Transportation Research Part E 41 (2005) 567–583

www.elsevier.com/locate/tre

Towards collaborative, intermodal hub networks

A case study in the fast moving consumer goods market
a,b,* a,c a
Bas Groothedde , Cees Ruijgrok , Lóri Tavasszy
a
TNO Environment and Geosciences, Van Mourik Broekmanweg 6, P.O. Box 49, 2600 AA Delft, The Netherlands
b
Delft University of Technology, Faculty of Civil Engineering and Geosciences, The Netherlands
c
Tias Business School, Tilburg University, The Netherlands

Abstract

Collaborative hub networks can provide an answer to the need to decrease logistics cost and maintain
logistics service levels by shifting consolidated flows to modes that are better suited for handling large vol-
umes (rail, barge, coastal shipping), so economies of scale can be obtained. This necessity has been
increased by the tendency of globalization of industries, smaller shipments sizes, high frequencies, and
the fragmentation of flows. Through collaboration the necessary synchronization between expensive but
fast and flexible means of transport and inexpensive, but slow and inflexible means can be combined in
an intermodal hub network. This paper shows the rationale behind these collaborative hub networks, based
on the literature on the design of many-to-many hub networks. The resulting methodology is explained
through presenting the results of the design and implementation of collaborative hub network for the dis-
tribution of fast moving consumer goods using a combination of trucking and inland barges. This concept,
first proposed by Vermunt [Vermunt, A.J.M., 1999. Multilognet, the intelligent multimodal logistics net-
work, an important node in the worldwide logistics net, Vermunt Logistiek Advies v.o.f., working paper
(in Dutch)], won the European Intermodal Award of the European Intermodal Association in 2003, and
after extensive research was launched in The Netherlands as a commercial pilot by logistics service provider
Vos Logistics and barge operator Riverhopper in January 2004.

2005 Elsevier Ltd. All rights reserved.

*
Corresponding author. Address: TNO Environment and Geosciences, Van Mourik Broekmanweg 6, P.O. Box 49,
2600 AA Delft, The Netherlands.
E-mail address: [email protected] (B. Groothedde).

1366-5545/$ - see front matter
2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tre.2005.06.005
568 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

Keywords: Hub network design; Collaboration; Intermodal transportation; Network development; Parallel transpor-
tation

1. Introduction

During the last thirty years, the imperatives of customer service and cost efficiency have pushed
firms to change their strategy and logistics organization, resulting in centralization of production
and distribution, reduction of inventory and time based competition. Although, for many compa-
nies these changes in strategy have been a part of a broader response to growing global opportu-
nities and increased levels of international competition, they have led to increased competition
and marketing pressure, dictating shorter lifecycles and order cycle times, a larger variety of prod-
ucts and production in smaller quantities (Muilerman, 2001). The evolution of logistics networks
during this period can be characterized by a strong rationalization of business processes. This
ongoing rationalization has led to a constant search for economies of scale and scope in the supply
chain, which has been an important parallel development in line with the changes in international
competition and manufacturing. For example, the logistics costs in Europe as percentage of sales
have dropped 40% from 1981 to 2001 (Piper Jaffray, 2002). However, cost efficiency and reliability
are under pressure due to increasing demands, congestion and vehicle-restrictions. The problems
and inefficiencies in todayÕs road transportation make it necessary to look for alternatives in this
sector. One such alternative could be formed by a collaborative hub network, in which inland
shipping (barge transportation) is used for the inter-hub transportation and economies of scale
are achieved, and parallel to this hub network, direct trucking is used to maintain responsiveness
and flexibility. The collaboration between shippers and the carriers in these networks is essential
in order to guarantee the synchronized organization of the network.
In the project Feasibility of Inland Shipping Networks, the problem at hand was the design of
such a hub network using barges for the transportation of palletized fast moving consumer goods
(FMCG), with the aim to establish economies of scale and scope through collaboration (TNO
Inro, 2001, 2002, 2003). Retail organizations like Albert Heijn, Schuitema and Laurus and manu-
facturers like Heineken, Interbrew, Grolsch, Unilever, Coca Cola, Sara Lee and Kimberly-Clark
participated in the project that started in 2000. The objective of the project was to develop and
implement a hub network in which economies of scale and a sufficient level of reliability was
reached through collaboration. In Section 2 of the paper we discuss the concept of a collaborative
hub network and illustrate the concept of parallel intermodal transport to achieve economies of
scale by the use of inland shipping and maintain responsiveness by the use of trucking. In Section
3 we focus on the retail sector in The Netherlands and in Section 4 we discuss the logistics costs and
economies of scale present in the hub network. In Section 5 we present the results of the project and
our findings concerning the important mechanisms in these collaborative logistics networks are dis-
cussed in Section 6. Finally the paper is concluded with the implications for research and policy.

2. The concept of an intermodal hub network

The concept to consolidate palletized flows between manufacturers and retailers, using inland
barges and form a collaborative hub network, was first proposed by Vermunt (1999). Until then,
 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 569

the initiatives for pallet-transport by inland navigation comprised dedicated transportation for
only one participant. Examples of previous initiatives in the Netherlands were the transportation
by barge of palletized dairy products of Unilever in 1993, soft drinks of Coca Cola in 1995 and
beer products of Bavaria in 1999. All these initiatives failed, for a number of reasons, but in es-
sence these companies could not achieve the necessary scale on their own to operate a cost efficient
intermodal network.
The case study, described in this paper, aimed at the development of a hub network in which
the necessary scale is achieved by combining the flows of different shippers and in which relatively
small dedicated pallet barges follow a high frequency schedule, with fast and cost efficient trans-
shipment on the hubs connecting the inland barges with road transport. These barges are
equipped with fully automated pallet positioning and handling systems, minimizing the transship-
ment time and costs on the hub and maximizing the annual shipments.
In Fig. 1 two situations are illustrated: the current situation, in which product flows between
manufacturing locations and retail warehouses are transported by truck. The problem is that,
due to the demands on frequency, drop size and reliability the cost-efficiency is under pressure.
To counter this development the product flows of different shippers can be consolidated and
shipped through a collaborative hub network, in which economies of scale can be achieved.
The extra costs that are introduced because the products are sent through the hub network for
extra handling, transportation to and from the hub, should be compensated by the economies
of scale of the inter-hub transportation.
In Fig. 2 the concept is illustrated in more detail. In this figure two alternatives are illustrated;
the first is direct transportation from manufacturer to the retail warehouse; the second alternative
consists of shipment through the hub network making use of barges. When making use of the hub
network the products are shipped from the manufacturer to the nearest hub by truck (A); the

Fig. 1. Fragmented flows form manufacturer to retail warehouse and consolidation of flows using a hub network.
570 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

Fig. 2. The concept of an intermodal hub network.

pallets are then transshipped on a dedicated pallet barge and transported from the hub near the
origin to the hub near the destination (B); unloaded and finally shipped by truck to the final des-
tination (C). As these barges have a capacity of 600 pallets (about 20 truckloads) economies of
scale can be achieved on this segment of the intermodal route. It is however essential to combine
this intermodal route with direct trucking that provides logistic services for short distances and
excess demand that cannot be accommodated through the hub network (D). In other words: in
this concept the combination is made between the capacity of inland barges and the responsive-
ness and flexibility of road transport; economies of scale and scope can thus be guaranteed. The

Fig. 3. Principle of parallel intermodal by peak shaving.

 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 571

two modes operate in parallel and make it possible to accommodate the large and predictable vol-
umes and the peaks in demand.
Because the shipment time increases considerably when barges are used the order lead-time will
usually be exceeded if the order is shipped after it is received. On average it takes 10–15 h to cover
150 km using these barges. To cover this same distance by truck would take approximately 2 h.
This gap (t4–t3, depicted in Fig. 3) between the shipment time via the hub network (t4–t1) and
the order lead-time (t3–t1) forms a great problem, especially in the retail where lead-times of 3 h
are no exception. To avoid this problem, the shipment should be sent in advance, before the order
is placed. The part that can be well forecasted can be sent through the hub network in advance, the
unpredictable part of demand can be accommodated by direct trucking (see Fig. 3).

3. The fast moving consumer goods market in The Netherlands

Retailers in The Netherlands are being faced with many significant changes today; increased
competition is creating pressure on retailers to simultaneously control cost and improve customer
service. The market for this segment of products, dominated by a relative small number of retail-
ers in The Netherlands, has great potential for the start of a collaborative hub network. First of
all, high volumes and frequent shipments make this segment very attractive for a hub network.
Second, as the number of retailers is limited, the flows can be consolidated relatively easily.
And finally, the problems in todayÕs road transportation concerning reliability and costs are
becoming ever more prominent making the chances for an alternative mode of transport more
favorable. In addition to these three drivers, not only the limited number of retailers plays an
important role, upstream in the market, the number of manufacturers in The Netherlands is re-
duced significantly due to mergers and acquisitions and more and more product categories are
dominated by only a few companies.
In order to comprehend the possibilities of a collaborative hub network an extensive market
analysis was first performed. This analysis, which examined the Dutch retailers and manufacturers
in the fast moving consumer goods market, resulted in an estimate of the potential market, map-
ping the production, incoming product flows of the warehouses and pallet transportation, broken
down into 28 product categories. Together, these manufacturers and retailers transport a volume of
26.6 million pallets annually between the mentioned 250 production locations and 210 distribution
centers (TNO Inro, 2003). In total, over 95 companies were involved in the project and the infor-
mation provided by the manufacturers and retailers made it possible to make a relatively accurate
estimate of the pallet flows between the manufacturing location and the retail distribution centers.

4. Quantifying economies of scale in hub networks

The hub network is designed for many-to-many distribution problems (Daganzo, 1999; OÕKelly
and Miller, 1994) in order to reduce logistics costs and as OÕKelly and Bryan (1998) so aptly put it:
economies of scale, due to the amalgation of flows, provide a raison dÕetre for hub systems. Consol-
idation in these networks allow more efficient and more frequent shipping by concentrating large
flows onto relatively few links between hubs. Although use of indirect (that is via a hub) shipments
572 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

may increase the distance traveled and extra handling increases the costs, the economies of scale
due to the larger volume shipments can reduce total cost. These configurations reduce and sim-
plify network constructions costs, centralize commodity handling and sorting and allow carriers
to take advantage of economies of scale (Campbell et al., 2002; Eberly et al., 2000). According to
Horner and OÕKelly (2001) a cost function ought to be responsive to flow, and should reward the
volume on links by giving them a lower rate. By simplifying inter-hub, hub operational, and han-
dling costs and assuming that these costs are independent of flows, the current models may not
only miscalculate the total network costs, but may also erroneously select optimal hub locations
and allocations. In this paper we incorporate the cost functions suggested by OÕKelly and Bryan
(1998) for inter-hub flows but in addition the economies of scale on the hub nodes themselves are
included by making the costs functions on the hubs responsive to flow (e.g. handling, transship-
ment), which is especially important in intermodal hub networks, where transshipment costs are a
dominant cost driver.

4.1. The design problem

The hub network design problem, in its general form, involves: (1) finding the optimal locations
for the hub facilities; (2) assigning non-hub origins and destinations to the hubs; (3) determining
linkages between the hubs; and, (4) routing flows through the network. The standard hub network
can be defined as the product of three simplifying restrictions: (1) all hubs are fully interconnected;
(2) all nodes are assigned to one hub; and, (3) there are no direct connections between non-hub
nodes (Fig. 4).
When optimal shipment sizes are used on all links, the cost on each link is a concave function of
flow, and the least cost strategy is always ship all parts direct or ship all parts via the hub network
(Blumenfeld et al., 1985; Newell, 1980). However, when there are restrictions (capacity, frequency
and shipment size) on the inter-hub links connections this all or nothing principle can not be used.
In the network design problem discussed in this paper we relax all three restrictions; (1) hubs are
not fully interconnected; (2) nodes can be assigned to more than one hub; and, (3) direct connec-
tions between non-hub nodes are allowed. This network is called a Protocol H hub network
according to the classification system presented by OÕKelly and Miller (1994).

Fig. 4. Considered hub network (Protocol H network).

 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 573

4.2. Cost functions

In addition to the product flows between the different firms, an important aspect of the project
was to gain insight into the logistics costs of pallet transportation via the hub network and compare
these costs with the current situation (e.g. direct trucking). The current rates were provided by sev-
eral participants, making it possible to compare and calibrate the cost functions used in the project.
In the design problem discussed in this paper we distinguish costs for logistics activities like, han-
dling, loading, unloading, inter-nodal transport, inter-hub transport, ordering and inventory costs.
We consider a situation in which there are i = 1, . . . , I origins. Each origin can transport its
product to each of J destination indexed j = 1, . . . , J. The volume of demand for product i by des-
tination j is denoted dij (e.g. the origins i represents the manufacturing locations, the destinations j
represents the retail warehouses and dij the flows between both locations). The cost of shipping dij
units directly from origin i to destination j is denoted cij. As an alternative to direct shipment there
are k = 1, . . . , K hubs or terminals of which a = 1, . . . , A sequences can be made from the hub near
the origin to the hub near the destination. So a
k forms a sequence of hubs k = 1, . . . , K sub-
jected to a = hk1, . . . , kn, k1i. This restriction was included to accommodate the returns and pack-
ing and to simplify the implementation process: the first hub in the sequence must also be the last
hub in the sequence completing the cycle.
The cost of shipping between the hubs on sequence a, denoted ca , is based on the time that the
shipment is on sequence a and the actual transportation time (total time minus the transshipment
time at other hubs) between the origin hub and the destination hub. The total utilization of the
sequence la determines the costs per item shipped through the transportation network. If ship-
ments are sent via the hub network, additional transit time result in longer travel times, while
the products wait to be consolidated. Therefore, an additional inventory holding cost is incurred
if a shipment from origin i bound for destination j is sent via the hub network. In addition, costs
for handling at the hubs cth, transportation from and to the hubs (cik, ckj) and planning and con-
trol cp&c are incurred.
The total logistics costs in network, C nij , can be formulated as follows:
X
I X
J X
I X
J X
A
C nij ¼ cij aij d ij þ ca ð1  aij Þd ij ð1Þ
i¼1 j¼1 i¼1 j¼1 a¼1

In this formula (1), aij denotes the fraction of demand dij that is transported directly from origin i
to destination j and (1  aij) in the second part of formula (1) denotes the fraction that is shipped
via the hub-network. The fraction aij of demand that is sent through the hub network via sequence
a, based on the capacity available on this sequence, is calculated if cij > ca . The shipment costs of
direct trucking between i (manufacturing location) and j (warehouse) comprise the fixed costs per
shipment, and the variable cost. For this direct trucking an average shipment size is assumed. If
the costs for direct trucking between origin and destination are calculated (and calibrated using
actual costs information), the costs for shipping via the hub network are calculated. These costs,
denoted ca , comprise the fixed costs ðcf Þ per time unit multiplied by the total shipment time (Ta).
The variable costs per time unit ðcv Þ multiplied with the transport time (ta), and the utilization of
the capacity available on sequence a, denoted la.
  
ca ¼ cf T a þ cv ta la þ cik þ ckj þ cth þ cinv þ cp&c ð2Þ
574 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

Fig. 5. Representative cost breakdown of intermodal transport in dedicated pallet hub network. For an average
shipment size (25 pallets), using an 89% utilized network with a distance between origin and destination of 152 km.

The costs for transport from origin i to hub k, denoted cik, and transport from hub k to
destination j, (ckj), are calculated using the same methodology used to determine the direct road
transportation. Next, additional costs are incurred for the handling of the products on the
hub (cth), inventory costs (cinv), and costs for planning and control (cp&c). An important
determinant of the costs for transportation via the hub network is the utilization factor la, a
variable dependent on the capacity that is available on sequence a, and the utilization of this
capacity.
An illustration of a representative cost breakdown of direct trucking versus the costs of ship-
ment through the hub network is illustrated in the figure below. For both alternatives the different
components are illustrated. For direct trucking the annual fixed costs and the labor costs both are
about one third of the costs per pallet. For shipment through the hub network the fixed costs
(including labor) only make up for about 25%. The transshipment costs at the hub and the road
transportation (shipping the pallets to and from the hub, depicted A and C in Fig. 2), dominate
the costs with 49% of the costs per pallet. This illustrates the necessity of cost-efficient and fast
transshipment at the hub and minimization of the distance between the origin and destination
and the hub (Fig. 5).

4.3. The development and implementation of the network

In this specific project the development path of the network was as important as the optimal
solution. In order to make a successful implementation of this new concept possible it was essen-
tial to start the development (and implementation) of the hub network with a feasible and
cost-efficient network, and as the available capacity in dedicated pallet-ships at the start of the
implementation process was limited, capacity restrictions were taken into account in the solution
methodology. In order to convince manufacturers, retailers and logistic service providers of the
potential cost savings that could be achieved using this new concept, the start of the hub network
should already yield cost savings. The methodology that was developed could be best described as
 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 575

Fig. 6. Starting solution: a two hub network using one barge.

an improvement heuristic; it starts with feasible and cost efficient solution and seeks to improve
upon it1. Fig. 6 depicts an example of a start solution of a network between two hubs, forming
sequence a = hk1, k2, k1i. In this figure the thickness of the line between two hubs indicates the po-
tential cost savings if a barge line between both hubs is implemented, using one barge with a uti-
lization of la = 0.95 and ltthi ¼ 0:95 (both barge operations and handling equipment have high
utilization). In other words: the cost savings for all ij are calculated in the most favorable situation
for these origins and destinations, using the shortest path between two hubs with a high
utilization.
If, in this situation the costs for direct trucking are still less than shipment via the hub network
ðca > cij Þ the hub network will never be an efficient alternative so those origin-destination rela-
tions are discarded. The first step consists of finding the best 2-hub connections (the thick dark
line in Fig. 6), the next step consists of increasing the capacity by adding another barge and cal-
culating the new costs costs/pallet for all ij and compare these new costs, based on the all invest-
ments, labor, variable cots and equipment needed, with the costs of direct trucking from the
baseline. If no improvement is found a search is started to extend the network, not by adding
capacity, but by adding a hub to the 2-hub sequence (Fig. 7). For every extension the costs for
all origin and destination relations are calculated and again compared with the costs of direct
trucking. Only those OD-relations for which the hub network is cost-effective will make use of
the network. If finally no improvements are found by adding capacity or hubs, the development
of sequence is finished.
In Fig. 8 the development path of a specific sequence is presented (e.g. the result of the proce-
dure described above for one sequence). In this figure the development of the average reduction in
logistics costs per item of shipments using the hub network is illustrated. In the subjoined table the
number of hubs in the sequence, the number of barges (capacity) and the origin-destination

1
See Sridharan (1995), Klincewicz et al. (1986), Neebe and Khumalawa (1981), Mickey and Thomas (1987),
Abdinnour-Helm (2001), Ernst and Krishnamoorthy (1996), Skorin-Kapov and Skorin-Kapov (1994) and Eberly et al.
(2000).
576 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

Fig. 7. Adding a hub to the sequence.

Fig. 8. Reduction in logistics costs per item during the development of the sequence. Illustrated by the number of hubs,
barges and OD-relations in the sequence.

relations are depicted. In the starting solution a reduction of 2% is realized with a 2-hub system
using only one barge. For 15 OD-relations2 this network would result in a cost decrease. The next

2
In this paper we frequently use the expression origin-destination relation, with the abbreviation OD-relation. An OD-
relation is a flow of one type of product from a manufacturer to a retail warehouse.
 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 577

step in the development of this network is formed by an increase of the capacity on the network
making it possible for 23 relations to reduce their logistics costs by 3% on average. In the next step
of the development a hub is added to this sequence.
This 3-hub sequence then forms a competitive hub network, with an average decrease of 5% in
logistics costs, for 29 OD-relations, compared with direct trucking. If the network is extended an-
other hub is added, resulting in an average decrease of approximately 15% in logistics costs for 65
OD-relations. Finally for 79 OD-relations an average reduction of 20% in logistics costs realized.
This is the natural limit of this sequence, after this step in the development no adjustments (in
hubs or barges) result in a decrease in system costs.
In Fig. 9 the development of the cycle time (between barges) is illustrated. The starting network
is a 2-hub network with 1 barge. This network has a cycle time of 46 h. If, in the next step, a barge
is added to the sequence the cycle time is reduced to about 24 h.
The third step consists of an additional hub in the network, which leads to a longer route
between three instead of two hubs, and therefore a longer cycle-time. If additional barges are
added to the sequence the cycle-time is further reduced. In the final design the cycle-time is
10 h. In the same figure the development of the throughput in pallet per year of sequence 1 is illus-
trated. Starting with about 100,000 pallets per year, this sequence develops towards an annual
throughput of about 500,000 pallets with six barges between the four hubs. In the table below
the characteristics of all sequences are presented. These 11 sequences make up for the final design
(Table 1).

Development of sequence 1
50 600
Cycle-time between barges (hour)

45
500
Pallets per year (x1000)

40

35
400
30

25 300

20
200
15

10
100
5

0 0
1 2 3 4 5 6 7 8
Development of sequence
Cycle time Annual throughput

Fig. 9. Development of the cycle-time between barges of sequence.

578 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

Table 1
Characteristics of the 11 sequences that form the final design
Sequence Throughput Direct trucking Hub network Delta Total annual Cycle Number Number
(·1000 costs costs (euro/pallet) saving time (h) of barges hubs
pallets/year) (euro/pallet) (euro/pallet) (millions
of euro)
1 499 21.7 20.5 1.2 1.69 10:00 6 4
2 502 16.2 13.6 2.6 2.02 8:00 7 5
3 477 16.1 13.0 3.1 2.16 8:00 7 5
4 495 14.9 12.0 2.9 2.09 10:00 7 5
5 494 15.5 12.2 3.4 2.31 6:00 8 4
6 325 17.1 15.5 1.6 1.05 10:00 6 5
7 356 15.5 14.2 1.4 1.03 10:00 6 5
8 466 38.2 36.1 2.1 2.99 10:00 6 5
9 394 28.4 23.9 4.5 2.55 9:00 7 4
10 530 28.3 22.8 5.5 4.22 10:00 6 4
11 330 26.3 21.1 5.2 2.46 12:00 6 5

5. Results of the project

In the previous section the development of a single sequence was illustrated and finally the 11
sequences were presented. This final design, based on the pallet flows of all manufacturing loca-
tions and retail warehouses is illustrated in Fig. 10. In this design seven national operating se-
quences (in The Netherlands) and four international sequences are included. The international
sequences connect hubs located in The Netherlands with hubs in Germany and Belgium.
The separate sequences are depicted by the thick gray scaled lines connecting the hubs nodes.
The remaining manufacturing locations and retail warehouses, not connected to the sequences of
the hub network, illustrated in the figure above, are still serviced by direct trucking; the hub net-
work does not form a cost-effective alternative for all origins and destinations. A total of 4.87 mil-
lion pallets would be transported annually by the sequences represented in Fig. 10 and a total
saving of about EUR 24.51 million a year could be achieved. The total market in this hub network
consists of 26.6 million pallets. In other words, 18% of the potential 26.6 million pallets would be
facilitated by the collaborative hub network3.
Based on the included manufacturing locations and retail warehouses (the OD-relations) the
transportation distance of OD-relations that make use of the hub network averages about
168 km. In the resulting network essentially only direct trucking is used for OD-relations up
through 100 km. The hub network is intensively used over 120 6 dij 6 220 km. Annually, on aver-
age 442,000 pallets are transported on these sequences, with an average saving of approximately
EUR 3.1 per pallet (in comparison with direct trucking), which means an average reduction of
logistics costs of 14% for those OD-relations that make use of the hub network. The results also

3
An indication of the lower bound was found by implementing a 2-hub sequence between the two nearest hubs for all
OD-relations and calculating the costs for shipment through this network, under the assumption that the utilization of
the barges and the equipment on the hub is very high (la = 0.95 and ltthi ¼ 0:95). In this hub network about 35% of all
flows would go through the hub network because the lower logistics costs in comparison to direct road trucking.
 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 579

Fig. 10. The collaborative intermodal hub network design.

revealed that the lead time between vessels varies between 6 and 12 h. An analysis of the possibil-
ities of cross-border pallet transportation was also conducted; resulting in four international ser-
vices illustrated in Fig. 10 (between The Netherlands, Belgium and Germany). In contrast to the
Dutch domestic market, in which barges carrying 550 pallets can be used, 1100-pallet barges can
be deployed for these routes. The inland waterways to and from Belgium and Germany are high
quality waterways and allow for the deployment of large barges subjected to minimal time loss,
due to bridges, floodgates or locks, the number of trips per year and thus the number of pallets
per sequence can be substantially increased in comparison with domestic transport.
580 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

6. Making hub networks work: towards new collaborative structures

The main motivation behind the tendency to look for collaboration between partners in logistic
networks is achieving economies of scale and scope (Zinn and Parasuraman, 1997; Gulati et al.,
2000). Through the combination of activities it is possible to share costs, through sharing of infor-
mation it is possible to avoid unnecessary costs and through avoiding sub-optimization and acting
as one organization the business units that co-operate can work more efficiently and become more
effective at the same time.
In many cases the easiest start for collaboration is to look for possibilities of sharing fixed as-
sets. The parties involved in the collaboration have already invested in these resources and
through maximizing the number of users of theses fixed assets, the cost per unit will drop if the
total capacity exceeds the demand for these facilities. But also in this simple form of collaboration
issues of trust and fear for dependency can form blocking factors. In Fig. 11 the development of
the logistics costs in the network is illustrated and it is shown that for each of the individual par-
ticipants the cost of working on their own is higher than would have been the case if they would
have chosen an alternative system.
If the participants decide to share resources one can expect the cost per unit to drop consider-
ably. There is however a natural limit to this cost decrease. First of all, the law of diminishing
returns applies and secondly there are problems that are likely to arise if the number of partici-
pants in the consortium increases. This is indicated as the natural threshold in Fig. 11. One
can imagine that this threshold is not only due to the fact that it becomes more difficult to agree
on issues of co-operation and investments with many partners, but also that there will be difficul-
ties in allowing someone to enter a successful partnership (the old participants will require from
the new participant an entry fee that compensates them for the initial investments and for the
opportunity to share the profits with them). If the marginal decrease of profit gets lower and

Fig. 11. Development of costs per item, and a threshold in the development of a hub network using a 3PL.
 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 581

the entrance fee gets higher, a natural limit will be reached. In such a case it makes sense to expect
some third party in logistics services (3PL), to step in. A 3PL can neutralize potential conflicts of
interest between the participants, is specialized in finding suitable clients for the system and also
can act as chain manager that organizes the logistics organization for the whole system, taking
into account the different customer requirements of each of the clients.
In fact this is exactly what happened in the case study described in this contribution. There the
logistic service provider (Vos Logistics) took the initiative to commercially exploit the concept and
offered their services to the manufacturer shippers not only with the obligation to find other
clients but also to organize the process of direct trucking, barge transportation and pre- and
end-haul on their behalf. What we can learn from this example is that in finding an optimal
configuration of collaborating partners in a multimodal network there are two complementing
principles: (1) try to find partners that share mutual interests and are willing to invest in the
collaborative network and (2) find a professional service provider that tries to combine the inter-
ests of participants in a (multimodal) network (Underhill, 1996).

7. Implications for research and policy

In the paper we discussed the trends and drivers that make it necessary for manufacturers and
retailers to constantly search for economies of scale. In certain situations a collaborative hub net-
work can form a solution for the problems in trucking today; through network collaboration
economies of scale and scope in logistics networks can be achieved. The use of a relatively slow
mode of transportation does not automatically mean that the lead time increases. If a combina-
tion is sought between inland shipping and road transport scale and scope can be achieved and
responsiveness and flexibility can be guaranteed through road transport; in multimodal networks
the primary objective is not to find the best mode in or/or type choice situations, but to find the
best mix of modes in and/and type combinations.
The principle of achieving economies of scale and scope through collaboration can be applied in
many types of situations but is especially suitable for multi-modal networks. There are examples
where this principle is applied in a global/international network context. One example is the Sony
case described in the Trilog report (Trilog Consortium, 1999). In this case a combination of long
distance container transport by sea is combined with air transport in much the same way as de-
scribed in the Distrivaart example above. The demand that can be predicted well in advance uses
the sea mode, and the excess demand uses the air mode. Other possibilities are the combination of
truck transport and rail transport or the combination of truck transport and short sea transport.
All these examples have in common that multi-modal network solutions apply alternative solutions
parallel to each other instead of the consecutive usage of the alternative modes of transport that is
normally applied. In order for these concepts to be successful they rely heavily on transparent lo-
gistic information systems and the willingness of the partners involved to exchange planning data.
Research should go into a better understanding of: (1) advancing hub network design methods
for full logistics costs (including coordination costs) in realistic networks and, (2) testing the
principles of collaborative multimodal networks that were developed here for other situations
in which the same design method could be used. One can think of multi-modal networks for
short sea and rail or for the combination of air and sea transport, (3) the real potential of cost
582 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583

reductions through collaboration (testing the principles in practice), (4) the mechanisms of collab-
oration in hub networks and finally (5) the cross-over possibilities from transport to other sectors
(energy).

Acknowledgements

Prof. Jos Vermunt played an indispensable role during the entire project and his suggestions
and comments helped to improve the paper significantly and are greatly appreciated. This
research was supported, in part, by TNO, (the Institute of Applied Scientific Research in the
Netherlands), Connekt, (the innovation network for traffic and transport), and the Holland Inter-
national Distribution Council. This support is gratefully acknowledged.

Appendix A

The heuristic used to design the hub network and illustrate the development path consists of
two strategies. The first strategy is to add a hub to the sequence based on the potential cost sav-
ings. As said the process starts with the m-best sequences that form the initial open sequences. The
next step is to insert or add a hub to the open sequence. For all open sequences the m-best inser-
tions and additions are evaluated. If the adjustments are selected the new cycle time, the capacity
and logistic costs are calculated, resulting in the new system costs ðC nþ1 ij Þ. If an improvement in
system costs is found the sequence is added to the open sequences for further development. If how-
ever no improvement in system costs are found, and the capacity not yet increased, an additional
ship is added and again the system costs are calculated. If this increase in capacity does not lead to
a decrease of the system costs, the sequence is discarded.
Adding a hub results in a new cycle time (tcycle) and therefore the new maximum capacity has to
be calculated (Amax). Next the new costs for transport via the hub network, "ij, are calculated ðca Þ
and compared with the costs for direct road transport (cij). If the total logistics costs in the net-
work or system costs C nþ1ij < C nij then the new sequence is saved, if however C nþ1
ij > C nij , then one
more adjustment is made and that is increasing the capacity on the sequence by adding a ship. If in
this adjusted sequence with increased capacity C nþ1 ij > C nij then the sequence is discarded. If, by
nþ1 n
adding capacity, C ij < C ij the sequence is saved.
Eventually no improvements will be found in the system costs of the open sequences and the
optimization ends. The results of the optimization is a collection of development paths that start
with a 2-hub sequence and end with h hubs and B barges.

References

Abdinnour-Helm, S., 2001. Using simulated annealing to solve the p-hub median problem. International Journal of
Physical Distribution and Logistics Management 31 (3), 203–220.
Blumenfeld, D.E., Burns, L.D., Diltz, J.D., Daganzo, C.F., 1985. Analyzing trade-offs between transportation,
inventory and production costs on freight networks. Transportation Research Part B 19B, 361–380.
 B. Groothedde et al. / Transportation Research Part E 41 (2005) 567–583 583

Campbell, J.F., Ernst, A.T., Krishnamoorthy, M., 2002. Hub location problems. In: Drezner, Z., Hamacher, H.W.
(Eds.), Facility Location, Applications and Theory. Springer-Verlag, Berlin, Heidelberg.
Daganzo, C.F., 1999. Logistics Systems Analysis, Third Revised and Enlarged Edition. Springer-Verlag, Berlin
Heidelberg.
Eberly, J., Krishnamoorthy, M., Ernst, A., Boland, N., 2000. The capacitated multiple allocation hub location problem:
formulations and algorithms. European Journal of Operational Research 120, 614–631.
Ernst, A.T., Krishnamoorthy, M., 1996. Efficient algoritms for the uncapacitated single allocation p-hub median
problem. Location Science 4 (3), 139–154.
Gulati, R., Nohria, N., Zaheer, A., 2000. Strategic networks. Strategic Management Journal 21, 203–215.
Horner, M.W., OÕKelly, M.E., 2001. Embedding economies of scale concepts for hub network design. Journal of
Transport Geography 9, 255–265.
Klincewicz, J.G., Luss, H., Rosenberg, E., 1986. Optimal and heuristic algorithms for multiproduct uncapacitated
facility location. European Journal of Operational Research 26, 251–258.
Mickey, R.W., Thomas, L.W., 1987. Solving quadratic assignment problems by Ôsimulated annealingÕ. IIE Transactions
(March), 107–119.
Muilerman, G.J., 2001. Time-based logistics, an analysis of the relevance, causes and impacts, Dissertation Delft
University of Technology, Delft University Press.
Neebe, A.W., Khumalawa, B.M., 1981. An improved algorithm for the multi-commodity location problem. Journal of
Operational Research Society 32 (2), 143–149.
Newell, G.F., 1980. Traffic flow on transportation networks. Series in Transportation Studies, vol. 5. MIT Press,
Cambridge Massachusetts.
OÕKelly, M.E., Miller, H.J., 1994. The hub network design problem, a review and synthesis. Journal of Transport
Geography 2 (1), 31–40.
OÕKelly, M.E., Bryan, D.L., 1998. Hub location with flow economies of scale. Transportation Research Part B:
Methodology 32 (8), 605–616.
Piper Jaffray, OECD, IFC World Fact book, Mercer Management Consulting, (2002). The Global Logistics Market,
presentation.
Skorin-Kapov, D., Skorin-Kapov, J., 1994. On tabu search for the location of interacting hub facilities. European
Journal of Operational Research 73, 502–509.
Sridharan, R., 1995. The capacitated plant location problem. European Journal of Operational Research 87, 203–213.
Trilog Consortium, (1999). TRILOG-Europe end report, TNO Inro, Delft.
TNO Inro, 2001. Groothedde, B., Bovenkerk, M., Kuipers, B., Iding, M., van Dongen, A. Feasibility of Inland
Shipping Networks: Distrivaart, An Analysis in the Fast Moving Consumer Goods, Delft (in Dutch).
TNO Inro, 2002. Groothedde, B., Ruijgrok, C., Iding, M., van Dongen, A. Feasibility of Inland Shipping Networks:
ÔDistrivaartÕ, Logistic Performance and Service, Delft (in Dutch).
TNO Inro, 2003. Groothedde, B. Rustenburg, M. Distrivaart network development, the road to an intermodal hub
network, Delft (in Dutch).
Underhill, T., 1996. Strategic Alliances; Managing the Supply Chain. Penn Well, Tulsa, Oklahoma.
Vermunt, A.J.M., 1999. Multilognet, the intelligent multimodal logistics network, an important node in the worldwide
logistics net, Vermunt Logistiek Advies v.o.f., working paper (in Dutch).
Zinn, W., Parasuraman, A., 1997. Scope and intensity of logistic-based strategic alliances. Industrial Marketing
Management 26 (2), 137–147.