Econ. 164 (0),
258–270.
Carrano, A.L., Thorn, B.K., Woltag, H., 2014. Characterizing the carbon footprint of wood pallet logistics. Forest Prod. J. 64 (7–8), 232–241.
Choong, S.T., Cole, M.H., Kutanoglu, E., 2002. Empty container management for intermodal transportation networks. Transp. Res. Part E: Logis. Transp. Rev.
38 (6), 423–438.
Corbiere-Nicollier, T., Gfeller Laban, B., Lundquist, L., Leterrier, Y., Månson, J.-A.E., Jolliet, O., 2001. Life cycle assessment of biofibres replacing glass fibres as
reinforcement in plastics. Resour., Conserv. Recycl. 33 (4), 267–287.
Dejax, P.J., Crainic, T.G., 1987. A review of empty flows and fleet management models in freight transportation. Transp. Sci. 21 (4), 227–248.
Dong, J.-X., Song, D.-P., 2009. Container fleet sizing and empty repositioning in liner shipping systems. Transp. Res. Part E: Logis. Transp. Rev. 45 (6), 860–
877.
Duhaime, R., Riopel, D., Langevin, A., 2001. Value analysis and optimization of reusable containers at Canada Post. Interfaces 31 (3), 3–15.
Elia, V., Gnoni, M.G., 2015. Designing an effective closed loop system for pallet management. Int. J. Prod. Econ. 170 (Part C), 730–740.
Gasol, C.M., Farreny, R., Gabarrell, X., Rieradevall, J., 2008. Life cycle assessment comparison among different reuse intensities for industrial wooden
containers. Int. J. Life Cycle Assess. 13 (5), 421–431.
Gnoni, Maria Grazia, Lettera, Gianni, Rollo, Alessandra, 2011. A simulation comparison analysis of effective pallet management scenarios. In: 2011 IEEE
International Conference on Industrial Engineering and Engineering Management, pp. 1228–1232.
Gstettner, S., Kuhn, H., 1996. Analysis of production control systems Kanban and CONWIP. Int. J. Prod. Res. 34 (11), 3253–3273.
Guzman-Siller, C.F., Twede, D., Mollenkopf, D.A., 2010. Differences in the perception of pallet systems between US and Canadian grocery retailers. J. Food
Distrib. Res. 41 (3).
Harris, J.S., Worrell, J.S., 2008. Pallet Management System: A Study of the Implementation of UID/RFID Technology for Tracking Shipping Materials within
the Department of Defense Distribution Network (Master’s Thesis). Naval Postgraduate School.
Jin, Shousong, Ge, Haiqing, Jiang, Meixian, Feng, Dingzhong, 2008. A discussion about Chinese Pallet Pool Co., Ltd. In: 4th International Conference on
Wireless Communications, Networking and Mobile Computing, 2008. WiCOM’08, pp. 1–4.
Krieg, G.N., Kuhn, H., 2008. Performance evaluation of two-stage multi-product Kanban systems. IIE Trans. 40 (3), 265–283.
Krishnamurthy, A., Suri, R., 2006. Performance analysis of single stage Kanban controlled production systems using parametric decomposition. Queueing
Syst. 54 (2), 141–162.
Kroon, L., Vrijens, G., 1995. Returnable containers: an example of reverse logistics. Int. J. Phys. Distrib. Logis. Manage. 25 (2), 56–68.
Lanzenauer, V., Haehling, C., Wright, D.D., 1978. Developing an optimal repair-replacement strategy for pallets. Naval Res. Logis. Quart. 25 (1), 169–178.
Li, J.-A., Leung, S.C.H., Wu, Y., Liu, K., 2007. Allocation of empty containers between multi-ports. Eur. J. Oper. Res. 182 (1), 400–412.
Peterson, A.J., 2005. An Examination of Reverse Logistics Factors Impacting the 463-L Pallet Program (Master’s Thesis). Air Force Institute of Technology.
Raballand, G., Aldaz-Carroll, E., 2007. How do differing standards increase trade costs? The case of pallets. World Econ. 30, 685–702.
Ram, B., 1992. The pallet loading problem: a survey. Int. J. Prod. Econ. 28 (2), 217–225.
Ray, C.D., Michael, J.H., Scholnick, B.N., 2006. Supply-chain system costs of alternative grocery industry pallet systems. For. Prod. J. 56, 52–57.
Rupert, R., 2015. Technical Communication with Manager of Unit Load. Millwood Inc. (Date of communication, 06-November-2015).
Savaskan, R.C., Bhattacharya, S., Van Wassenhove, L.N., 2004. Closed-loop supply chain models with product remanufacturing. Manage. Sci. 50 (2), 239–252.
Sugimori, Y., Kusunoki, K., Cho, F., Uchikawa, S., 1977. Toyota production system and Kanban system materialization of just-in-time and respect-for-human
system. Int. J. Prod. Res. 15 (6), 553–564.
The United States Census Bureau, 2012. Industry Statistics Portal. <http://www.census.gov/econ/isp/sampler.php?naicscode=321920&naicslevel=6>
(accessed 07-November-2015).
Tornese, F., Carrano, A.L., Thorn, B.K., Pazour, J.A., Roy, D., 2016. Carbon footprint analysis of pallet remanufacturing. J. Clean. Prod. 126, 630–642.
Wang, H., Wang, H.-P. (Ben), 1991. Optimum number of Kanbans between two adjacent workstations in a JIT system. Int. J. Prod. Econ. 22 (3), 179–188.