Expert Systems with Applications 38 (2011) 2319–2329

Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

The impact of Collaborative Transportation Management on supply chain

performance: A simulation approach
Felix T.S. Chan a,⇑, T. Zhang b
a
Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hung Hom, Hong Kong
b
Department of Industrial and Manufacturing Systems Engineering, The University of Hong Kong, Pokfulam, Hong Kong

a r t i c l e i n f o a b s t r a c t

Keywords: Collaborative Transportation Management (CTM) is based on the interaction and collaboration between
Collaborative Transportation Management trading partners and carriers participated in the supply chain, appropriate application of CTM can
Performance measurement improve the flexibility in the physical distribution and minimize the inefficiency of supply chain man-
Modeling and simulation agement. This paper proposes new concepts of CTM and carriers’ flexibility. A simulation approach is
Supply chain management
used to (i) evaluate the benefits of the proposed CTM, (ii) explain the concept of carrier’s flexibility,
Carriers’ flexibility
and (iii) optimize the delivery speed capability. Based on a simple supply chain including one retailer
and one carrier, three different simulation models have been developed with changeable delivery lead
time as follows: (1) Unconstrained delivery speed capability without CTM. (2) Constrained delivery
speed capability without CTM; and (3) Constrained delivery speed capability with CTM. Simulation
results reveal that CTM can significantly reduce the retailer’s total costs and improve the retailer’s ser-
vice level.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
Global logistics in business operation has been playing a crit-
ical role in responding to the even changing market demand in
the world of globalization and mass customization. The efficiency
and flexibility of global distribution holds the key to success in
international trade. Collaborative Transportation Management
(CTM) is not only a new collaboration strategy between the ship-
per and carrier, it is also a new business model (Feng & Yuan,
2007).
In recent years, the collaboration among disparate partners
within the supply chain and e-supply chain has been widely dis-
cussed. Interestingly, the transportation and its impact on the en-
tire supply chain have seldom been explored. For instance, two
trading partners in a supply chain generally execute Collaborative
Planning, Forecasting, and Replenishment (CPFR), in order to im-
prove the inventory cost, revenue and service. However, the con-
nection with transportation and distribution management is
often neglected. Consequently, the missing link of transportation
blurs the lines between planning and execution of the supply
chain. The financial and operational performances for the sellers’
and the buyers’, therefore, would be highly affected (Bishop,
2004; Browning & White, 2000).
⇑ Corresponding author. Tel.: +852 2766 6605.
E-mail addresses: mffchan@inet.polyu.edu.hk (F.T.S. Chan), tingzhang930@
yahoo.com.cn (T. Zhang).
Can the replenishment appear at the right time and in the right
place? Often, the order is in the process, but its status is unknown
due to unavailable carrier capabilities or delayed resulted from low
carriers’ flexibility. In a high changing demand market, retailers
must suffer from high backorders of customer’s demand with high
penalty cost. Another consequence is to increase transportation
costs by using secondary carriers, whose contract rates are not as
advantageous as primary carriers. In order to minimize the ineffi-
ciency of transportation caused by insufficient interaction and col-
laboration, trading partners of the supply chain should consider
transportation management as part of the collaboration. Through
the integration and cooperation of the buyer, seller and carrier,
the flexibility and overall value of business chain would be
enhanced.
In this paper, a simple supply chain with stochastic market de-
mand will be developed including one retailer and one carrier.
Three different simulation models allowing changeable delivery
lead time will be built which present three different simulation
scenarios: (1) unconstrained delivery speed capability without
CTM; (2) constrained delivery speed capability without CTM;
and (3) constrained delivery speed capability with CTM. Different
performances of the above situations will be analysis and dis-
cussed, including retailer’s total cost and service level. The three
simulation scenarios in this paper are similar as Feng et al.
(2005), but the simulation models, software used to build the
model and indicators to measure the performance are totally dif-
ferent. The delivery capability changeable in our model is the

0957-4174/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2010.08.020
2320 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329
delivery speed capability which is measured by delivery lead
time; the shorter the lead time, the higher the capability. While,
the models built by Feng et al. (2005) allowed the changeable
delivery amount, the reason of which is they focused on minimiz-
ing the inventory and the holding cost. While, one focus of this
paper is to minimize the penalty cost caused by the backorder
of customer’s demand, therefore the indicators to measure the
performance in our simulation models are the retailer’s total cost,
including penalty cost, inventory holding cost and order cost, and
the retailer’s service level.
Another main contribution of this paper is the proposal of a new
concept of carrier’s flexibility. Simulation models built in this paper
also explain the new concept of carrier’s flexibility which starts
with order/shipment forecasts including capabilities of planning
and scheduling. Briefly speaking, the main idea of carrier’s flexibil-
ity is adjustment of the planned delivery capabilities to match the
changing demand. When the demand exceeds the planned capabil-
ity of the carriers’, carriers can adjust the delivery planning strat-
egy with CTM so that the available delivery capability can meet
the demand. They can even adjust the available delivery capability
to the maximum delivery capability, in order to reduce the gap be-
tween planned delivery capability and available delivery
capability.
This paper consists of five sections. Section 1 is an introduction.
The Section 2 is the review of related literatures. Section 3 is the
development of three models of supply chain with CTM. Section
4 is the analysis of simulation results, and finally Section 5 presents
the conclusion and suggests the future research.
2. Literature review
2.1. CTM
2.1.1. Definition of CTM
According to the Collaborative Transportation Management
White Paper (2004), CTM is defined as a holistic process that brings
together supply chain trading partners and service providers to
drive inefficiencies out of the transport planning and execution
process.
Not only is CTM a new partner strategy between the shipper
and carrier, it is also a new business model. This model includes
the carrier as a strategic partner for information sharing and col-
laboration in the supply chain. The application of CTM promises
to reduce transit times and total costs for the retailer and its sup-
pliers while increasing asset utilization for the carriers. The pro-
grams benefits all three parties involved: the retailer, the
supplier and the carrier (Tyan, Wang, & Du, 2003).
2.1.2. Objective of CTM
The objective of CTM is to improve the operating performance
of all parties involved in the relationship by eliminating inefficien-
cies in the transportation component of the supply chain through
collaboration. Transportation service represents a major compo-
nent of order lead time—the time that elapses from an order place-
ment until the goods are ultimately delivered to a customer. Much
of the variability in order lead time is attributed to variation in
transit times. With more and more companies operating on a
just-in-time basis, there is less room for error in the delivery pro-
cess (CTM White Paper, 2004).
It is important for companies to work together to eliminate
inefficiencies, reduce cost, and ensure excellence in the move-
ment of goods. In order to achieve the positive results of CTM,
the processes between participating companies should be in real
time, extendible, automated and cost-effective (Rabinovich,
2005).
2.1.3. Simulation of CTM
Feng et al. (2005) developed a modified simulation model of the
‘‘beer game”, allowing the changeable delivery capacity. The
supply chain performance indicators that they measured are total
supply chain costs, including inventory costs and backlog costs,
and transportation capacity utilization. Results of the simulation
reveal that CTM can significantly reduce the total supply chain
costs and improve transportation capacity utilization.
The three simulation scenarios presented in this paper are sim-
ilar as Feng et al. (2005), but the simulation models, software used
to build the model, and indicators to measure the performance are
totally different.
(a) The inventory policy considered here is the continuous
review policy or the ‘‘s-S” policy or the fixed quantity policy.
(b) The delivery capability changeable in our proposal models is
the delivery speed capability which is measured by delivery
lead time; the shorter the lead time, the higher the capabil-
ity. While, the delivery amount is fixed.
(c) The indicators to measure the performance in simulation
models are the retailer’s total cost, including inventory hold-
ing cost, penalty cost and order cost, and retailer’s service
level.
(d) One objective in this paper is to minimize the penalty cost
caused by backorder of customer’s demand instead of the
minimizing inventory level.
2.2. Supply chain collaboration
2.2.1. Definition of supply chain collaboration
Supply chain collaboration is prevalent in today’s business
model. An organization not only optimizes itself but also collabora-
tively with other organizations to have larger optimization plan-
ning (Chan, Chung, & Wadhwa, 2004). In order to achieve an
integrative settlement, collaboration has been defined as an at-
tempt to fully satisfy the concerns of the parties involved in ex-
change (Esper & Williams, 2003). The process of collaboration,
pointed by several authors, is the decision making among interde-
pendent parties (Jiang & Jiang, 2005; Koulinitch & Sheremetov,
1998; Kwon & Lee, 2002). It involves joint ownership of decisions
and collective responsibility for outcomes (Stank, Keller, & Daugh-
erty, 2001). The key characteristics of collaboration identified are
coherence, communication, task management, resource manage-
ment, schedule management, and real-time support (Graham,
2006).
Basically, there are three types of collaborations: the horizontal,
vertical and lateral collaborations (Hsu & Hsu, 2009). The type of
collaboration is mainly decided by the collaboration scenario and
the attributes of the participants. Each type of collaborations is de-
fined below:
 Horizontal collaboration: occurs when two or more unrelated or
competing organizations cooperate to share their private infor-
mation or resources, such as joint distribution centers.
 Vertical collaboration: occurs when two or more organizations
such as the manufacturer, the distributor, the carrier and the
retailer share their responsibilities, resources, and performance
information to serve relatively similar end customer.
 Lateral collaboration: aims to gain more flexibility by combin-
ing and sharing capabilities in both vertical and horizontal
manners.
As pointed out by Thomas and Griffin (1996), collaboration is
creating significant value in the relationships along the value
chain. Many studies have also discovered positive impact of strate-
gic alliance between enterprises on their market performance
 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329
(George, Zahra, Wheatley, & Khan, 2001; Park & Cho, 1997; Sant-
oro, Borges, & Rezende, 2006; Sarkar, Echambadi, & Harrison,
2001).
2.2.2. CTM in supply chain collaboration
A single member of the supply chain alone cannot do much to
resolve supply chain problems. This is why collaboration among
partners in a supply chain has become a topic of great interest
for many and an essential element of company strategy for others
(CTM White Paper, 2004). Previous studies on supply chain collab-
oration have focused mainly on the collaboration among supply
chain parties including the suppliers, manufacturers, wholesalers/
distributors and retailers. Such as Chan et al. (2004) and Kim,
Banerjee, and Burto (2008) studied the partnership between sup-
plier and buyer. Elofson and Robinson (2007) studied the Collective
Customer Collaboration (C3) system. As a matter of fact, supply
chain consists of not only customers in downstream flows, but also
third-party organizations, such as logistics and transportation pro-
viders (Esper & Williams, 2003; Mentzer, Foggin, & Golicic, 2000).
Researchers including Browning and White (2000), Esper and Wil-
liams (2003) and Bishop (2004) have all pointed out the need to
incorporate Collaborative Transportation Management (CTM) with
Collaborative Planning, Forecasting and Replenishment (CPFR)
among trading partners in the supply chain. While CPFR is primar-
ily buyer- and seller-based, CTM involves the transportation ser-
vice providers including carriers and 3PLs to ensure efficient and
effective shipment delivery.
2.3. Lead time
2.3.1. Lead time in inventory models
The element of time was present in the earliest inventory mod-
els facing stochastic demand. For example, Karlin and Scarf (1958)
consider a multiple period model with non-zero lead-time. The
treatment of lead-time as a deterministic decision variable within
inventory models began with Liao and Shyu (1991). In their model,
the order quantity is fixed so that lead-time is the only decision
variable. They introduce the concept of crashing cost to stochastic
inventory models, where crashing cost is the cost increase associ-
ated with reducing lead-time. In a series of papers beginning with
Ben-Daya and Raouf (1994), the model introduced by Liao and
Shyu (1991) is extended such that both order quantity and lead-
time are treated as decision variables. More recently, Jang and
Kleinz (2004) studied how to determine the production quantity
and the processing time as to minimize expected customer re-
sponse time and expected plant costs. Rabinovich (2005) improved
consumer direct fulfillment performance in internet retailing by
emergency transshipments and demand dispersion. Babai,
Syntetos, Dallery, and Nikolopoulos (2009) studied a single-stage
and single-item inventory system with non-stationary demand
and lead-time uncertainty.
2.3.2. Time- based competition
One field of previous literature related to this paper is the time-
based competition literature, which examines delivery speed as a
source of competitive advantage. Li (1992) investigated the role
of inventory in time-based competition. Kalai, Kamien, and Rubi-
novitch (1992) studied the effect of processing speed on price
and firm’s performance, such as market share and profit. Lederer
and Li (1997) included scheduling as a strategic variable when cus-
tomers are heterogeneous. More recently, So and Song (1998)
examined competition with delivery-time guarantees. Cachon
and Harker (2002) considered competition between two firms with
price- and time-sensitive demand, and investigate the impact of
outsourcing on competition. All the previous work consider the
2321
role of delivery speed in markets where customers incur delay
costs.
The difference of this paper from previous work can be briefly
presented as below:
(a) The purpose of this research is looking for a long-term rela-
tionship between retailers and carriers to minimize the
retailers’ cost through the two parties’ collaboration, such
as mutual planning of the carrier’s capabilities and the retai-
ler’s inventory level.
(b) The objective of improvement of delivery capability is not
the competition among carriers but to minimize the ineffi-
ciency in the physical distribution process and to improve
the retailer’s performance.
(c) This research is in retailer environment, not in manufactur-
ing environment. It means that we focus on the ‘‘delivery
into the retailer”, not ‘‘delivery out of the factory”. In this
connection, the measurements are about the retailer’s
performance.
2.4. Flexibility
2.4.1. Definition of flexibility
Several attempts have been made in the literature to define,
model, and measure the flexibility with a view to understand
its true nature and its effect on the performance of the manufac-
turing system. Many definitions of flexibility can be found in the
literature. For instance, Carlsson (1989), cited flexibility as: (1)
those attributes of a production technology which accommodate
greater output variation, as the firm’s response to uncertainty,
especially in the form of fluctuations in demand, but also market
imperfections; (2) a property of initial positions as it refers to the
cost, or possibility of moving to various second period positions.
One position is more flexible than another if it makes available
a larger set of future positions at any given level of cost. Upton
(1995) defined flexibility as the ability to change with little pen-
alty in time, effort, cost or performance. More recently, Sushil
(2000) defined flexibility as the exercise of free will or freedom
of choice on the continuum to synthesis the dynamic interplay
of thesis and antithesis in an interactive and innovative manner,
capturing the ambiguity in systems and expanding the continuum
with minimum time and efforts.
Flexibility becomes particularly relevant when supply chain is
considered, which consists of a network of supply, production,
and delivering firms (Christopher, 1992). In this case, many sources
of uncertainty have to be handled, such as market demand, sup-
plier lead-time, product quality, and information delay (Giannoc-
caro, Pontrantrandolfo, & Scozzi, 2003). Flexibility allows
switching of production among different plants and suppliers, so
that management can cope with internal and external variability
(Chen, Egbelu, & Wu, 1994).
Based on the above definitions, it can be summarized that the
role of flexibility in a system is to enable the system to manage
the change (certain or uncertain), in an effective and efficient man-
ner. The change in the environment includes change in both the
internal environment (resource bottlenecks, etc.) and the external
environment (customer preferences, etc.). Effective manner refers
to the extent to which the effect of change has been successfully
managed, and efficiency refers to the time, cost and effort required
to do this (Wadhwa, Rao, & Chan, 2005).
2.4.2. Classification of flexibility
In spite of a large body of literature on flexibility, the major ones
are about manufacturing flexibility, for example, Browne, Dubois,
Rathmill, Sethi, and Stecke (1984) proposed eight flexibility types
to describe the nature of a manufacturing system that is still one
2322 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329
of the most widely used classification of flexibility types. Benjaafar
and Ramakrishnan (1996) describe 19 types of flexibility. More re-
cently, Wadhwa et al. (2005) studied the exact mechanism that en-
ables manufacturing flexibility to reduce the lead time by
simulation experiments.
Another kind of flexibility is about supplier’s flexibility. The
flexibility of supplier for logistics can be referred to as the routing
flexibility at the shop floor level, i.e. the ability of using alternative
routes to move the work-in-process through different resources
offering the same processes (Das & Nagendra, 1997; Ho & Moodie,
1996). According to Garavelli (2003) logistics flexibility of suppli-
ers is then defined as the possibility of shifting the production of
an item (component or final product) to different sites at a given
stage, to reduce the negative impact on SCM performance. Chan,
Bhagwat, and Wadhwa (2009) focused on a similar concept of
the switching production among different flexible suppliers at dif-
ferent stages of production, with the aim of reducing the lead time
and evaluating the impact of information system and suppliers’
physical manufacturing capabilities.
However, not many literature reported on the carriers’ flexibil-
ity. One kind of related work are about flexibility shipment in the
inventory model. The earliest model with emergency shipment
may be built by Barankin (1961). In this paper, a single period
model was developed in which a shipment is received in the begin-
ning of the period and an emergency order is placed at some time
during the period. Khouja (1996) determined the profit maximiz-
ing order quantity for a single period model with an emergency
supply option and shows that this quantity is smaller than the
solution to the newsvendor model. Lau and Lau (1997a) considered
the following extension of the newsvendor model: a customer re-
ceives an order at the beginning of the season and has the oppor-
tunity to place an additional order at some point during the season.
The objective is to determine order quantities for both ordering
opportunities as to maximize expected profit, where stochastic de-
mand is represented by either normal or beta distribution. Lau and
Lau (1997b) considered the same situation with uniform demand
and compare results with and without additional replenishment.
Lau and Lau (1998) extended their work involving normally dis-
tributed demand by determining the optimal reorder point, and
they include the additional dimension of set-up cost. More re-
cently, Rabinovich (2005) improved consumer direct fulfillment
performance in internet retailing by emergency transshipments
and demand dispersion.
Nevertheless, these work are from the view of flexible inventory
planning, few work are about the carrier’s shipment planning, even
some of them offered decision options were without actually refer-
ring to them as a form of flexibility. For example, Lau et al. (2009)
dealt with the problem of optimization of vehicle routing and pro-
posed a multi-objective evolutionary algorithm to solve the multi-
objective optimization problem. Feng et al. (2005) developed a
simulation model to meet stochastic demand of delivery amount
by adjusting delivery planning but not referred the process as car-
rier’s flexibility.
This paper proposed a new concept of carrier’s flexibility and
developed three different simulation models to explain it. The
main idea of the proposed concept of carrier’s flexibility in this pa-
per is adjustment of the planned delivery capabilities to match the
changing demand.
3. The simulation model with CTM
3.1. Description of the problems
In order to reduce the high penalty cost caused by the de-
mand backorders, the retailer has to re-engineer the process of
an enterprise. The focus should be on the collaborative of logis-
tics strategies, sharing information in the supply chain and
improvement of delivery flexibility. When the retailer’s planned
delivery lead time shorter than the planned capability of the car-
riers, carriers can adjust the transportation planning strategy
through CTM so that the available delivery capability can match
the retailer’s demand. They can even adjust the available deliv-
ery capability to the maximum available delivery capability, in
order to reduce the gap between planned delivery lead time
and available delivery lead time. This paper skips the description
of various negotiation details of the partners of CTM and directly
refers to the result of transportation collaboration, because this
paper focuses more on constructing the simulation models with
CTM.
3.2. Simulation logic
Three models will be built to simulate three different situations
as follows:
Model one: unconstrained capability without CTM.
Model two: constrained capability without CTM.
Model three: constrained capability with CTM.
The simulation logic is divided into two sessions: one is demand
generation as show in Fig. 1; the other is order processing and
shipping as presented in Figs. 2–4. The demand generation sessions
are the same in three models, while the differences are presented
in the order processing and shipping.
Fig. 1 shows the demand generation session. The process begins
at each simulation day with the generation of an entity that repre-
sents a customer. Then that customer’s demand is generated. The
model then determines whether the demand can actually be filled.
If there is enough inventory, the demand can be filled and be taken
away from inventory; if not, the demand backorder along with its
penalty cost are recorded. The cost and service level statistics will
be updated accordingly.
Fig. 2 presents the order processing and shipping in model one.
Since continuous review inventory policy is considered here, the
model checks inventory level at each simulation day. Then it com-
pares the inventory level with the reorder point. If the inventory le-
vel is lower than the reorder point, the order processing begins
with a required delivery lead time. There is unconstrained delivery
capability in model one, so the actual delivery lead time (Ti) is the
required delivery lead time (Di). After time Ti shipment, the inven-
tory level is updated.
Fig. 1. Demand logic.
 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329 2323

3.3. Variables

3.3.1. Notation
Di required delivery lead time at cycle i
Ti actual delivery lead time at cycle i
Vi available delivery lead time at cycle I; Vi  unif(u, U)
u shortest delivery lead time, also the lower limit of
available delivery lead time ay cycle i
U the upper limit of available delivery lead time at cycle i
TD(t) total customer demand
FD total immediately filled customer demand
f(t) immediately filled customer demand everyday
I(t) total on-hand inventory
inv(t) on-hand inventory everyday
B(t) total backorder
b(t) backorder everyday
Q replenishment quantity
Q1 order-up-to-level, which refers to ‘‘S” in the ‘‘S-s
policy”, also the initial value of on-hand inventory
Q2 reorder point
S.s safety stock
H(t) total holding cost
Fig. 2. Model one. h holding cost per unit per day
P(t) total penalty cost
p penalty cost per unit per day
O(t) total order cost
fc fixed order set-up cost
o(t) delivery cost per order
C(t) total cost
ô service level
li mean demand in one cycle; li  norm(m, ð12)
m mean value of mean demand per cycle
ð1 standard deviation of mean demand per cycle
k(t) customer demand everyday; kt  norm(li, ð22)
ð2 standard deviation of everyday’s customer demand

3.3.2. Initial values

The initial values of above variables are as follows:

m 10 units
ð1 3 units
ð2 1 unit
Q1: 310 units
Q2: 80 units
S.s 10 units
h $0.025
p $2.5
Fig. 3. Model two. fc $10

3.4. Model formulation

Fig. 3 presents the order processing and shipping in model
two. The part before the order processing begins is similar as 3.4.1. CTM
model one. Since, there is constrained delivery capability in mod-
el two which is presented by constrained delivery lead time avail- Di ¼ IðtÞ=li : ð1Þ
able (Vi), the model has to determine whether the required Refer to Eq. (1), the required delivery lead time (Di) is the on-hand
delivery lead time can be met. If Di is not shorter than Vi, the ac- inventory divides by the mean demand in that cycle (li).
tual delivery lead time Ti will be equal to the requirement Di. On
the other hand, if Di is shorter than Vi, then Ti will be equal to Vi Vi  unif ðu; UÞ: ð2Þ
which is longer than Di. Refer to Eq. (2), the available delivery lead time (Vi) is random gen-
Fig. 4 shows the order processing and shipping in model three. erated between shortest delivery lead time (u) and longest delivery
The part before the order processing begins is still similar as mod- lead time (U) according to the uniform distribution.,
els one and two. In model three, there is constrained delivery capa-
bility with CTM, so the required delivery lead time Di can be met as Ti ¼ Di; if Di=Vi: ð3Þ
long as it is not shorter than the shortest delivery lead time u Ti ¼ Di; if u5Di5Vi: ð4Þ
which presents the maximum delivery capability. Ti ¼ u; if Di < u: ð5Þ
2324 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329

Fig. 4. Model three.

Refer to Eq. (3), the actual delivery lead time (Ti) is equal to the re- Refer to Eq. (9), total immediately filled customer demand
quired delivery lead time (Di) when Di is longer than the available (FD(t)) equals to the cumulation of the part of everyday’s customer
delivery lead time (Vi). The carrier need not adjust the delivery demand which can be met immediately(f(t)).
planning. Z t
Refer to Eq. (4), the Di can still be met even when it is shorter IðtÞ ¼ inv ðtÞ  dt;
than Vi as long as it is longer than the shortest delivery lead time 0
(u). The carrier can shorten the Vi to Di through the CTM.
Refer to Eq. (5), the Ti equals to u when Di is too short. The car-
rier can only adjust the Vi to u. inv ðtÞ ¼ iðtÞ  kðtÞ: ð10Þ
The limit condition in Eqs. (1)–(5) is as below:
Refer to Eq. (10), total on-hand inventory (I(t)) equals to the cumu-
Di; Ti; Vi; U; u = 0: lation of inventory everyday(int(t)).
Ti; Vi = u: In which, inventory everyday(inv(t)) = inventory  demand
Z t
The actual delivery lead time (Ti) and the available delivery lead BðtÞ ¼ bðtÞ  dt;
time (Vi) the carrier can supply cannot be shorter than the shortest 0
delivery lead time (u).
bðtÞ ¼ kðtÞ  iðtÞ: ð11Þ
u 5 Vi 5 U:
Refer to Eq. (11), total backorder(B(t)) equals to the cumulation of
The available delivery lead time in cycle i (Vi) is between the lower backorder everyday(b(t)).
limit (u) and the upper limit (U). In which, backorder everyday(b(t)) = demand-inventory

3.4.2. Demand and inventory Q ¼ Q 1  Q2: ð12Þ

  Refer to Eq. (12), replenishment quantity (Q) is the difference be-
li  norm m; ð12 : ð6Þ
tween the order-up-to-level of inventory (Q1)and the reorder point
(Q2).
Refer to Eq. (6), the mean demand in one cycle (li) is a stochastic
variable and follows the normal distribution with the mean of m
and the standard deviation of ð1. 3.4.3. Cost structure
  
HðtÞ ¼ h IðtÞ: ð13Þ
kt  norm li ; ð22 : ð7Þ 
PðtÞ ¼ p BðtÞ: ð14Þ
Refer to Eq. (7), the customer demand everyday in one cycle (kt) fol- OðtÞ ¼ fc þ oðtÞ: ð15Þ
lows the normal distribution with the mean of li and the standard
deviation of ð2. In which o(t) = k/Ti.
Refer to Eq. (13), Holding cost(H(t)) equals to holding cost per
Z t
unit per day(h)*accumulative on-hand inventory(I(t)).
TDðtÞ ¼ kðtÞ  dt: ð8Þ
0 Refer to Eq. (14), Penalty cost(P(t)) equals to penalty cost per
Z t unit per day(p)*accumulative demand backorder(B(t)).
FDðtÞ ¼ f ðtÞ  dt: ð9Þ Refer to Eq. (15), Order cost(O(t)) equals to fixed order set-up
0
cost(fc) plus delivery cost(o(t)).
Refer to Eq. (8), total customer demand (TD(t)) equals to the cumu- In which, delivery cost = k/actual delivery lead time, and k is a
lation of customer demand everyday (k(t)). constant parameter.
 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329
3.5. Supply chain performance measurements
Two performance indicators as below:
CðtÞ ¼ HðtÞ þ PðtÞ þ OðtÞ; ð16Þ
^ ¼ FDðtÞ=TDðtÞ:
o ð17Þ
Refer to Eq. (16), Total cost(C(t)) equals to the sum of holding cost,
penalty cost and order cost.
Refer to Eq. (17), Service level(ô) equals to filled customer de-
mand/ total customer demand.
3.6. Assumptions
(a) To facilitate the control of simulation model, the length of
one simulation cycle (i) is one month which equals to 30
simulation days.
(b) From test the simulation steady state by changing the simu-
lation length in certain initial conditions and random num-
bers, we observe the simulation results and found that the
model reaches steady-state after three simulation years,
therefore, we set the total simulation length ten years, which
equals to 3600 simulation days, to guarantee the model run
for a enough long time.
(c) There is only one delivery demand in one simulation cycle
and there is no delivery backorder. This assumption is simi-
lar to one made by Hadley and Whitin (1963, p. 162), in their
analysis of the single lead time case.
(d) Lead time has two parts—order lead time and delivery lead
time. To simplify the model, we assume order lead time is
zero, so the lead time in our paper refers to the delivery lead
time. The main factor makes impact on the delivery lead
time is the delivery speed capability of the carrier.
(e) Order cost is the sum of fixed set-up order cost which is con-
stant and delivery cost which is the expression of actual
delivery lead time. Therefore, the order cost is only affected
by one decision variable, that is the delivery lead time. In
fact, the shorter the delivery lead time, the higher the order
cost.
4. Results analysis
4.1. Comparison of the three models
The results of simulation are shown in Table 1.
The service level in model one (0.98393) is higher than both
model two (0.87984) and model three (0.94393). The reason is in
model one, there is unconstrained capability, i.e. the required
delivery lead time is the actual delivery lead time, the backorder
level is low. But the service level isn’t one hundred percents in
model one because of the deviation of customer demand.
The total cost in model one (26,735) is lower than model two
(32,890), because of the low penalty cost, but higher than model
three (26,364), because of the high order cost. It indicates that
it’s possible to trade off among the major cost components and find
the best solution of total cost.
Table 1
Comparison of the three models.
Model One Two Three
Capability limit No Yes Yes
CTM No No Yes
Total cost 26,735 32,890 26,364
Service level 0.98393 0.87984 0.94393
2325
There is constrained capability in both the models two and
three, but model three is with CTM. The delivery forecasting infor-
mation is shared so that the available delivery capability can be ad-
justed to the maximum delivery capability, which will reduce the
demand backorders, resulting in a low penalty cost. Hence, the to-
tal cost in model three is lower than that in model two.
4.2. Improvements of CTM
In the following section, constrained capability scenario will be
discussed to study how CTM improves the total cost and service le-
vel along with the time.
Fig. 5 indicates the total cost with CTM is lower than that with-
out CTM. Moreover, the gap is wider along with the time.
Fig. 6 presents the service level with CTM is obviously higher
than that without CTM. Furthermore, both of the two service levels
converge finally (0.94393 with CTM and 0.87984 without CTM).
One point that should be noticed here is the two curves reach
the steady-state around 1000 simulation days, however, the gap
between the two situations is still wide even before the curves
reach the steady-state.
4.3. Sensitivity analysis
The slope of a curve refers the dependent variable’s sensitivity
to independent variable. For example, when we set the dependent
variable is total cost and the independent variable is delivery capa-
bility, the steep curve indicates the change of delivery capability
affects heavily on the total cost, i.e. the total cost is very sensitive
to delivery speed capability.
In the following section, we study the sensitivity of total cost
and service level to three parameters—the maximum level of deliv-
ery speed capability, the penalty cost per unit per day and the stan-
dard deviation of customer demand per cycle, in order to study
how the CTM improve the supply chain performance when these
important parameters change.
4.3.1. Change the maximum level of delivery speed capability
We will measure delivery speed capability by delivery lead
time, i.e. shorter lead time means higher capability. We change
the shortest delivery lead time but the longest one is infinite. In
the situation with CTM, it means that the carrier can meet the re-
quired delivery lead time when it is not shorter than the shortest
delivery lead time. While in the situation without CTM, it means
that the carrier’s available delivery lead time changes among the
shortest delivery lead time and a fixed delivery lead time, 10 days
in this case, according to the uniform distribution. When we adjust
the penalty cost per unit per day, we fix the holding cost per unit
per day.
Fig. 7 indicates the total cost increases regardless of application
of CTM or not. However, it is obviously improved when CTM is
adopted, especially in the middle stage. But in the last stage which
means the delivery capability is too low, the total cost of two
curves converge. Another point we should notice is the total cost
increases with the delivery lead time all the time if there is no
CTM. While, if there is CTM, the total cost reaches a bottom around
six days, which means we can get a ‘‘best solution” in some deliv-
ery lead time. In this case, the best solution is the delivery lead
time of six days in which the total cost is 25,642, saving 21% than
31,052 in the situation without CTM.
Fig. 8 indicates the service level is obviously improved when
CTM is adopted, especially in the middle stage. Similar to Fig. 7,
in the last stage, the curves converge.
2326 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329

Fig. 5. Change of the total cost with.

Fig. 6. Change of the service level with time.

4.3.2. Change penalty cost per unit per day the increase is much higher than the situation with CTM which
Fig. 9 indicates the total cost in both situations increase when indicates the total cost without CTM is more sensitive to penalty
the penalty cost increase. While, in the situation without CTM, cost. It suggested that the CTM will improve the total cost
 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329
Fig. 7. Sensitivity of total cost to delivery lead time.
Fig. 8. Sensitivity of service level to delivery lead time.
Fig. 9. Sensitivity of total cost to penalty cost.
obviously when the penalty cost is high, for example, the penalty
cost can be very high in a high competition environment; if we
cannot meet the customer’s need once, we will lose the customer
for ever.
4.3.3. Change standard deviation of customer demand per cycle
Fig. 10 indicates the total cost increases with the deviation of
customer demand per cycle. While, similar to Fig. 9, the increase
is quicker in the situation without CTM. It suggests that CTM will
obviously improve the total cost when there is a high unstable de-
mand, for example, the demand fluctuates heavily due to the sea-
sonal effect in the market.
2327
Fig. 10. Sensitivity of total cost to deviation.
4.4. Optimum delivery speed capability in CTM model
High delivery capabilities cost a lot for carriers, and it is a huge
waste if they are not necessary, while, the shipment planning and
demand will be hardly met if the delivery capabilities are too low,
and the delivery backorders will also result in huge cost for carri-
ers. In the past, the planning of proper delivery capabilities is very
difficult resulted from lack of communication and information
sharing between carriers and retailers, while, the collaboration of
the two parties facilitates the planning work, and the seeking of
optimum delivery capability becomes possible.
Now, let us examine the question what’s the optimum delivery
speed capability the carriers should supply in CTM scenarios. This
question can be addressed under two fields, one is to analyze how
the major cost components change with delivery speed capability.
Those cost components changing significantly are the ones sensi-
tive to delivery speed capability, hence, it is possible to get the
optimum delivery speed capability when these cost components
are tightly controlled. The other field in this section is to observe
the total cost’s sensitivity to delivery speed capability in different
situations with various penalty cost per unit per day and various
standard deviation of customer demand per cycle. The optimum
delivery speed capability is different in different cost structure
and demand environments, hence, the carriers should make the
optimum planning according to the real world.
4.4.1. Sensitivity of major cost components
Two important points will be considered—one is how the
amount of the major cost components change with delivery speed
capability, the other is how their sensitivity change. It should be
noted that we measure the delivery capability with delivery lead
time.
Fig. 11 indicates the holding cost decreases with the delivery
lead time. The reason is the replenishment amount is fixed, so
the inventory level decreases when the order cycle is long.
The order cost also decreases. The expression of order cost tells
us the same information.
The penalty cost increases sharply which indicates it is very
sensitive to the delivery speed capability.
4.4.2. Sensitivity of total cost in various penalty cost
In the following part, how the total cost’s amount and sensitiv-
ity to delivery lead time change in different situations with various
penalty cost, will be discussed.
Fig. 12 presents how the total cost changes with the delivery
lead time in three situations—the penalty cost per unit per day in-
creases to 100%, 200% and 300% of the initial value. It can be ob-
served the total cost in each curve almost keeps the same when
2328 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329
Fig. 11. Sensitivity of major cost components to delivery lead time.
Fig. 12. Change of total cost with delivery lead time when penalty cost differs.
the delivery lead time is before six days, which means the total cost
is not so sensitive to delivery lead time. It suggests that merely
from the view of penalty cost, to decrease delivery lead time to less
than six days does not help a lot to save total cost, moreover, it will
be a waste. While, it can also be observed that the total cost in-
creases sharply after seven days in all the three curves, further-
more, the gaps among the three curves become wider along with
delivery lead time, which means the total cost is more sensitive
to delivery lead time when the penalty cost increases. Hence, it
suggests that to improve the delivery capability is very important
when the available delivery lead time is longer than seven days
and the improvement is more pressing when the penalty cost is
high.
Fig. 13. Change of total cost with delivery lead time when deviation differs.
4.4.3. Sensitivity of total cost in various customer demand deviation
In the following part, the impact of the standard deviation of
customer demand per cycle will be discussed.
Fig. 13 presents how the total cost changes with the delivery
lead time in three situations – the standard deviation of cus-
tomer demand per cycle is one, two and three, respectively. It
can be observed that the total cost increases with the deviation
and the lowest total cost reaches at an earlier point (day eight
when deviation is one; day seven when deviation is two; day
six when deviation is three) which means the optimum delivery
lead time should be less when the deviation increases. It sug-
gests that to improve the delivery capability is very important
when the demand is unstable.
5. Conclusions and suggestion of future research
CTM is based on the collaboration between trading partners
and carriers of a supply chain in order to minimize the ineffi-
ciency of physical distribution and improve the flexibility of the
supply chain. Currently, more and more partners in the supply
chain operate on a just-in-time basis. With the expectation of
shortening planning cycles and minimizing demand backorders,
transportation efficiency becomes one of the crucial factors for
efficient supply chain management. It is important that business
partners cooperate to strengthen communication, share informa-
tion, and ensure the efficiency of physical delivery. That is why
collaboration among partners in a supply chain has become a to-
pic of great interest for many and an essential element of com-
pany strategy.
The purpose of this research is looking for a long-term relation-
ship between retailer and carrier to minimize the retailer’s total
cost through the two parties’ collaboration. The objective of opti-
mization of the delivery capabilities planning is to improve the re-
tailer’s performance.
This paper proposes new concepts of CTM and carriers’ flexibil-
ity. A simulation approach is used to evaluate the benefits of the
proposed CTM, explain the concept of carrier’s flexibility and opti-
mize the delivery speed capability. Results of the simulation reveal
that CTM can significantly reduce the retailer’s total costs and im-
prove its service level. Not as many previous papers in which the
holding cost made a heavy impact on the total cost, the perfor-
mance indicators in our models are sensitivity to the penalty cost,
the reason of which is our models are based on a high changing de-
mand market.
Simulation models built in this paper also explain the new con-
cept of carrier’s flexibility which starts with order/shipment fore-
casts including capabilities planning and scheduling. Briefly
speaking, the main idea of carrier’s flexibility is adjustment of
the planned delivery capabilities to match the changing retailer’s
shipment demand. The simulation results demonstrated that the
mutual planning of the carrier’s capabilities with CTM can improve
carrier’s flexibility.
The future research is suggested in the following aspects:
(a) Development of a multi-echelon supply chain with the man-
ufactures, distribution centers, carriers, and retailers verti-
cally and more than two competitors horizontally.
(b) The performance indicators are related to each partner in the
supply chain, respectively, and the whole supply chain as
well.
(c) In the horizontal supply chain, some competition issues
should be researched, such as whether the retailer or the
carrier has the right to decide the order cost, application
of CTM as the completion advantage between two carriers,
etc.
 F.T.S. Chan, T. Zhang / Expert Systems with Applications 38 (2011) 2319–2329 2329

References Kalai, E., Kamien, M., & Rubinovitch, M. (1992). Optimal service speeds in a
competitive environment. Management Science, 38, 1154–1163.
Karlin, S., & Scarf, H. (1958). Inventory models of the Arrow–Harris–Marschak type
Babai, M. Z., Syntetos, A. A., Dallery, Y., & Nikolopoulos, K. (2009). Dynamic re-order
with time lag. In K. J. Arrow, S. Karlin, & H. Scarf (Eds.). Studies in the
point inventory control with lead-time uncertainty: Analysis and empirical
mathematical theory of inventory and production (pp. 155–178). Stanford, CA:
investigation. International Journal of Production Research, 47(9), 2461–2483.
Stanford University Press (Chapter 10).
Barankin, E. W. (1961). A delivery-lag inventory model with emergency provision.
Khouja, M. (1996). A note on the newsboy problem with emergency supply option.
Naval Research Logistics Quarterly, 8, 285–311.
Journal of the Operational Research Society, 47, 1530–1534.
Ben-Daya, M., & Raouf, A. (1994). Inventory models involving lead-time as a
Kim, S. L., Banerjee, A., & Burto, J. (2008). Production and delivery policies for
decision variable. Journal of the Operational Research Society, 45, 579–582.
enhanced supply chain partnerships. International Journal of Production
Benjaafar, S., & Ramakrishnan, R. (1996). Modelling measurement and evaluation of
Research, 46(22), 6207–6229.
sequencing flexibility in manufacturing system. International Journal of
Koulinitch, A. S., & Sheremetov, L. B. (1998). Coordination and communication
Production Research, 34, 1195–1220.
issues in multi-agent expert system: Concurrent configuration design advisor.
Bishop, S. B. (2004). Collaborative transportation management benefits. Supply Chain
Expert Systems with Applications, 15, 295–307.
Systems, Inc., <http://www.vics.org>.
Kwon, O. B., & Lee, K. C. (2002). MACE: Multi-agents coordination engine to resolve
Browne, J., Dubois, D., Rathmill, K., Sethi, P., & Stecke, K. E. (1984). Classification of
conflicts among functional units in an enterprise. Expert Systems with
flexible manufacturing systems. FMS Magazine, 114–117.
Applications, 23, 9–21.
Browning, B., & White, A. (2000). Collaborative transportation management – A
Lau, H. C. W., Chan, T. M., Tsui, W. T., Chan, F. T. S., Ho, G. T. S., & Choy, K. L. (2009). A
proposal. Logility Inc., <http://www.vics.org>.
fuzzy guided multi-objective evolutionary algorithm model for solving
Cachon, P., & Harker (2002). Competition and outsourcing with scale economies.
transportation problem. Expert Systems with Applications, 36, 8255–8268.
Management Science, 48, 1314–1333.
Lau, A., & Lau, H. (1997a). Reordering strategies for a newsboy-type product.
Carlsson, B. (1989). Flexibility and the theory of the firm. International Journal of
European Journal of Operational Research, 103, 557–572.
Industrial Organization, 7, 179–203.
Lau, A., & Lau, H. (1997b). A semi-analytic solution for a newsboy problem with
Chan, F. T. S., Bhagwat, R., & Wadhwa, S. (2009). Study on suppliers’ flexibility in
mid-period replenishment. Journal of the Operational Research Society, 48,
supply chains: Is real-time control necessary. International Journal of Production
1245–1253.
Research, 47(4), 965–987.
Lau, A., & Lau, H. (1998). Decision models for single period products with two
Chan, F. T. S., Chung, S. H., & Wadhwa, S. (2004). A heuristic methodology for order
ordering opportunities. International Journal of Production Economics, 55,
distribution in a demand driven collaborative supply chain. International Journal
57–70.
of Production Research, 42(1), 1–19.
Lederer, P., & Li, L. (1997). Pricing, production, scheduling, and delivery-time
Chen, C. F., Egbelu, P. J., & Wu, C. T. (1994). Production planning models for a central
competition. Operational Research, 45, 407–420.
factory with multiple satellite factories. International Journal of Production
Li, L. (1992). The role of inventory in delivery-time competition. Management
Research, 32(6), 1431–1450.
Science, 38, 182–197.
Christopher, M. (1992). Logistic and supply chain management. London: Pitman.
Liao, C., & Shyu, C. (1991). Stochastic inventory model with controllable lead-time.
Collaborative Transportation Management White Paper, Version 1.0 (2004). CTM
International Journal of Systems Science, 22, 2347–2354.
sub-committee of the voluntary inter-industry commerce standards (VICS) logistic
Lodree, E., Jr., Jang, W., & Kleinz, C. M. (2004). Minimizing response time in a two-
committee. <http://www.vics.org>.
stage supply chain system with variable lead time and stochastic demand.
Das, S. K., & Nagendra, P. (1997). Selection of routes in a flexible manufacturing
International Journal of Production Research, 42(11), 2263–2278.
facility. International Journal of Production Economics, 48, 237–247.
Mentzer, J. T., Foggin, J., & Golicic, S. (2000). Collaboration: The enablers,
Elofson, G., & Robinson, W. N. (2007). Collective customer collaboration impacts on
impediments, and benefits. Supply Chain Management Review(September/
supply-chain performance. International Journal of Production Research, 45(11),
October), 52–58.
2567–2594.
Park, N. K., & Cho, D. S. (1997). The effect of strategic alliance on performance: A
Esper, T. L., & Williams, L. R. (2003). The value of collaborative transportation
study of international airline industry. Journal of Air Transport Management, 3(3),
management (CTM): Its relationship to CPFR and information technology.
155–164.
Transportation Journal, 42(4), 55–65.
Rabinovich, E. (2005). Consumer direct fulfillment performance in Internet
Feng, C. M., & Yuan, C. Y. (2007). Application of collaborative transportation
retailing: Emergency transshipments and demand dispersion. Journal of
management to global logistics: An interview case study. International Journal of
Business Logistics, 26(1), 79–113.
Management, 24(4), 623–638.
Santoro, F. M., Borges, M. R. S., & Rezende, E. A. (2006). Collaboration and
Feng, C. M., Yuan, C. Y., & Lin, Y. C. (2005). The system framework for evaluating the
knowledge sharing in network organizations. Expert Systems with Applications,
effect of collaborative transportation management on supply chain. Journal of
31, 715–727.
the Eastern Asia Society for Transportation Studies, 6, 2837–2851.
Sarkar, M., Echambadi, R., & Harrison, J. S. (2001). Alliance entrepreneurship and
Garavelli, A. C. (2003). Flexibility configurations for the supply chain management.
firm market performance. Strategic Management Journal, 22, 701–711.
International Journal of Production Economics, 85, 141–153.
So, K., & Song, J. S. (1998). Price, delivery time guarantees and capacity selection.
George, G., Zahra, S. A., Wheatley, K. K., & Khan, R. (2001). The effects of alliance
European Journal of Operational Research, 111, 28–49.
portfolio characteristics and absorptive capacity on performance: A study of
Stank, P. T., Keller, S. B., & Daugherty, P. J. (2001). Supply chain collaboration and
biotechnology firms. The Journal of High Technology Management Research, 12(2),
logistics service performance. Journal of Business Logistics, 22(1), 29–48.
205–226.
Sushil (2000). Concept of systemic flexibility. Global Journal of Flexible Systems
Giannoccaro, I., Pontrantrandolfo, P., & Scozzi, B. (2003). Uncertainty in supply chain
Management, 1, 77–80.
inventory management: A fuzzy approach. European Journal of Operational
Thomas, D. J., & Griffin, P. M. (1996). Coordinated supply chain management.
Research, 149, 185–196.
European Journal of Operational Research, 94, 1–15.
Graham, C. (2006). Agent co-ordination aided distributed computational
Tyan, J. C., Wang, F. K., & Du, T. (2003). Applying collaborative transportation
engineering design. Expert Systems with Applications, 31, 776–786.
management models in global third-party logistics. International Journal of
Hadley, G., & Whitin, T. (1963). Prentice Hall international series in management. In
Computer Integrated Manufacturing, 16(4–5), 283–291.
Analysis of inventory systems. Englewood Cliffs, N.J.: Prentice-Hall.
Upton, D. M. (1995). Flexibility as process mobility: The management of plant
Ho, Y. C., & Moodie, C. L. (1996). Solving cell formation problems in a manufacturing
capabilities for quick response manufacturing. Journal of Operational Research,
environment with flexible processing and routing capabilities. International
12, 205–224.
Journal of Production Research, 34(10), 2901–2923.
Wadhwa, S., Rao, K. S., & Chan, F. T. S. (2005). Flexibility-enabled lead-time
Hsu, H.-P., & Hsu, H.-M. (2009). Systematic modeling and implementation of a
reduction in flexible systems. International Journal of Production Research, 43,
resource planning system for virtual enterprise by predicate/transition net.
3131–3162.
Expert Systems with Applications, 35, 1841–1857.
Jiang, Y. C., & Jiang, J. C. (2005). A multi-agent coordination model for the variation
of underlying network topology. Expert Systems with Applications, 29, 372–382.