Towards Automated Logistics Service Comparison
Decision Support for Logistics Network Management
Christopher Klinkmüller, Stefan Mutke, André Ludwig and Bogdan Franczyk
Information Systems Institute, University of Leipzig, Grimmaische Straße 12, 04109, Leipzig, Germany
Keywords: Service Comparison, Business Process Similarity, Logistics Management.
Abstract: A recurring task when managing logistics networks in which logistics companies jointly offer services is the
comparison of logistics services based on their underlying processes. The comparison is necessary for the
integration of processes, the selection of logistics providers and the evaluation of a company's performance.
Due to a high diversity of logistics services and their properties as well as due to the high amount of
services automated logistics service comparison is needed to support this task. This paper presents basic
requirements and evaluates the state of the art with regard to these requirements. In addition, an initial
solution approach providing a solid base for future work is outlined.
1 INTRODUCTION
Logistics management plays a central role for most
companies in manufacturing industries. It organises
flows of goods and information across corporate
value chains. Increased value orientation,
progressive globalisation, ongoing concentration on
core competencies, higher requirements towards the
quality of service, and innovation in information and
communication technology led to a high diversity
logistics management has to deal with (Pfohl, 2004).
As a consequence, logistics companies such as
warehouses or carriers start to arrange themselves in
logistics networks in order to jointly offer a logistics
service bundle that is able to meet customers'
expectations. These logistics networks are usually
managed by Logistics Network Service Providers
(LSP) like third and fourth party logistics providers
(Gudehus and Kotzab, 2009). LSPs do not
necessarily have to provide own physical logistics
assets such as trucks for the service delivery. Instead
they need to have a wide knowledge of logistics
processes and of information technology enabling
them to act as the central point of contact to the
customer and to coordinate logistics companies in
order to flexibly configure services within the
network with regard to the customers' requirements.
The main task for LSPs is therefore the network
management which was introduced by (Sydow and
Duschek, 2011) and which comprises four tasks: the
selection of logistics companies which should be
part of the network; the regulation of tasks necessary
to implement the demanded logistics services; the
allocation of these tasks to the companies within the
network; and the evaluation of the network.
A recurring problem within those tasks is the
comparison of logistics services. When selecting
logistics services the LSP needs to check whether
the services offered by companies fit to those
required by the network. Within the allocation it
needs to be examined whether those services are
suitable to implement services needed by customers.
Furthermore, LSPs have to find similar logistics
services which indicate options to obtain economies
of scale during the regulation. Finally, a central task
when evaluating services is to verify that they still
correspond to their initial design. As a manual
comparison of logistics services can be quite
cumbersome due to the high amount and diversity of
logistics services and their properties the objective
of this paper is to briefly outline an automated
approach to the comparison of logistics services to
support decision making within the management of
logistics networks. In particular, the contribution of
this paper is the evaluation of state of the art based
on basic requirements as well as the introduction of
an initial approach satisfying these requirements.
The paper is structured as follows. In section 2
the requirements towards the comparison of logistics
services are outlined. Afterwards, section 3
evaluates related work with regard to these
requirements. The approach is introduced in section
259
Klinkmüller C., Mutke S., Ludwig A. and Franczyk B..
Towards Automated Logistics Service Comparison - Decision Support for Logistics Network Management.
DOI: 10.5220/0003982502590264
In Proceedings of the 14th International Conference on Enterprise Information Systems (ICEIS-2012), pages 259-264
ISBN: 978-989-8565-10-5
Copyright
c
 2012 SCITEPRESS (Science and Technology Publications, Lda.)

4. Finally, section 5 concludes the paper and gives
an outlook on next steps.
2 REQUIREMENTS
This section introduces the basic requirements
towards the automated service comparison within
logistics network management. These requirements
were determined by conducting expert interviews
and case studies in the context of two research
projects and in collaboration with a logistics network
emphasizing the practical need for an appropriate
approach. The requirements are outlined in the
following and examples that are partly based on the
ARIS SmartPath reference processes are used to
illustrate the purpose of the requirements.
Requirement 1 (Flow semantics): The most
important requirement is that the comparison of
logistics services has to be based on the examination
of the behaviour of the business processes which
implement the services independently of which
business process notation is used. Processes as sets
of activities performed in coordination by a single
company (Weske, 2007) and collaborations of them
allow to capture the flows of goods and information
which are implemented by a logistics network in
order to perform the main task of logistics, namely
transferring goods in space and time (Gudehus and
Kotzab, 2009). The reason for explicitly looking at
the behaviour of processes is that logistics processes
are usually characterized by a high degree of
variability. A typical example is to compare
consignment processes to identify consolidation
options. In order to deal with different types of
goods there might be some activities whose
execution depends on the type. In such a case two
processes might be quite different from a structural
perspective as the number of types that can
potentially be handled by a company might differ
from those of another company. Comparing the
behaviour instead helps to determine cases which
both processes can handle. The behavioural view
also allows to compare the actual process execution
with process templates in case of unexpected
runtime variations. This would probably not be
possible from a structural perspective as the
variations are commonly not captured in a model.
The actual behaviour instead can be reconstructed
from data within information systems. Additionally,
notation-independence is needed because the
companies within the network usually employ
different notations, e.g. BPMN, EPC etc, affecting
the identification of appropriate services.
Requirement 2 (Context semantics): While the
flow semantics consider how a service is delivered,
it is also essential to take account of what is done.
Common process notations allow to label activities
using phrases like "transport goods" and "pick
order". This is not sufficient in logistics where it is
necessary to consider the context in which a process
is executed, e.g. during regulation two transport
processes can only be consolidated if their routes are
close to each other or during selection it is necessary
to determine if a company is able to process
individual orders in a special format. Hence, the
second requirement is that activities are compared
under consideration of a detailed functionality
description rather than simply relying on their labels.
Requirement 3 (Level of abstraction): The third
requirement refers to the first two requirements. It
demands that the approach must take the different
levels of abstraction that services can be viewed
from into consideration. For example, there might be
the option to consolidate a simple transport service
with a composed service which consists of a couple
of services, but which depicts a similar transport.
Considering the flow semantics in such a case, a
simple process must be compared to a process
collaboration. Furthermore, companies may provide
more process details than necessary to the LSPs that
are mainly interested in a coarse-grain view onto the
activities and the points of interaction. In this case a
few activities from an LSP's view could correspond
to a complex flow of activities offered by the
companies. At the context level there is also a
difference between the representation of services
offered by companies and of those requested by
customers. While services of companies usually
illustrate companies' capabilities, services demanded
Table 1: Summary of the requirements.
Key phrase Description
Req. 1 Flow semantics The comparison must be based on a notation-independent analysis of the behaviour of the
business processes implementing the logistics services.
Req. 2 Context semantics Process activities have to be compared on the base of a detailed functionality description rather
than relying on labels.
Req. 3 Levels of abstraction The different levels of abstraction services can be described on need to be regarded.
Req. 4 Presentation of results The results must enable analysts to investigate reasons for the similarity of two processes.
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
260

by LSPs are specified with regard to a certain
contract. A simple example to illustrate this is a
transport service. A carrier would usually name the
region in which it is able to conduct transports, e.g.
Central Europe etc., while in a contract there is
usually a demand for a specific tour, e.g. from
Hamburg to Prague. This requirement is most
important when tasks are allocated to companies.
Requirement 4 (Representation of results): In
order to support an LSP in decision making it is not
sufficient to present the result of a comparison of
two services as a single number indicating the
degree of overlap or difference between the services.
Such a number might indeed be useful to preselect
suitable services. Unfortunately, in this case the
reasons for the classification are hidden behind a
single number, which makes it hard for analysts to
further investigate on the most suitable solution.
Thus, the fourth requirement is that an analyst must
be able to examine reasons for commonalities and
differences for decision making using the results.
To summarize this section Table 1 provides an
overview of all four requirements.
3 RELATED WORK
After having outlined the requirements in the last
section existing work is presented and assessed on
the base of these requirements here. Because of the
flow semantics being the central requirement and all
other requirements being based on it the focus is on
approaches that compare processes.
In literature a couple of equivalence notions for
comparing processes can be found, e.g. bisimulation
(Hidders, Dumas, van der Aalst, ter Hofstede and
Verelst, 2005). Following (van Dongen, et al., 2008)
those notions can be excluded from the explanations
in this section for various reasons. The most
important one is that they compute the equivalence
of two processes, i.e. they answer the binary
question if two processes are equivalent or not. As
the fourth requirement states, it is important to make
a statement about the degree of equivalence and to
give hints for further investigation. This is clearly
not satisfied by those notions. Thus, this section
deals with approaches in the field of process
similarity that measure the degree of equivalence.
The first approach outlined here is presented in
(van der Aalst, et al., 2006). Here processes are
compared on the base of finite sets of traces. These
sets usually comprise a certain number of actual
process executions, but can also be derived from
simulations or user defined scenarios. To compare
two processes using sets of traces two metrics are
defined. Both are asymmetric and measure the
similarity based on one of the processes. Besides
counting the number of transition connections that
appear in traces of the original as well as in the
compared model the metrics also account for the
transitions that are enabled within the traces.
In (Dijkman, et al., 2009) the Graph Edit
Distance which indicates how many operations are
needed to transform one process graph into another
one is used to calculate the similarity. To calculate
this metric, the mapping of nodes of two graphs is
determined in four different ways each of them
relying on activity labels.
In (van Dongen, et al., 2008) an approach is
presented that relies on so called causal footprints.
These footprints consist of all nodes of a process
graph and two sets for each node. The first set
comprises all nodes which can be executed before and
the second set comprises those which can be executed
Table 2: Assessment of existing approaches.
Approach Flow semantics Context semantics Levels of abstraction Presentation of results
(Dijkman, Dumas and
García-Bañuelos, 2009)
- Structure
- Business process graphs
- Labels - Not considered - A symmetric metric
(Ehrig, Koschmider and
Oberweis, 2007)
- Structure
- Petri nets
- Labels - Not considered - A symmetric metric
(Kim and Suh, 2010) - Structure
- Special ontologies
- Context information - Not considered - A symmetric metric
(Lu, Sadiq and Governatori,
2009)
- Structure & behaviour
- Process variant scheme
- Labels
- Context information
- Not considered - A symmetric metric
(van der Aalst, de Medeiros
and Weijters, 2006)
- Behaviour
- Petri nets
- Not considered - Not considered -Two asymmetric
metrics
(van Dongen, Dijkman and
Mendling, 2008)
- Behaviour
- Causal footprint
- Labels - Not considered - A symmetric metric
(Zha, Wang, Wen, Wang
and Sun, 2010)
- Behaviour
- Transition adjacency relations
- Not considered - Not considered - A symmetric metric
TowardsAutomatedLogisticsServiceComparison-DecisionSupportforLogisticsNetworkManagement
261

after the current node. Transforming the footprints
into vectors makes it possible to calculate the
similarity as the cosine of the angle between these
vectors. During the transformation matching activities
that are based on labels and in case of EPCs also on
information derived from events surrounding a
function are employed. A similar approach is
introduced in (Zha, et al., 2010). There a process is
represented as a transition adjacency relation
comprising pairs of activities of a process that can be
executed directly one after the other. The similarity is
defined as the ratio between the cardinality of the
intersection of two transition adjacency relation sets
and the cardinality of their union.
In the field of process variants an approach to
determine whether a certain process variant meets a
query which is a collection of features is introduced
in (Lu, et al., 2009). These features can be classified
as behavioural, structural or contextual features. For
all classes algorithms to measure the similarity are
proposed and the general similarity is then defined
as the ratio of the similar features and the number of
features in the query.
Some approaches rely on ontologies used to
describe processes. In (Ehrig, et al., 2007) Petri net
models are represented using an ontology. The
similarity of two concepts from different models is
the weighted sum of the syntactic, the linguistic
(synonym and homonym relations) and the structural
(taking related process concepts into account)
similarity. The similarity of two processes is the sum
of the similarities of concept pairs determined
beforehand by mapping concepts from one model to
those from the other one. In (Kim and Suh, 2010)
five ontologies are defined to describe different
views onto a process including organizational,
domain, structural, resource and service aspects.
Based thereon matchmaking is employed to classify
the match between properties of two processes and
to sum the corresponding similarity degrees.
The assessment of these approaches with regard
to the requirements outlined beforehand is
summarized in Table 2. As can be seen in this table,
approaches exist which examine the behaviour of
processes independently from a certain business
process notation by relying on a representation that
can be derived from such notations. While most of
the approaches use labels to match activities or
assume the match to be done beforehand, two
approaches consider context information. However,
none of the approaches fulfils both requirements.
Regarding the demand for supporting different
levels of abstraction it can be seen that none of the
approaches addresses this requirement. Lastly, all
approaches calculate a single degree of similarity but
do not provide further information. The approach
presented in (van der Aalst, et al., 2006) is slightly
more advanced as it calculates the similarity for each
of the processes being compared.
It is subject to future work and a relevant open
issue to develop an approach which is designed with
regard to all requirements. A first blueprint for such
an approach is introduced in the next section.
4 PROPOSED APPROACH
The basic approach to the automated comparison of
logistics services and the reference of each step
within the approach to the requirements are
presented in Figure 1.
The first step is the transformation of the process
models into notation-independent models with
activity annotations. Candidates for a meta-model
are Petri nets, transition systems etc. On the base of
such a meta-model different transformations have to
be written in order to ensure that process models of
various notations can be compared as demanded by
the first requirement. Existing approaches, like
(Raedts, Petkovic, Usenko, van der Werf, Groote
and Somers, 2007) where BPMN models are
transformed into Petri net models, can be reused.
A further important part of the first step is to
annotate the models during the transformation in
order to add information about the logistics
functionality as necessary due to the second
requirement. The annotation is based on the IOPE-
model which is used within several service
specification approaches like the Unified Service
Description Language (Cardoso, Barros, May and
Kylau, 2010). This model allows for describing
activities with regard to their inputs and outputs as
well as the preconditions and effects as
representations of the state of the world that need to
remain valid before and after activity execution.
Furthermore the IOPE-model allows for applying
the scheme introduced by (Hömberg, Hustadt, Jodin,
Kochsiek, Nagelö and Riha, 2007). This scheme can
be used to describe logistics functionality in terms of
the information and goods that flow through an
activity (input and output) as well as in terms of the
changes made to the time and the space as well as
the states of the information and goods (precondition
and effect). To make these annotations interpretable
for machines, different ontologies as explicit
specifications of a conceptualization (Gruber, 1993)
need to be employed. Regarding the third
requirement the concepts of these ontologies must
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
262

reflect different levels of abstraction and must be
related to each other, e.g. an ontology to describe
states regarding space must enable a modeller to
define regions and routes for transports. While the
region is necessary to describe a company's abilities
the route is needed to specify contract related
requirements. This ontology should also connect the
concepts route and region so that a machine is able
to determine whether a company can handle routes
in a certain region. The actual annotation can then be
done in different ways. If the source model is
already annotated in some way, these annotations
also need to be transformed, e.g. if different
ontologies are used, ontology matching algorithms
(Euzenat and Shvaiko, 2007) will need to be
integrated. In case of missing annotations they can
be added manually while annotations on the base of
the proposed ontologies can simply be copied.
Afterwards the second step is to normalize the
models. This is done because of the third requirement.
The goal of this step is to transform fine-grain process
models into more coarse-grain ones in order to bring
both models to the same level of abstraction. A simple
rule could be to summarize activities that are arranged
in sequence without any points of decision or
interaction in between. One of the approaches
supporting this step is presented in (Koliadis and
Ghose, 2007) where effects of activity executions
within processes are summarized supporting the
summary of the overall preconditions and effects.
The third step prepares the process models for
the actual comparison by matching activities of one
process to the ones of the other process. The
rationale here is to calculate the similarity of all
activity pairs on the base of their annotations.
Afterwards the optimal mapping is determined by an
appropriate heuristic whereby optimal means that
the sum of the similarity of all mapped pairs is as
high as possible. This step is oriented towards the
approach outlined in (Dijkman, et al., 2009) where
the optimal mapping of activities is computed on the
base of the syntactic and linguistic similarity of their
labels. By relying on the annotations this step also
accounts for the second requirement.
The fourth step is the comparison of the models
on the base of their behaviour. As presented in the
previous section there already exist approaches to
compare processes from a behavioural perspective,
like the one presented in (van der Aalst, et al., 2006).
Nevertheless, extension is necessary to consider the
fourth requirement, i.e. besides the computation of a
degree of similarity the main reasons for the result
must also be collected.
The last step is then to present the results to the
customer using an appropriate visualisation that not
only presents the degree of similarity but also the
indicators that were collected in the previous step.
As the services might rely on different notations it is
important to present the results in a way that allows
an analyst to investigate them although he or she is
not familiar with the used process notations.
As the comparison of the original service to a set of
other services is done in pairs and as there might be
a lot of services that need to be compared the
computation time can be high. In order to reduce it
different strategies are possible. The first one is to
estimate the similarity beforehand and only take
those services into consideration which are believed
to be similar to a certain degree, like it is done in
(Yan, Dijkman and Grefen, 2010). A further option
is to preselect services based on the purpose of the
comparison, e.g. in the allocation and in the
selection only services representing a company's
capabilities are regarded. The last option mentioned
here is to configure the features taken into account
within the approach like it is proposed in (Lu, et al.,
2009). Of course all these strategies can be
commonly employed. It is also possible to proceed
iteratively and refine the result set step by step.
5 CONCLUSION & NEXT STEPS
This paper motivated why it is necessary to support
Figure 1: Basic approach for the service comparison.
Transformation
Process normalization
Comparison
Visualization
Activity mapping
TowardsAutomatedLogisticsServiceComparison-DecisionSupportforLogisticsNetworkManagement
263

the management of logistics networks with an
automated approach for logistics service
comparison. The central requirements determined in
cooperation with a logistics network were
introduced. Subsequently, the state of the art was
evaluated with regard to these requirements. As a
first step towards the automated comparison an
approach which consists of five steps was proposed.
These steps include some pre-processing in form of
the transformation into a notation-independent
representation as well as the normalization of the
representation and the activity mapping to equalize
the different levels of abstraction. Afterwards the
comparison is done using the notation-independent,
normalized and mapped process models. The final
step is the visualization making the results
interpretable for experts.
The first step to implement the basic approach is
the selection of a notation-independent represen-
tation and of a basic comparison algorithm. On the
one hand this represents the main functionality of
the approach and on the other hand it constitutes a
solid base for adding the other requirements. It is
planned to evaluate the approach in each
development step in order to ensure the benefit for
logistics management. Hence, experts opinions and
the results of the automated approach will be
compared on the base of scenarios derived from
logistics reference processes and from case studies
conducted within the logistics network.
ACKNOWLEDGEMENTS
The work presented in this paper was partly funded
by the German Federal Ministry of Education and
Research under the project InterLogGrid (BMBF
01IG09010F) and by the European Regional
Development Fund under the project LOGICAL
(3CE396P2).
REFERENCES
Cardoso, J., Barros, A., May, N. and Kylau, U. (2010).
Towards a Unified Service Description Language for the
Internet of Services: Requirements and First Developments.
Proceedings of the 2010 IEEE International Conference on
Services Computing, 602-609.
Dijkman, R., Dumas, M. and García-Bañuelos, L. (2009).
Graph Matching Algorithms for Business Process
Model Similarity Search. Proceedings of the 7th
International Conference on Business Process
Management, 48-63.
Ehrig, M., Koschmider, A. and Oberweis, A. (2007).
Measuring similarity between semantic business process
models. Proceedings of the fourth Asia-Pacific
conference on Comceptual modelling - Volume 67, 71-
80.
Euzenat, J. and Shvaiko, P. (2007). Ontology matching.
Heidelberg, Germany: Springer.
Gruber, T. R.(1993). A translation approach to portable
ontology specifications. Knowledge Acquisition, 5,199-220.
Gudehus, T. and Kotzab, H. (2009). Comprehensive
Logistics. Springer: Dordrecht.
Hidders, J., Dumas, M., van der Aalst, W.M.P., ter
Hofstede, A.H.M. and Verelst, J. (2005). When are two
workflows the same? Proceedings of the 2005
Australasian symposium on Theory of computing -
Volume 41, 3-11.
Hömberg, K., Hustadt, J., Jodin, D., Kochsiek, J., Nagelö,
L. and Riha, I. (2007). Basisprozesse für die
Modellierung in großen Netzen der Logistik. Technical
University Dortmund: Technical Report.
Kim, G. and Suh, Y. (2010). Ontology-based semantic
matching for business process management. ACM
SIGMIS Database, 41, 98 - 118.
Koliadis, G. and Ghose, A. (2007). Verifying Semantic
Business Process Models in Inter-operation.
Proceedings of the IEEE International Conference on
Services Computing SCC 2007, 731-738.
Lu, R., Sadiq, S. and Governatori, G. (2009). On Managing
Business Processes Variants. Data & Knowledge
Engineering, 68, 642-664.
Pfohl, H.-C. (2004). Logistikmanagement - Konzeption und
Funktion. Berlin: Springer.
Raedts, I., Petkovic, M., Usenko, Y.S., van der Werf,
J.M.E.M., Groote, J.F. and Somers, L.J. (2007).
Transformation of BPMN Models for Behaviour
Analysis. Proceedings of the 5th International
Workshop on Modelling, Simulation, Verification and
Validation of Enterprise Information Systems, 126-137.
Sydow, J. and Duschek, S. (2011). Management
interorganisationaler Beziehungen : Netzwerke -
Cluster - Allianzen. Stuttgart: Kohlhammer.
van der Aalst, W.M.P., de Medeiros, A.K.A. and Weijters,
A. J. M. M. (2006). Process Equivalence: Comparing
Two Process Models Based on Observed Behavior.
Proceedings of the 4th International Conference on
Business Process Management, 129-144.
van Dongen, B., Dijkman, R. and Mendling, J. (2008).
Measuring Similarity between Business Process
Models. Proceedings of the 20th international
conference on Advanced Information Systems
Engineering, 450-464.
Weske, M. (2007). Business Process Management. Berlin,
Germany: Springer.
Yan, Z., Dijkman, R. and Grefen, P. (2010). Fast business
process similarity search with feature-based similarity
estimation. Proceedings of the 2010 international
conference on On the move to meaningful internet
systems - Volume Part I, 60-77.
Zha, H., Wang, J., Wen, L., Wang, C. and Sun, J. (2010). A
workflow net similarity measure based on transition
adjacency relations.
Computers in Industry,61, 463-471.
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
264