Vous êtes sur la page 1sur 14

Computers in Industry 84 (2017) 2538

Contents lists available at ScienceDirect

Computers in Industry
journal homepage: www.elsevier.com/locate/compind

Impact analysis of ERP post-implementation modications: Design,


tool support and evaluation
Minou Parhizkara , Marco Comuzzib,*
a
Department of Computer Science, City University London, Northampton Square, EC1V 0HB London, United Kingdom
b
School of Management Engineering, Ulsan National Institute of Science and Technology (UNIST), 50 UNIST-gil, Ulju-gun, Ulsan 44919, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Article history:
Received 10 June 2016 ERP systems evolve in the post-implementation phase because of changing business requirements. Post-
Received in revised form 25 October 2016 implementation changes are likely to decrease the quality of ERP systems and of the data that they use,
Accepted 8 November 2016 which negatively impacts organisational performance. We propose a framework for impact analysis of
Available online xxx ERP post-implementation modications. Our framework allows mapping dependencies among ERP
system components and, based on these dependencies, automatically assessing the impact of a proposed
Keywords: change on both the design-time structure and run-time landscape of the system through a novel set of
ERP impact metrics. The framework also provides semi-automatic support to safely terminating the running
Impact analysis
process instances affected by change. The framework is evaluated with expert users in two pseudo-real
ERP-post-implementation
ERP system implementations.
ERP change
Enterprise systems evolution 2016 Elsevier B.V. All rights reserved.

1. Introduction by the system. It degrades, for instance, when different functions


manipulating the same object in an ERP system are not compatible,
ERP systems are packaged software systems supporting most that is, using function A makes a data object unusable for function
operations of organisations, such as nance, logistics, human B. ERP data quality refers to standard data quality dimensions of
resource management, and manufacturing [5]. Most medium to the ERP database, e.g., the accuracy or currency of customer
large organisations already have gone through at least one cycle of information [17]. Poorly managed post-implementation modi-
ERP software implementation and nd themselves in the so-called cations reduce the ERP system and data quality, leading to chaotic
post-implementation phase [7]. systems that are hard to manage [18,40,45]. Therefore, there is a
ERP systems inevitably change in the post-implementation need to support stakeholders, such as business analysts and
phase to align the ERP functionality to changing business developers, in assessing the impact of ERP post-implementation
requirements [30,40]. Post-implementation changes may be change and to guide them during the implementation of proposed
driven by internal reasons, e.g., a new and more efcient modications. This need is currently only partially answered by the
warehouse management policy is suggested by the management existing literature on ERP systems modelling and design.
and must be reected in the ERP system, or by external factors, There is a large body of literature, in fact, dealing with advanced
such as suppliers, technology providers or regulatory agencies [15]. methods and tools to model ERP system components [31,36], align
For instance, an ERP system may require updates to accommodate ERP system functionality to business requirements [27,37] and
new regulations devised by nancial authorities to counter optimise ERP systems installations [34,44]. This literature,
emerging money laundering practices. however, focuses on the ERP implementation and early usage
The literature on ERP post-implementation focuses mainly on phases. In these phases, designers have more freedom in
post-implementation CSF (Critical Success Factors), e.g., conguring and possibly customising the ERP system to address
[15,28,33,29]. ERP system and data quality typically are possibly conicting requirements. In the post-implementation
considered CSF of the post-implementation phase [16,36]. ERP phase, the existing design-time structure of an ERP system and its
system quality refers to the quality of the data processing enabled currently instantiated business operations is crystallised and may
limit the options available to designers. A change of the process
handling purchase requisitions, for instance, may affect all the
* corresponding author. open purchase cases with suppliers. Hence, the impact of this
E-mail addresses: minou.parhizkar.1@city.ac.uk (M. Parhizkar), change on the ERP system should be carefully analysed and,
mcomuzzi@unist.ac.kr (M. Comuzzi).

http://dx.doi.org/10.1016/j.compind.2016.11.003
0166-3615/ 2016 Elsevier B.V. All rights reserved.
26 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

furthermore, currently running cases affected by a change should since maintenance packages are normally designed by software
be allowed to terminate safely before a change can be imple- vendors with the aim of minimising the impact on existing
mented. installations.
To close the research gap identied above, this paper presents a As far as the impact of enterprise systems change on
framework, i.e., a set of methods and a tool, to support operational aspects of an organisation is concerned, de Boer
stakeholders in the impact analysis of ERP post-implementation et al. [11], Krishna et al. [20] and Dam et al. [10] have proposed
changes. Our framework allows mapping the dependencies among approaches to change identication and propagation in enterprise
design-time and run-time entities of an ERP system and architectures (EA). While de Boer et al. and Krishna et al.s work
automatically detecting dependencies among them relevant for focus on change impact analysis, Dam et al.s work also considers
a given change request. Based on the identied dependencies, the repairing dependencies in an EA that have been disrupted by
framework supports the user in analysing the impact of a proposed changes. These works consider a methodological approach similar
change through a semi-automated impact analysis mechanism. to the one presented in this paper, identifying mapping of
The scope and depth of a proposed change can also be analysed dependencies among EA elements, taxonomy of possible changes,
through a novel set of ERP post-implementation modication and mechanism to assess change propagation as main constitu-
impact assessment metrics. ents. None of them considers dening metrics to quantify the
The implemented tool, which shows the feasibility of our impact of proposed changes.
framework, has been evaluated by ERP experts in hands-on EA change impact analysis presents two limitations. Firstly, the
sessions with different change management scenarios to evaluate scope of EA is much broader than ERP or, more generally, enterprise
usability and t for purpose of our framework in ERP post- systems. Therefore, EA change impact analysis tend to consider ERP
implementation change management. systems as black boxes in the organisational software landscape,
The paper is organised as follows. Related work is reviewed in failing to focus in depth on the internal structure of these systems.
the next section. The design of the framework is presented in Secondly, EA typically focuses on design-time concerns. This
Section 3. Section 4 discusses the tool implementation. The means that EA change impact analysis fails to address the impact of
evaluation with ERP experts is discussed in Section 5 and change on ongoing business operations, such as long-running
conclusions are nally drawn in Section 6. instances of business processes that are not yet completed at the
moment of change.
2. Related work From the standpoint of engineering change, ERP systems are
perceived as complex software products embedded in a complex
Business-IT strategic alignment is a long standing issue in the organisational environment. Software systems change impact
enterprise systems literature [23]. More recently, process align- analysis is a discipline of software engineering focusing on
ment has emerged as a research discipline in which organisational understanding, predicting and possibly quantifying the impact
business processes are considered the main vehicle to reach this of source code modications [6,16]. While most works in this area
alignment [8,38,39,41]. By embedding organisational business focus on source code change impact analysis, there are also efforts
processes, ERP systems can be seen as one of the main technical considering impact analysis on models of the software behaviour,
artefact to reach organisational process and strategic alignment such as UML diagrams [19,43]. Various techniques have also been
[44]. Managing post-implementation changes is a way of proposed to deal with inconsistencies and evolution in the
preserving process alignment. However, while the literature on requirements engineering phase [20,42]. In the context of ERP
process alignment mainly analyse ex-post how process alignment systems, Dor et al. [13] propose a technique for change impact
impacts organisational performance [8,39], in this paper we take a analysis and propagation analysis based on program slicing. The
design perspective, aiming at providing solutions to facilitate the proposed technique is tested extensively on the SAP codebase.
management of post-implementation changes. Software systems impact analysis focuses on the impact of
Even though research about improving design solutions for ERP proposed changes on the existing source code, while our approach
systems is often limited by the lack of access to real ERP focuses on the impact of proposed changes at the operational level,
implementations and by the complexity of real world ERP that is, on the conceptual entities dening an ERP system and the
installations, literature contributions in the areas of ERP concep- currently running business processes. As such, we argue that
tual modelling and ERP risk management are abundant. software systems impact analysis should follow impact analysis
Soffer et al. [36] propose an ERP modelling language for ERP based on our framework. That is, once the operational impact of a
systems, which considers business processes and business objects proposed change has been analysed, the next step is to analyse the
as the main conceptual constituents of an ERP system. Dependen- impact of the change implementation on the existing source code.
cies among processes and objects can also be captured. An Note, in fact, that changes to an ERP system may not necessarily
algorithmic approach to align ERP systems capabilities with involve modication of the source code, but they can be
enterprise requirements using this modelling language also has implemented through conguration of existing options or bolt-
been proposed by the same authors [37]. Both works focus on the on to functionality provided by an external third party [7].
ERP implementation phase and the proposed language does not
model run-time entities and their dependencies with design-time 3. Framework design
business processes and objects. A model of ERP systems compris-
ing business processes, functions and business objects is also The design of our framework is divided into the following three
adopted by Manuel and AlGhamdi [25] and Park and Kusiak [31] in parts:
the context of designing ERP systems for improved process
integration.  Design of the ERP dependency meta-model (Section 3.1): this
A large stream of research can be found in the area of risk meta-model allows dening dependencies among the entities of
assessment for ERP maintenance [24,22]. From the point of view an ERP system both at the levels of design-time static structure
of the target of change, these works consider planned maintenance and run-time operations;
activities, such as system upgrades, rather than ad-hoc modica-  Design of the impact analysis mechanisms (Section 3.2): given a
tions to support emerging business requirements. Moreover, the proposed post-implementation change modication, this mech-
impact of maintenance activities has to be considered limited, anism allows to automatically identify the ERP components
M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538 27

Fig. 1. ERP dependencies meta-model.

affected by change and to semi-automatically support the user in Post-implementation changes concern changes to align the ERP
safely terminating the currently running process instances system to changing business requirements [7,15]. These changes
affected by the change; usually concern the static design structure of an ERP system. Based
 Design of the impact assessment metrics (Section 3.3): this on this assumption, the set of possible post-implementation
metrics quantify the impact of a proposed change based on the modications is derived by combining the primitive operations
results of impact analysis. create, update, delete with the design-time elements of the meta-
model, i.e., the Changeable Entities. Note, in fact, that post-
implementation change management is not concerned with ad-
3.1. ERP dependency meta-model hoc interventions on running business process instances, but it
rather focuses on permanent changes of the design-time entities
The meta-model of dependencies among entities constituting identied by the meta-model.
an ERP system is shown in Fig. 1. Besides involving, i.e., having an impact, on other design-time
The model integrates common perspectives of different ERP entities, a proposed modication may also affect one or more
models proposed in the literature. The design-time part of our running process instances. For each instance, it is possible to
meta-model follows from considering a traditional 3-tier archi- identify a critical point of execution in respect of a proposed
tecture of ERP systems of user interface, application and data change. This critical point is a particular point in the process
layers. The application layer in the middle is further decomposed execution beyond which the execution of a process instance can be
into business processes realised as the orchestration of business safely terminated after the implementation of the proposed
functions. Similar decompositions have been considered for change. A more precise characterisation of this critical point is
instance by Manuel and AlGhamdi [25] and Park and Kusiak provided in the next section while discussing the impact analysis
[31]. Strnadl [38] also has proposed a model of organisational IT mechanisms.
based on processes, services (functions) and business information Table 1 presents a set-theoretic representation of the model of
(data objects). Our meta-model allows for hierarchical aggregation Fig. 1, which is used to support the presentation of the impact
of processes, that is, functions in business processes may refer to analysis mechanism and the formalisation of the metrics.
the execution of sub-processes. Such a hierarchical aggregation is In the remainder of the paper we consider a running example of
typical of processes as service-based compositions [26]. typical sales and procurement processes integrated by an ERP
As far as the run-time is concerned, process instances represent system in a traditional manufacturing rm. Fig. 2 shows a BPMN-
individual execution of business processes, e.g., the processing of a based representation of a possible design-time scenario in the
specic purchase order. Process instances create instances of running example. Functions are represented as BPMN activities,
individual functions, which may require the instantiation of whereas business objects are represented as BPMN data objects. In
specic data objects, or documents, for their execution. the example, the sales process starts with a customer inquiry,
28 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

Table 1
Notation summary.

Element Notation Related predicates


D E
ERP system ERPSYS PRO; FUN; OBJ; PIN; FIN; DOC
 
Business processes PRO prop ; p 1; . . . ; P
 
Business functions FUN f unf ; f 1; . . . ; F use pro fun 
PRO  FUN
Business objects
OBJ fobjo g; o 1; . . . ; O use fun obj 
FUN  OBJ
 
Active instances of a process prop PINp pinp;i ; i 1; . . . ; I called by  PIN  PIN
PIN [p1 PINp
P instance of
pro  PRO  PIN
 
Active instances of a function f unf FINf f inf ;j ; j 1; . . . ; J instance of fun  PIN  FIN
FIN [f 1 FINf
F
 
Active documents instances of an object objo DOC o doco;k ; k 1; . . . ; K instance of obj  FIN  OBJ
DOC [o1 DOC o
O

Set of changeable entities CHE PRO [ FUN[
OBJ fcheg
Post-implementation change chc cheOLD ; cheNEW ; update
chc cheOLD ; delete
chc cheNEW ; create
Set of post-implementation changes CH fchc g; c 1; . . . ; C involves  CH  CHE
affects  CH  PIN
involveschc ; che; che 2 CHE
affectschc ; pin; pin 2 PIN
Critical point of process instance pinp;i cp : CH  PIN7!FIN
D E
Impact of a change chc an ERP system ERPSYS (at design- IMPchc ; ERPSYS DTI; RTI RTIchc fcheeCHEjinvolveschc ; chegDTIchc fpinePINjaf f ectschc ; ping
timeDTI, and run-timeRTI).

Fig. 2. Running example (design-time scenario).

which may eventually result in a purchase order. If the ordered 3.2. Impact analysis mechanism
goods are not in stock, then they have to be manufactured on
demand. This involves procuring the required raw materials from In our framework, given an ERP system ERPSYS and a proposed
supplier(s) (procurement sub-process) and manufacturing the change ch, impact analysis is the problem of determining which
goods (production sub-process). Once the goods are sent to the predicates involves (ch, che) and affects (ch, pin) evaluate to true,
customer, an invoice is issued and the process terminates when the that is, to dene the design- and run-time impact DTIch and
payment is received. RTIch, and to support the safe termination of the process
In the procurement sub-process a request for quotation is rst instances affected by change in RTIch.
issued to the supplier and processed, e.g., compared to previous In the remainder of this paper we focus on the update type of
quotations, until an agreement on the supplier bid is reached. This ERP post-implementation changes, which covers also the create
triggers the creation of a purchase order. The process terminates and delete types of changes. Any update modication, in fact, can be
when the invoice received from the supplier is paid and registered modelled as the sequential combination of deleting an existing ERP
as an accounting entry. Note that the Supplier PO business object is entity and creating a new entity as a substitute.
required by both the receive goods and verify supplier invoice
functions for consistency verication purpose.
M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538 29

Fig. 3. Impact analysis mechanism.

In our running example, let us consider the following change ch: These process instances have to be safely brought to completion
D E minimising the disruption introduced by the change (see the
ch SupplierPOOLD ; SupplierPONEW ; update ; Safely Complete and Migrate Run-Time Instances phase in Fig. 3).

that is, an update of the Supplier PO business object. This change 3.2.2. Phase: assess design-time compatibility
may be required by a new directive from a nancial authority to Not all changes are critical for a given ERP system. Some
show tax information and prices including tax calculated based on changes, in fact, may be perfectly backward compatible with the
the nancial regime of the suppliers country. entities affected by the change and, therefore, with the existing
Fig. 3 shows a process-based view of impact analysis in our ERP system. For example, a compatible update of the Supplier PO
framework using the BPMN notation. Phases that can be fully data object in our running example may concern adding a new eld
automated are modelled as BPMN service tasks (see the cog icon to capture more details about the supplier identity. The new
in the top left corner of the task symbol), whereas phases that may Supplier PO is able to replace the old version, since the processes
require human intervention are modelled as BPMPN user tasks (see and functions using it can simply ignore the new eld. In other
the user icon) In the remainder of this section, we discuss in
words, we say in this case that new version chenew of an entity is
detail each phase of the mechanism.
compatible with the old version cheold .
The notion of compatibility of proposed changes with an
3.2.1. Phase: calculate dependencies
existing system is a well-known issue in the literature about
In this phase, the design-time and run-time dependencies
change management [10,13]. Assessing compatibility between
DTIch and RTIch of the proposed change ch are calculated based data objects [9], functions or services [32], and business processes
on the ERP dependency model. [12] implies to verify the syntactic and semantic similarity
Regarding design-time dependencies, let us consider the case in between two versions of the same entity. Syntactic similarity
which the change ch involves a data object obj (in our running refers to similarity among the specications of different entities,
example, obj = SupplierPO). In this case, all the functions f unf using e.g., the matching among the signatures of two functions or the
obj as either input or output are affected by the change. Moreover, syntactic matching among activity labels in a process models.
all the processes prop using the functions f unf affected and, Semantic similarity refers to similarity of the meaning of two
recursively, all the processes using the affected processes as entities, e.g., two business functions are semantically similar if
sub-processes at some point are also affected by the change ch. In they implement the same function between input and output
other words, identifying the design-time impact DTIch implies to parameters, or two process are semantically similar if they produce
navigate the relations use pro fun, use fun obj and is sub process the same results for the client.
in the meta-model of Fig. 1. In particular, the sub-process relations Compatibility checking in our framework is a semi-automated
allows to take into account transitive dependencies among activity. While, in fact, some tasks related to compatibility
processes using affected functions. assessment may be fully automated, e.g., checking the syntactic
In our running example (see Fig. 2), the change ch similarity between the schema of two data objects or two process
SupplierPOOLD ; SupplierPONEW ; update affects the functions Create models, other tasks require human intervention, e.g., judging
Purchase Order, Receive Goods and Verify Supplier Invoice and the whether the execution of two functions or processes creates the
processes procurement and sales. same result for the user.
As far as run-time dependencies are concerned, RTIch Based on whether a proposed change ch is compatible with the
comprises all the currently running, i.e., started but not yet existing ERP system, our framework involves two different paths
terminated, instances of the processes identied as part of DTIch. (see Fig. 3). The trivial case is the one of compatibility of the old
30 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

version of an entity cheold with the new version chenew . This case conguration, which can be done very quickly, even automatically,
involves the implementation of the change and the migration of to developing new code or customising the existing code base of
affected running instances to the new version of the entity, after the ERP system (see [7] for a taxonomy of possible ERP change
which the old version can be deleted. The other case is the one of implementation strategies).
non-compatibility, for which the old version of an entity can be
deleted only after the process instances affected by change have 3.2.4. Phase: migrate run-time instances (compatibility case)
been brought to a safe state in respect of the proposed change. If a new entity chenew is fully compatible with the ERP system
ERPSYS, or at least semantically and syntactically similar to the old
3.2.3. Phase: implement change (compatibility and non-compatibility version cheold to a certain acceptable degree, then we assume that it
case) is able to seamlessly substitute the existing entity without any
This task implies that the new entity in a proposed change ch is further specic impact.
implemented in the ERP system, e.g., a new version of the Supplier Migrating a design-time entity, e.g., a business process, to a new
PO business object is congured or a new process for customer entity, e.g., a business object, means to make sure that an entity
orders is designed and deployed in our running example. uses the new entity in any future instance. For example, a
Note that, in practice, implementing a new entity in an ERP compatible update of the Supplier PO business object in our
system may range from changing a few parameters in the system running example implies that all running instances of the

Fig. 4. Running example (run-time scenario).


M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538 31

procurement process can start seamlessly using the new version of purpose of the Safely complete and migrate run-time instances
the Supplier PO and that all future instances utilise the new Supplier phase is to identify critical process instances and to bring them to a
PO version. A compatible update of the Supplier PO may concern, safe completed execution. This is achieved by ensuring that the
for instance, adding a new eld to capture more details about the entity cheOLD is not deleted until all critical process instances have
supplier identity. Running instances are not likely to be affected by reached past their respective critical point cp.
this change, since they can simply ignore the new eld, while new Fig. 4 shows a possible run-time scenario in the running
instances are able to use the new version of the Supplier PO without example. There are 4 running instances of the procurement process,
any further issue. pinp;1 to pinp;4 . The last function in this process using, as either
In case of non-compatibility, the critical phase in Fig. 3 is the input or output, the entity Supplier PO involved in the change ch is
one regarding the safe termination of currently affected running the function Verify Supplier Invoice. Therefore, according to the
instances: denition given above, the activity Verify Supplier Invoice is the
critical point for all the instances of the process procurement.
3.2.5. Safely complete and migrate run-time instances (non- Instances of the process procurement are safe if their execution is
compatibility case) beyond the Verify Supplier Invoice activity, and critical otherwise.
Non-compatibility of a change ch with the existing system Fig. 4a shows the case of individual process instances, that is,
ERPSYS implies that currently running instances affected by a instances that have not been created because of sub-processes of
change ch may not be able to terminate their execution because other processes. Instance pinp;2 is safe, since its execution has
unable to use a new version chenew of an entity. already passed the verify supplier invoice function. Instance pinp;1 ,
In this case, for each affected running instance, we dene a on the contrary, is not safe. Its critical point is then set to the Verify
critical point cp of execution with respect to the proposed change Supplier Invoice function, which still has to be executed.
ch. This critical point cp is dened as the last activity in a process Instances pinp;3 and pinp;4 exemplify the case of hierarchical
instance that uses the entity che involved in the change ch. process integration. For the sake of simplicity, we consider only
We say that a process instance affected by change is safe if its one process in the hierarchy in this example, that is, instances of
execution is currently beyond the critical point cp determined by the procurement process pinp;3 and pinp;4 are generated by
the change ch, while it is not safe (or critical) otherwise. The instances of the sales process, i.e., pins;1 and pins;2 , respectively.

Fig. 5. Duplication of calculate critical points.

Fig. 6. Timeline of safely migrating critical instances.


32 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

In the case of hierarchical process integration (see Fig. 4b), been qualitatively appraised by ERP domain expert during our
determining the safeness of a process instance in respect of change evaluation, it is not the purpose of this paper to propose validated
ch may be more elaborated. Specically, two cases may occur. measurement scales for these metrics. We only aim at showing
In the rst case, the execution of the called process instance, i.e., that our framework enables the denition of useful metrics and we
pinp;3 in Fig. 4b, is already beyond the critical point. In this case, the leave the validation of their measurement scale and their
called instance and all the instances calling it in the hierarchy are application for advanced risk management in ERP post-imple-
safe. mentation to future work.
In the second case, e.g., instance pinp;4 , the execution of the We distinguish three level of impact of a given change ch, i.e.,
called process is not beyond the critical point. In this case, the system, phase, and entity.
instance is not safe and a critical point has to be set for it and, At the entity level the impact of a change ch is broken down into
recursively, for all the instances in the process integration the percentage of processes, function, objects, process instances,
hierarchy that use this instance. In our example, therefore, the function instances or documents that are potentially affected by ch.
instance pinp;4 is not safe. Its critical point is set to the Verify That is, we dene the following metrics at the component level:
Supplier Invoice activity, whereas the critical point of pins;2 , which jfpro 2 PROjinvolvesch; progj
represents in this case the process integration hierarchy, is set to IMPP ch
jPROj
the procurement sub-process.
Note that, in case a calling process instance also utilises the
entity che involved in the change at some point, then its critical jff un 2 FUNjinvolvesch; f ungj
point is set, following the denition, as the last point in the process IMPF ch
jFUNj
using che. Fig. 5 shows a modied version of our running example
referring to this case. The critical point of the calling instance pins is
set to the function Pick and Pack Goods, since it is the last one to
jfobj 2 OBJjinvolvesch; objgj
utilise the Supplier PO object in the process. IMPO ch
jOBJj
As a result, a list of critical affected process instances is
identied with a respective critical point. While the safe process
instances can be immediately migrated to the version of the entity
jfpin 2 PINjaf f ectsch; pingj
chenew involved in the change ch, the critical affected instances can IMPPIN ch
jPINj
be migrated to the new entity chenew only after their execution has
passed their respective critical point. While determining the
safeness or criticality of an affected process instance is done
jff in 2 FINjaf f ectsch; f ingj
automatically, determining whether the execution of a process IMPFIN ch
jFINj
instance has passed the critical point requires human intervention.
Fig. 6 shows a timeline of the safe completion phase in our
running example. Let us refer to t0 as the time instant in which the
jfdoc 2 DOCjaf f ectsch; docgj
assessment of the safeness or criticality of affected process IMPDOC ch
jDOCj
instance is completed. The safe instances pinp;2 , pinp;3 and pins;1
can be immediately migrated to the new entity chenew. For the Metrics at the entity level are aggregated using SAW (Simple
critical process instances pinp;1 , pinp;4 and pins;2 , we have to wait for Additive Weighting) [21]. The rst level of aggregation is
their execution goes beyond the identied critical point before associated to the phase dimension, that is, we distinguish among
migrating. Let us assume that this happens at times tp;1 , tp;4 , ts;2 , impact at design-time (IMPDT ) and run-time (IMPRT ) as follows:
respectively. The maximum of this time instants, i.e., tp;1 in our X
IMPDT ch vl  IMPl ch
example, determines the instant at which the old entity cheold can
l2fP;F;Og
be deleted. After tp;1, in fact, the entity cheold is not needed by any
remaining critical process instances to safely terminate their
execution. X
IMPRT ch vm  IMPm ch
m2fPIN;FIN;OBJ g
3.2.6. Phase: delete old components (compatibility and non-
compatibility cases) where weights vl and vm capture the relative importance of
Finally, once the change has been implemented and all the impact of different classes of entities in determining impact in a
affected running instances are migrated to new versions of the specic phase.
entities, then the old entities can be deleted from the system. Impact assessment metrics can be further aggregated at the
Similarly to the case of implementation, this task may range for system level. The global impact of a proposed change on an ERP
changing a few parameters in the ERP system conguration, to a system can be expressed as:
complex operation affecting the code base of the ERP system.
IMPSYS ch vDT  IMPDT ch vRT  IMP RT ch
3.3. Impact assessment metrics where the weights vDT and vRT capture the relative importance of
a change impact on the design-time and run-time entities in an
Decisions about change implementation are eventually taken ERP system, respectively.
by project managers and executives [14] who may not have specic Values of the weights characterising the impact assessment
knowledge about the internal structure and the runtime landscape metrics can be determined and validated by surveying domain
of an ERP system. As such, information obtained from impact experts or by the decision makers responsible for the impact
analysis should be synthesised into a compact set of impact analysis. The former case applies in the case of standard change
assessment metrics for decision makers. In this regard, our scenarios, which may be relevant to most organisations using an
framework denes a set of hierarchical metrics to quantify the ERP system, whereas the latter method is preferable when the
impact of a proposed change ch. Even though these metrics have
M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538 33

Table 2
Impact metrics values in running example.

Metric value Notes


P
IMP ch 2=3 0:66 The change does not affect the production process

IMPF ch 3=16 0:19 3 functions of the procurement process are affected by change because using Supplier PO (create purchase
order, receive goods, verify supplier invoice)
O
IMP ch 1=14 0:07 The Supplier PO business object is the only one involved in the change

IMPPIN ch 6=6 1:0 All process instances are affected by the change ch
FIN
IMP ch 12=46 0:26 3 function instances in each of the 4 procurement instances are affected by change

IMPDOC ch 3=38 0:08 Only the Supplier PO documents are involved in the change

IMP DT ch 0:7  0:66 0:2  0:19 0:1  0:07 0:507 Decision makers give more importance to impact at the process layer vp 0:7 as compared to functions
and business objects (vf 0:2 and vo 0:1)
IMP RT ch 0:45  1:0 0:45  0:26 0:1  0:08 0:575 Decision makers give more importance to run-time impact of process and function instances
(vPIN vFIN 0:45) as opposed to documents (vDOC 0:1)
IMPSYS ch 0:3  0:507 0:7  0:575 0:5546 Decision makers give more importance to the impact on ongoing operations (vRT 0:7) as opposed to the
structure of the ERP system (vDT 0:3)

Fig. 7. Architecture of software tool.

perception of change impact is driven by organisation specic the status of execution of currently running process instances in
concerns, e.g., industry, type of processes or products. the ERP system. The former facilitates the denition of the ERP
In our running example, the design-time scenario (see Fig. 3) dependency model, whereas the latter is required by the impact
involves 3 business processes, i.e., sales, procurement and produc- analysis mechanism (see Section 3) to determine the state of
tion, 14 business objects and 16 functions (not including the execution of running instances affected by change when deter-
functions of the production process, which is not expanded in mining the safeness of currently affected process instances.
Fig. 3). In the run time scenario (see Fig. 4), there are 6 process External business intelligence tools can access information about
instances, 46 function instances and 38 documents involved, impact analysis reports and metrics to run more advanced impact
assuming that each business object is instantiated once by process analytics.
instances. The values of the metrics for the change ch The tool is implemented using a state of the art model driven
SupplierPOOLD ; SupplierPONEW ; update are calculated in Table 2 application development platform for enterprise system integra-
tion.1 The tool has been instantiated for the pseudo-real ERP
4. Tool support systems SAP ERP 6.0 for GBI Inc. and Microsoft Dynamics NAV R2
2013 for Cronus Inc. Both companies have been developed for years
To show the feasibility of the approach, our framework has been by the respective software vendor for professional training
embedded in a software tool, i.e., a decision support system purposes. Although ctitious, their ERP instantiations contain a
supporting business analysts in the assessment and implementa- sufcient level of detail, in terms for instance of number and type
tion of ERP post-implementation changes. of completed and open orders, products, clients, suppliers, and
The architecture of the tool is shown in Fig. 7. processes, to simulate the operations of a complex medium to large
The tool interacts with two types of external software packages, organisation at a given instant in time. GBI Inc. is a typical large-
i.e., the ERP system and external analytic tools. As far as the ERP sized manufacturing company, assembling products from spare
system is concerned, as previously remarked, we adopt a non- parts, whereas Cronus Inc. is a medium-sized company trading a
invasive approach to impact analysis, whereby we cannot assume variety of products. Since we did not have access to the ERP APIs,
to have control over the internal implementation of the ERP the mapping of the two case onto our tool has been performed
system. The requirements imposed by our framework on the ERP manually. In both instantiations, we consider only sequential
system are (i) a reective interface to provide information about business processes, that is, processes constituted by a sequence of
design-time components, e.g., list of business processes, functions
and data objects and (ii) a management interface allowing to query
1
www.mendix.com.
34 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

Fig. 8. Snapshots of software tool.

function invocations. This simplifying assumption facilitates the applying the impact analysis mechanisms, i.e., the list of entities
calculation of critical points during impact analysis and has affected (run-time entities are shown on the right-hand side) and
allowed quicker understanding of the framework functioning by values of assessment metrics (left-hand side).
practitioners during the evaluation.
Fig. 8 shows snapshots of the implemented tool in the case of 5. Evaluation
SAP ERP 6.0 GBI Inc. Fig. 8a shows the interface listing all post-
implementation modications implemented so far and the input We provide two different types of evaluation of our tool. First, to
interface for a new proposed change. Fig. 8b shows the results of show its feasibility in practice, we evaluate to what extent our tool
M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538 35

Table 3
SAP coverage analysis.

Enterprise SAP Features Manageable by impact Comments


Level analysis framework?
Organizational Authorization/Roles and X Out of the scope of analysis
Layer Responsibilities
Organizational Level (Client, X Out of the scope of analysis
Company Code, Plant, etc.)
Presentation Graphical User Interface X Out of the scope of analysis.
Layer Transaction Data Header/Item
Business Logic Module U Framework can identify which modules are affected by change as a result of modications
Layer of functions constituting a module.
(design- Business Processes U Framework can identify which business processes are affected by change and can handle
time) Sub-process Integration U sub-process integration.
Business Function/Transaction U Framework can identify which business transactions are affected by change.
Business object Master Data U Framework can identify the impact of change of master data as business objects
Layer Organisational Data X Out of the scope of analysis
(design- Transaction Data X Out of the scope of analysis (post-implementation changes do not deal with ad-hoc
time) changes to running process instances, which manipulate transaction data)
Attributes of Data Item X Framework cannot differentiate the change impact of individual attributes of a business
object.
Operation Active Business Process U Framework can identify which process instances are affected by change.
Layer Active Business Transaction U Business transactions affected by change can be identied by the framework based on the
(run-time) process instances affected by change.
Active Transaction Data U Transaction data affected by change can be identied by the framework based on the
process instances affected by change.

Table 4
MS Dynamics NAV coverage analysis.

Enterprise MS Dynamics NAV features Manageable by Comments


Level impact analysis
framework?
Organizational Authorization/Roles and Responsibilities X Out of the scope of analysis
Layer
Presentation Graphical User Interface X Out of the scope of analysis
Layer Fact Box/Sorting/Navigation/Field
Business Logic Department U Framework can identify which modules are affected by
Layer change as a result of modications of functions
(design- belonging to a Department.
time) Business Processes U Framework can identify which business processes are
Business Process Integration U affected by change and can handle sub-process
integration.
Business Function/Action U Framework can identify which business actions are
affected by change.
Business Master Data: U Framework can identify the impact of change on master
Object Layer data as business objects
(design- - Chart of Account
time) - Cards (Vendor, Customer, Item)
- Orders (sales, Production, Purchase)

Organisational Data X Out of the Scope of analysis


Journals X Framework cannot specify what journal will be
(General, Sales, Purchase, Item) impacted by modication of data.
Attribute of Card X Framework cannot differentiate the change impact of
individual attributes of a business object (or card).
Operation Active Business Process U Framework can identify which process instances are
Layer affected by change.
(run-time) Active Business Transaction U Business transactions affected by change can be
identied by the framework based on the process
instances affected by change.
Active Transaction Data U Transaction data affected by change can be identied by
the framework based on the process instances affected
by change.
Posted Documents U Framework can identify posted documents affected by
change, since posted is only a particular state of a
business object.

is able to handle standard functionality provided by commercial consulting). Then, we present the results of an experiment with
ERP packages. This evaluation activity has involved a panel of 7 ERP ERP users to assess the t for purpose and usability of our tool in
professionals with average 10.7 years experience with different practice. This evaluation activity has involved the panel of 7 ERP
types of ERP systems (SAP, Oracle/JD Edwards, and Microsoft professional and 12 master students in Business Systems Design
Dynamics) in different industries (banking, higher education, ERP
36 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

who recently completed an introductory course on business authors and then discussed with the panel of ERP professionals to
process management and enterprise systems. achieve substantial agreement.
As far as the coverage analysis for SAP and MS Dynamics NAV is Tables 3 and 4 show the results of the coverage analysis. The
concerned, for each system, we rst have identied a set of coverage of the two systems, i.e. SAP and MS Dynamics NAV,
standard features. Then, we have evaluated whether our tool is appear to be aligned. More specically, our framework is able to
able to handle changes of the identied features. This process has deal with change in the core features of the application logic of an
been executed following a customised Delphi method, whereby ERP system, that is, business processes, related functions and
the list of ERP package features and the matching with the business objects. The framework is not able to address change in
functionality in our framework have been identied rst by the the presentation layer or at the level or roles and responsibilities in

Table 5
Results of experiment. (For interpretation of the references to color in the table, the reader is referred to the web version of this article.)
M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538 37

the organisational layer. We argue that change in the presentation 6. Conclusions


layer is always present when the core functionality of an ERP
system. It is related to issues of usability and acceptance [15], but it This paper has presented a framework, i.e., a set of methods and
does not affect the running operations of an organisation. Impact of a tool, to support stakeholders such as business analysts in
change on organisational governance, i.e., roles and responsibili- assessing the impact of post-implementation changes to an ERP
ties, is subject of our ongoing work. system. The framework denes a meta-model of the dependencies
As far as the business object layer is concerned, for both among ERP entities, a mechanism to assess the impact of proposed
systems, our framework covers post-implementation changes of changes and a set of metrics to synthetically quantify the scope and
master data, while organisational and transactional data are not depth of ERP changes. The implemented tool has been instantiated
covered. Master data are the ones more likely to evolve in the post- in the case of two commercial ERP software packages, i.e., SAP ERP
implementation phase, as they capture the concepts relevant for an 6.0 and Microsoft Dynamics NAV 2013. Based on the instantiations,
organisations daily operations, such as customers, suppliers, our framework is able to assess the impact of changes of the main
materials, and products. Organisational and transactional data are entities of an ERP system, such as business processes and functions,
not covered because they are of out of the scope of our analysis. In but it still fails to capture change at the level of roles and
particular, we argue that the organisational structure is likely to be responsibilities in the organisational. The evaluation of the tool
stable during the lifetime of an ERP system or, at least, it changes with ERP experts has revealed that our framework provides an
much less frequently than master data. Transactional data are out effective solution to an issue that is relevant for practice.
of scope since, as discussed before, impact analysis is not focused As suggested by the participants to the evaluation, our
on ad-hoc changes to running processes and their data, but rather framework can be extended by considering impact of post-
on changes to the static structure of the ERP system. implementation changes at the organisational level and the
The experiment with ERP professionals and students has been nancial implications of a change. Change at the organisational
executed in individual working sessions. In each session we rst level, in particular, entails analysing the impact of change on the
have presented the functionality of the tool using a running ERP responsibility and authorisation policies, which are closely
example of change scenarios, then asked the participant to run two knitted to the issue of security of an ERP system [35].
change scenarios independently, and then collected feedback from From a methodological point of view, while in this work we took
the participant using a structured questionnaire. The items of this a model-driven approach to impact analysis, future work may
questionnaire relate to typical dimensions of technology accep- investigate ERP post-implementation impact analysis from an
tance and success evaluation, i.e., system usefulness, ease of use, evidence-based perspective. Rather than modelling explicitly ERP
ease of learning and satisfaction with the system [1,24]. All items entities and their dependencies, these may be learned from the
are evaluated on a 5 point Likert scale from strongly disagree (1) to execution logs of ERP systems. In this way, impact analysis can be
strongly agree (5). based on actual dependencies learnt from process execution data.
While we selected ERP professionals because they represent the Based on execution frequency and other possible information in
natural type of users of our tool in the real world, the student group execution logs, such as resource or cost information, impact
has been selected because less likely to suffer from attribution bias analysis can be also characterised more in depth by dening other
during the evaluation. We argue, in fact, that the personal measures of dependencies among, such as strength or cost.
experience of ERP professionals in managing post-implementation
may have biased their understanding and utilisation of our Acknowledgment
framework. The students, on the other hand, possess a sufcient
knowledge background to understand the issues associated to ERP This work was supported by the 0000 Research Fund (Project
post-implementation change, but they do not have direct number 1.160044.01) of UNIST (Ulsan National Institute of Science
experience with these in the real world that may bias their and Technology).
experience with our tool.
Table 5 shows the results of the evaluation. The feedback References
obtained from both groups of participants is overall positive. The
few unsatisfactory results are found in the evaluation of ease of [1] Y.M. Abgaz, Change Impact Analysis for Evolving Ontology-based Content
Management, Dublin City University, 2013 (PhD Dissertation).
use. Because of its proof of concept nature, our prototype was not [2] D. Aloini, R. Dulmin, V. Mininno, Risk management in ERP project
developed considering usability as primary non-functional introduction: review of the literature, Inf. Manage. 44 (6) (2007) 547567.
requirements. Therefore, lower scores for the ease of use criteria [3] S. Ahmadi, E. Papageorgiou, C.H. Yeh, R. Martin, Managing readiness-relevant
activities for the organizational dimension of ERP implementation, Comp. Ind.
have been expected. 68 (2015) 89104.
It is remarkable that the usefulness evaluation (in particular [4] R.C. Beatty, C.D. Williams, ERP II: best practices for successfully implementing
items C4C6) and the satisfaction with the tool (items C17C20) an ERP upgrade, Commun. ACM 49 (3) (2006) 105109.
[5] P.J. Bhattacharya, P.B. Seddon, R. Scheepers, Enabling strategic transformations
are generally high for both students and professionals. This signals
with enterprise systems: beyond operational efciency, Proc. Int. Conf. Inf.
that, on the one hand, the tool addresses a relevant problem in Syst. 2010 (2010) 5567.
practice and, on the other hand, it provides an effective solution to [6] S.A. Bohner, Extending software change impact analysis into cots components,
Proceedings of the 2002 Software Engineering Workshop, in Conjunction with
it.
27th Annual NASA Goddard/IEEE Conference (2002) 175182.
The student group has provided generally higher evaluations. [7] L. Brehm, A. Heinzl, M.L. Markus, Tailoring ERP systems: a spectrum of choices
This can be explained by their relative lack of experience with and their implications, Proceedings of the 34th Annual Hawaii International
utilising ERP systems in the real world, which has prevented them Conference on System Sciences, January (2001) 19.
[8] A. Cataldo, R.J. McQueen, J. Hardings, Comparing strategic IT alignment versus
to realise the nuances of the issues that can be encountered in ERP process IT alignment in SMEs, J. Res. Pract. Inf. Technol. 44 (1) (2012) 4355.
post-implementation change management in real world scenarios. [9] Roberto Da Silva, Raquel Stasiu, Viviane Moreira Orengo, Carlos A. Heuser,
In this regard, the expert participants stressed that the tool could Measuring quality of similarity functions in approximate data matching, J.
Infometrics 1 (1) (2007) 3546.
be extended with more advanced functionality to assess the [10] H.K. Dam, L.-S. L, A. Ghose, Managing changes in the enterprise architecture
organisational impact of ERP changes and the nancial impact of modelling context, Enterp. Inf. Syst. 10 (6) (2016) 131.
proposed changes based on different possible implementation [11] F.S. de Boer, M.M. Bonsangue, L.P.J. Groenewegen, A.W. Stam, L. van der Torre,
Change impact analysis of enterprise architectures, Proceedings of the 2005
strategies.
38 M. Parhizkar, M. Comuzzi / Computers in Industry 84 (2017) 2538

IEEE International Conference on Information Reuse and Integration (2005) [29] C.S.P. Ng, G.G. Gable, T. Chan, An ERP-client benet-oriented maintenance
177181. taxonomy, J. Systems and Software 64 (2) (2002) 87109.
[12] R. Dijkman, M. Dumas, B. Van Dongen, R. Krik, J. Mendling, Similarity of [30] T. Oseni, M.M. Rahim, S.P. Smith, S. Foster, An initial empirical evaluation of the
business process models: metrics and evaluation, Inf. Syst. 36 (2) (2011) 498 inuence of ERP post-implementation modications on business process
516. optimisation, Proceedings of the 2014 European Conference on Information
[13] N. Dor, T. Lev-Ami, S. Litvak, M. Sagiv, D. Weiss, Customization change impact Systems (2014) (Retrieved from http://aisel.aisnet.org/ecis2014/proceedings/
analysis for ERP professionals via program slicing, Proceedings of the 2008 track06/4).
International Symposium on Software Testing and Analysis (2008) 97108. [31] K. Park, A. Kusiak, Enterprise resource planning (ERP) operations support
[14] W.W. Eckerson, Performance Dashboards: Measuring, Monitoring, and system for maintaining process integration, Int. J. Prod. Res. 43 (19) (2005)
Managing Your Business, John Wiley & Sons, 2010. 39593982.
[15] Y.M. Ha, H.J. Ahn, Factors affecting the performance of enterprise resource [32] P. Plebani, B. Pernici, URBE: web service retrieval based on similarity
planning (ERP) systems in the post-implementation stage, Behav. Inf. Technol. evaluation, IEEE Trans. Knowl. Data Eng. 21 (11) (2009) 16291642.
33 (10) (2014) 10651081. [33] J. Ram, D. Corkindale, M.L. Wu, Implementation critical success factors (CSFs)
[16] A.E. Hassan, R.C. Holt, Predicting change propagation in software systems, for ERP: do they contribute to implementation success and post-
Proceedings of the 20th IEEE International Conference on Software implementation performance? Int. J. Prod. Econ. 144 (1) (2013) 157174.
Maintenance, September (2004) 284293. [34] P. Samaranayake, Business process integration, automation, and optimization
[17] A. Haug, J. Stentoft Arlbjrn, A. Pedersen, A classication model of ERP system in ERP, Bus. Process Manage. J. 15 (4) (2009) 504526.
data quality, Ind. Manage. Data Syst. 109 (8) (2009) 10531068. [35] W. She, B. Thuraisingham, Security for enterprise resource planning systems,
[18] P. Inedo, B. Rapp, A. Inedo, K. Sundberg, Relationships among ERP post- Inf. Syst. Secur. 16 (3) (2007) 152163.
implementation success constructs: an analysis at the organizational level, [36] P. Soffer, B. Golany, D. Dori, ERP modeling: a comprehensive approach, Inf. Syst.
Comp. Hum. Behav. 26 (5) (2010) 11361148. 28 (6) (2003) 673690.
[19] I. Ivkovic, K. Kontogiannis, Tracing evolution changes of software artifacts [37] P. Soffer, B. Golany, D. Dori, Aligning an ERP system with enterprise
through model synchronization, Proceedings of the 20th IEEE International requirements: an object-process based approach, Comp. Ind. 56 (6) (2005)
Conference on Software Maintenance (2004) 252261. 639662.
[20] A. Krishna, S.A. Vilkomir, A.K. Ghose, Consistency preserving co-evolution of [38] C.F. Strnadl, Aligning business and it: the process-driven architecture model,
formal specications and agent-oriented conceptual models, Inf. Software Inf. Syst. Manage. 23 (4) (2006) 6777.
Technol. 51 (2) (2009) 478496. [39] P.P. Tallon, A process-oriented perspective on the alignment of information
[21] C.L. Hwang, K. Yoon, Multiple Attribute Decision Making: Methods and technology and business strategy, J. Manage. Inf. Syst. 24 (3) (2014) 227268.
Applications a State-of-the-art Survey, Springer Science & Business Media, [40] M. Themistocleous, Z. Irani, R.M. O'Keefe, R. Paul, ERP problems and
2012. application integration issues: an empirical survey, Proceedings of the 34th
[22] C. Lopez, J.L. Salmeron, Dynamic risks modelling in ERP maintenance projects Annual Hawaii International Conference on System Sciences (2001) 1018.
with FCM, Inf. Sci. 256 (2014) 2545. [41] R. Yu-Yuan Hung, T. Chung, B. Ya-Hui Lien, Organizational process alignment
[23] J. Luftman, R. Papp, T. Brier, Enablers and inhibitors of business-IT alignment, and dynamic capabilities in high-tech industry, Total Qual. Manage. Bus.
Commun. AIS 1 (1999) (Article 11). Excellence 18 (9) (2007) 10231034.
[24] Arnold M. Lund, Measuring Usability with the USE Questionnaire12. 2001 [42] A. Van Lamsweerde, R. Darimont, E. Letier, Managing conicts in goal-driven
Retrieved on May 1st 2016 from: http://garyperlman.com/quest/quest.cgi? requirements engineering, IEEE Tran. Software Eng. 24 (11) (1998) 908926.
form=USE. [43] E. Visser, J. Warmer, A. Van Deursen, A. Van Deursen, Model-driven software
[25] P.D. Manuel, J. AlGhamdi, A data-centric design for n-tier architecture? Inf. Sci. evolution: a research agenda, Proceedings of the International Workshop on
150 (3) (2003) 195206. Model-Driven Software Evolution in Conjunction with the 11th European
[26] N. Milanovic, M. Malek, Current solutions for web service composition, IEEE Conference on Software Maintenance and Reengineering (2007) 715.
Internet Comput. 8 (6) (2004) 5159. [44] J.H. Worley, K.A. Chatha, R.H. Weston, O. Aguirre, B. Grabot, Implementation
[27] P.A. Millet, Toward a model-driven, alignment-oriented ERP methodology, and optimisation of ERP systems: a better integration of processes, roles,
Comp. Ind. 64 (4) (2013) 402411. knowledge and user competencies, Comp. Ind. 56 (6) (2005) 620638.
[28] M. Moalagh, A.Z. Ravasan, Developing a practical framework for assessing ERP [45] Y. Zhu, Y. Li, W. Wang, J. Chen, What leads to post-implementation success of
post-implementation success using fuzzy analytic network process, Int. J. ERP? An empirical study of the Chinese retail industry, Int. J. Inf. Manage. 30 (3)
Prod. Res. 51 (4) (2013) 12361257. (2010) 265276.

Vous aimerez peut-être aussi