Dynamic Modelling to Support Organizational Decision Making.
The Stevedore Case
D.T,T.-van Bijck’, RJ. Streng” and H.G. Sol”
Delft University of Technology
“School of Engineering, Policy and Management “Department of Information Systems
P.O. Box 5015 P.O. Box 356
2600 GA Delft 2600 AJ Delft
“ The Netherlands. The Netherlands
Abstract ..
This paper addresses a simulation study that has been carried out in the Rotterdam Port to assess the value
of decision alternatives for terminal entrance facilities and procedures. The organizations involved in this
Project are dynamic systems by nature, so a static analysis and design technique does not suffice when
. supporting decision processes. The Dynamic Modellirig approach ‘was therefore applied in a case study.
1A problem in the port of Rotterdam
The Rotterdam-based company Europe Combined Terminals (ECT) is one of the largest stevedores in the
world. It takes care of the stowage of about 15.000 ocean vessels every year, which are loaded with
250.000 containers, 100.000 cars, 200.000 ton of forest products and 30 million tons of so-called
customized cargo. In that respect, ECT acts as a gate to the European market.
ECT operates 7 terminals in the port of Rotterdam. The largest terminal is called the "Eemhaven terminal",
which is the focus of this paper. Trucks enter the Eemhaven Terminal to pick up cargo that has been
brought in by ocean vessels (the import of cargo) or to deliver cargo that will be transported overseas by
these vessels (the export of cargo). Of course, all kinds of combinations are possible. Furthermore, it
should be noted that cargo trains and barges take care of about 20% of the total cargo flow. In this study,
we have restricted ourselves to the 80% of the cargo that is dealt with by trucks.
The Eemhaven Terminal is depicted in figure 1. The terminal has three entrance gates, where 850
container carriers and 350 conventional trucks enter the terminal every day. The 850 container carriers first
have to pay a visit to the so-called control terminal, which is located about three kilometers away from
the Eemhaven Terminal. On this control terminal, the administrative procedures take place. As an example,
we will briefly describe the procedures that have to be carried out when a truck driver wants to deliver
a container for export to a non-EC country. The driver parks his truck at the control terminal and hands
over the container documents to an ECT desk employee. The documents state the final destination of the
container, the contents of the container and the ocean vessel the container has to be stowed into. The
container is booked into the in-house computer system of ECT and a terminal stack position is assigned
to it, The customs documents for export are filled out, and handed over to the customs officer at the
control terminal, who stamps them to approve for export. The driver gets a copy of the documents and
‘a little note stating the entrance he has to take and the stack position of the container. He or she then
walks back to the truck and drives over public road to the Eemhaven Terminal. There the container is
unloaded and stacked at the right position on the shore, waiting to be stowed.
SYSTEM DYNAMICS '93, 93
Figure 1: Layout of the Eemhaven’ terminal area
It is the administrative process at the control terminal that concerns ECT management for two reasons:
The control terminal procedure causes a significant delay for the truck drivers. On peak hours
the waiting time can be up to two hours. Average waiting times are around 40 minutes.
At the moment, the customs officers allow the trucks to drive from the control terminal to the
Eemhaven Terminal over public road, although they are already approved for export. This will
probably change in the future, when the EC regulations require a stronger rule enforcement at
the borders of the European community.
In the near future there will be an opportunity to redesign the terminal layout and the administrative
Processes connected to the physical handling of cargo. About one third of the current terminal capacity
is used by one large customer of ECT. The stowage operations for this customer will move to a newly
_ built terminal elsewhere in the port by 1993. ECT management plans to redesign the terminal to overcome
the problems mentioned above. The most important aspect is the implementation of a central gate where
all in- and outgoing traffic passes through and where all administrative and security procedures are carried
out. The plan looks globally as follows:
The implementation of infrastructural changes that give direct access from the highway to the
terminal. To support this, the municipality of Rotterdam has started to build a fly-over (crossing
busy train tracks) to give better access to the terminals.
‘The construction of a customs office close to the central entrance.
The construction of ECT administration desks next to the central entrance. This requires the
filling of a small part of the harbor. This might also involve the implementation of new
procedures and information systems for handling containers and customized cargo.
A reduction of the administrative procedures by means of inter-organizational information
technology, in particular the use of fax messages as pre-arrival notices and the use of EDI.
(EDI: Electronic Data Interchange, the computer to computer communication of business
documents (see Streng et al. (1992)).
94
SYSTEM DYNAMICS '93
‘The investments that are necessary to carry out this plan are rather high. Furthermore, there is the risk that
a central gate cannot handle the peak loads in traffic and will cause severe congestion on one of the most
busy highways of the Netherlands. Therefore, the question is if the proposed project indeed solves the
problems sketched above and if it will do so in a satisfying way.
An ECT project team was formed to answer this question. The problem situation is characterized by a high
Igvel of complexity. There are a number of independent organizations involved (ECT, customs, trucking-
companies, forwarders, shipping-lines) that each have their own characteristics. The inter-arrival and
handling time of the trucks are non-deterministic and can only be analyzed on a statistical basis. The
problem situation itself has a close relationship with the dynamics of the primary process of ECT. These
dynamics are essential to the problem and therefore a static approach to the question will not provide the
answers that are needed.
For this reason, ECT has started.a research project ’where it was investigated if the dynamic modelling
approach that is under development at Delft University of Technology (the "LANE"-approach, see Seng
and Sol (1992)) could be used to answer the following question:
"Is a central entrance gate a good solution to the problem and is it possible to operate this gate in an
effective and efficient manner?"
2 A problem solving approach
In our case study, we use the problem solving cycle as described by Mitroff et al. (1974). Although it is
criticized for modelling the problem solving process too much as a rational activity with clear milestones,
we emphasize that we view it as an approach that strongly encourages the use of multiple iterations and
that can be used in solving ill-structured problems, Sol (1982) has elaborated on Mitroff’s process of
problem solving, among other things by defining more concisely the terms "Specification" and "Solution
Finding". The cycle is depicted in figure 2.
yn
es
JS
é
Problem Correspon| dence che Empirical
Perception | Model
Y
” Solution
iF
S ls
{Target system)
Figure 2: Mitroff’s problem solving cycle
When the owner of a problem is realizing his problem, the cycle starts at the perception stage. The first
thing that he or she will do is trying to formulate the problem by using concepts that he-is familiar with,
This gives him the vocabulary to describe the problem situation in detail and discuss it with other people.
This detailed description is called the empirical model and the activity to obtain it is called specification,
SYSTEM DYNAMICS '93 95
The empirical model contains the problem situation as the-problem owner sees it. Based on this model
he or she will try to find possible solutions. Because we want to be explicit on the distinction between
design and choice (see Simon-(1947)), we have subdivided the activity "Solution Finding” into: two
activities: “Solution Finding", defined as the generation of a number of possible courses of action; and
“Solution Assessment", defined as the activity that leads to a a qualification of the various proposed courses
of action and.the choice for the preferred one.
The solution that comes out best in the aupicigl model can be implemented in reality if the problem
owner views it as a satisfying solution. When the implementation is finished, the problem owner can direct
his attention to another (perhaps a new unforeseen). problem which closes the cycle: In. our case, the
management of ECT has:already done part of the problem solving cycle. They have perceived a problem
in reality. Based on the problem situation as they see it, they have found one altemative solution.
The study presented in this paper deals with the assessment of this solution. It has to be checked if it
indeed solves the problem in a satisfying way. For this, hard figures were needed of the terminal perfor-
mance in the future compared to performance in the existing situation. As the models developed by
management were not documented and not detailed enough to perform a rigid statistical analysis, we had
to iterate the conceptualization-specification cycle again to develop models that would allow for such a
statistical analysis. For these modelling efforts we used the LANE-approach, which is especially developed
to deal with organizational problems that have strong dynamic interactions with other organizations.
3 Dynamic Modelling to support (inter-)organizational decision making.
As mentioned before, the problems in the Eemhaven case extend above the. organizational level. The
problem and its solutions are strongly affected by-the way other organizations such.as customs, trucking
‘companies and other organizations will cooperate.
The following can serve as an illustration of this point. The trucking companies that bring the cargo to
the Eemhaven Terminal are not the customers of ECT. In the transportation business, there are separate
organizations called shipping-agents or forwarders that serve as an intermediary between the stevedore and
the trucking company. Therefore, ECT cannot speak directly to the trucking companies about the
improvements in the information system, but has to deal first with the shipping-agents.
So we are trying to solve problems that have solutions on the. inter-organizational level. However, the
value of these solutions can be found on the organizational level. Sol (1988) has made this distinction
more explicit by identifying three perspectives from which we can view the system:
. The macro-perspective concems the strategic planning of the organization in a societal sector.
From this perspective an organizational network is perceived: a group of cooperating
organizations that fundamentally retain their independence.
. The meso-perspective concems the coordination of workplaces in-an organization, with the aim
of improving the performance of the organization as a whole.
* The micro-perspective concerns tasks of individuals or small groups at the so-called work
places. The focus is on improving human performance.
96 SYSTEM DYNAMICS '93.
The proposed solution aims at improving the performance of the stevedoring company. It is clear that
inter-organizational efforts will be needed to reach this goal, which implies that there should be an
advantage for the other organizations as well.
‘When it comes to modelling the system under consideration, we have to deal explicitly with these three
different layers. We therefore propose a layered modelling approach called LANE, which stands for
Layered Actors, Networks and Entities approach. It-gives the modeler a rough guideline when he or she
wants to construct a model of the current situation or a model of a possible future situation. LANE is
based on two assessment techniques:
. In the first place the dynamic models created by the LANE approach are executable and can
be used to perform a statistical analysis of the system under consideration.
fe By means of visualization techniques, the modeler can present an animated view on the models
constructed and thus on the system under consideration. This can lead to a better understanding
and a better agreement of the possible future situations, especially when it comes to
implications that are difficult to express in hard figures. Wierda (1991) has proven that this
method leads to a better agreement on the problem and its possible solutions.
In the next section, we will describe how the LANE approach has been applied to the given problem.
4 The LANE-approach applied to the stevedore’s problem
The LANE approach distinguishes Actors, Entities and Networks. Actors are active objects that handle
passive objects, the Entities. Actors on the macro-level are the organizations and on the meso-level: the
departments within one organization. On the micro-level the actors are the employees of the organization.
In this way, the LANE approach explicitly deals with the three perspectives as distinguished by Sol. The
actors send entities such as containers and fax messages to each other and they send them over Networks,
such as the road network or the telephone network.
Actors and entities can be described by using a notation that is borrowed from the object-oriented world.
During the case study a tool called Fast-LANE has been developed to do the bookkeeping concerning the
objects. The actions, carried out by the actors can be described by means of task structures, as developed
by Bots (1989) as part of his task analysis technique. This technique is also used to depict the way of
working of the LANE approach in figure 3. The little bombs in the task descriptions indicate that the task
consists of sub-tasks and can therefore be "exploded". We can distinguish the three layers, as well as the
sub-division between "conceptualization" and "specification". In the following we will describe the various
activities that have been carried out in the ECT case. Note that the description seems to be linear, but that
in fact we have carried out numerous iterations.
According to the LANE approach, the problem situation was analyzed from the three perspectives as
defined above. In each perspective, a conceptualization as well as a specification activity is carried out.
The result of the conceptualization activity is a data void, qualitative representation of reality. The
specification activity combines this representation with the proper empirical figures.
SYSTEM DYNAMICS '93. 97
‘Agtoomenton what Agrooment on what Agrooment on the
‘organizations parts of the organization understanding model ?
Figure 3; Task structure of the LANE approach
4.1 The macro level
From a macro perspective, it is important to demarcate the problem area. We have to determine what
organizations are important for our problem, what networks connect them and what entities are being sent
from one organization to another over what network. The potential participants can be determined when
we look at the primary process of the interorganizational network: the transportation of cargo from shipper
to consignee. In the typical case, the shipper sends a request to a forwarder who will then arrange the
transport. The forwarder sends a shipping instruction to a shipping-agent and a transport request to a road
transport organization: The shipping agent sends a loading instruction to the stevedore. who takes care of
the stowage in-ocean vessels. The transport company sends a truck to: the shipper to collect cargo and.
bring it to the stevedore for overseas shipment. The organizations involved are displayed in table 1.
Stevedoring company (ECT)
Transport Companies
Shippers
Forwarders
Consignees
Customs
Table 1: Organizations on the macro level _
The next step is. then to achieve empirical figures about the operation of this complex system. Some
figures are given in section one, but a lot more data could be incorporated in the model. We have
determined which organizations play an important role in our problem domain. We have to zoom in to
the meso-level to get more detailed information about these organizations. The entity flows all come
together at ECT. Therefore there should be zoomed in to the meso level at least for the ECT organization.
The meso. perspective for ECT is presented in the next section.
98 SYSTEM DYNAMICS '93
4.2 The meso level
In this stage, we have to determine what actors and entities play major roles at ECT in handling cargo.
This can be done by an analysis of the primary process of ECT. The LANE approach proposes to do this
by following the entities through the system (material entities like trucks and containers as well as
information entities like documents and fax messages). The result of this analysis is a detailed description
of the processes that are carried out within the relevant parts of the system. This description can be
extended with empirical data like interarrival time distributions of trucks, driving and handling time
distributions etcetera. As an example, a terminal interarrival time distribution is given in figure 4. The next
activity is to zoom in to the actual actions that are carried out by the actors in the relevant parts of the
system in order to determine the factors that cause bottlenecks or delays. This is done by an analysis of
the micro level, which is discussed in the next section.
Terminal time
Aleman wetter contin tarminat
roertage of tris witnin atone
9020 30 a0 Se as 70 Go Oo 00 0 a0 1910 190 17000 a
2 mepirton eae on
Figure 4: Example of a terminal interarrival time distribution
4.3 The micro level.
From a micro perspective, we look at the tasks to be carried out at the workplace. When these tasks are
analyzed, empirical data about their duration can be combined with it in order to calculate delays and
queue lengths. The tasks are described using the task analysis technique as proposed by Bots (1989). The
task of an employee working at the administration desk for unloading services is given as an example in
figure 5.
4.4 From description to improvement
‘After all the analysis activities, the qualitative and quantitative data can be incorporated in a discrete event
simulation. model, connected with an animation of the current situation. This simulation model can be
validated by using standard statistical methods (replicative validation) and by showing the animation to
field experts (structural validation). An example of an animation screen can be found in figure 6. After
validation, experiments can be done with the descriptive model in order to obtain figures of the existing
situation. These figures should be in accordance with empirical data. In order to obtain the desired 90 %
confidence intervals, 20 replications of the model should be run.
SYSTEM DYNAMICS '93 99
Desk Worker
Container exempted 7
Desk worker
Notice of arrival with driver ?
Desk worker)
Find Find arrival
document
Desk worker}
Exempt
container
ECT System
Print
documents
Desk worker)
‘Send driver
to gate with,
documents
Figure 5: Task structure of a desk worker
Using the descriptive model as a base model, the alternatives for solution, like the implementation of EDI
or more fax messages in a central gate setting, can be tested by applying these to the model. The
conceptualization-specification has to be carried out one or more times. After this a number of ’what-if”
questions can be answered.
.
UESG Level
Figure 6: Example of an animation screen
100 SYSTEM DYNAMICS '93
5 Using the models in organizational decision making
Several changes and sub-solutions to the existing gate situation could be made. The changes were
evaluated by implementing them in the simulation/animation model. We will discuss the results of each
alternative briefly. An complete report of the results can be found in van Eijck (1992),
5.1 The centralized gate
The centralization of the gate was tested for performance improvement which includes handling times,
queue lengths and traffic flow. Table 2 presents a comparison between existing and future situations for
several key factors. A change of procedure at the administrative desks is incorporated in this solution. We
Tecognize some substantial improvements. We also see that in a central gate situation, customized trucks
have to wait longer because they are now in the same- queue as-container trucks.
Variable Existing Central Relative
situation gate improverhent
Avg. Handling time (Container trucks) (minutes) “108.51 72.82 32.9 %
Avg. Handling time (Customized trucks) (minutes) 96.82 128.73 -32.9 %
Handling Time (Average total) (minutes) 106.35 82.48 22.4 %
Max. Queue length (at parking place) 81 61 24.7 %
Table 2: Simulation results of central gate alternative
5.2 Desk capacity planning
The central gate model showed peak values for queue lengths and handling times in the morning hours,
This was the reason for testing the central gate with more personnel capacity in these hours. This capacity
shift caused another improvement of 4.2 % in average total handling time and another 22 % (!) in
maximum parking space requirement, Our suggestion is to give extra attention to administrative desk
capacity planning. .
5.3 Fax and EDI usage
We also tested queue length and handling time as a function of the amount of fax usage by transport
companies (the fax is sent ahead of the driver in order to obtain faster administrative handling at ECT
desks). In the first experiment (the central gate) 5 % of the trucks were announced before arrival by fax
message. Values of 10, 15.and 20 % were also tested.-Figure 7 shows the total handling time as a function
of the different fax usage percentages. We see a maximum extra improvement of 5.2 % at a usage rate
of 20 %. Furthermore, the effects were tested of possible usage of EDI applications to notice truck drivers
in advance. Rates of 5, 10, 15 and 20% of EDI usage were tested, at a fax usage rate of 10 %. An extra
improvement of 3.2 % could be reached, the results are presented in figure 8.
‘With fax usage, the maximum parking space requirements could be brought back to 49 trucks (which is
an extra improvement. of 19.6 %. With 10 % fax usage, the requirement for parking space could be
brought back to 37 if 20% of the trucks is noticed in advance with an EDI application (17.8 % extra
improvement.
SYSTEM DYNAMICS '93, 101
Total handling time Total handling time.
- a tencton of fax wow = a tenetion of ©! mow
iin ces 3 i nies ers =
Figure 6: Handling time and fax usage Figure 7: Handling time and EDI usage
6 Conclusions
Conclusions can be drawn with respect to the experimental. results as well.as to the methods that were
used to achieve these results. With respect to the results, it was rather surprising that fax and EDI
prearrival noticing has relatively little effect (less than 5 percent). The effects of replanning the gate and
reengineering the business processes were much more spectacular (for some values more than 30 percent).
The methods that we label as dynamic modelling: simulation, animation and the LANE approach, helped
a lot to get agreement about which part of the system should be examined more closely: in order to get
confident figures. Clearly it was impossible to investigate the whole situation as one complete system. The
simulation and animation facilities clarified possible solutions and outcomes to management and workfloor.
References
Bots, P.W.G., An Environment to Support Problem Solving, Ph.D. Thesis, Delft University of
Technology, The Netherlands, 1989.
VAN EUCK, D.T.T., Dynamic Modelling in a Rotterdam Port Area, MSc Thesis, Delft University of
Technology, The Netherlands, 1992.
MrTrorf, LI., F. BETZ, LR. PONDY AND F. SAGASTI, On managing science in the systems age: two sche-
mas for the study of science as a whole systems phenomenon, TIMS Interfaces, vol.4,no.3,1974.
Simon, H.A., Administrative Behavior, Macmillan, New York, NY, 1947.
SOL, H.G., Simulation in-Information Systems Development, Ph.D, Thesis, University of Groningen, 1982.
Soi, H.G., Information Systems Development: A Problem Solving Approach, Proceedings of the
Symposion on Systems Analysis and Design, Atlanta, 1988.
STRENG, RJ., AND H.G,-SOL, A Dynamic Modelling Approach to Analyze the Organizational and
Financial Implications of Electronic Data Interchange in Transport, Proceedings of ‘the
International Conference on Economics/Management and Information Technology, Tokyo, 1992.
STRENG, R.J., C.F, EKERING, E. VAN HECK AND J.F. SCHULZ (EDS), Scientific Research on EDI; bringing
worlds together, Samsom publishers, Alphen a/d Rijn the Netherlands, 1992.
STRENG, R.J., A Dynamic modelling approach to assess the value of EDI, A study in the Rotterdam port
community, Ph.D. Thesis, Delft University of Technology, The Netherlands, 1993.
WIERDA, F.W., Developing Interorganizational Information Systems, Ph.D. Thesis, Delft University of
Technology, The Netherlands, 1991.
102 SYSTEM DYNAMICS '93
