Towards a Software Framework for Automatic Business Process Redesign

ACEEE Int. J. on Information Technology, Vol. 02, No. 01, March 2012

Towards a Software Framework for Automatic
Business Process Redesign
Marwa M.Essam1, Selma Limam Mansar2
1
Faculty of Information and Computer Sciences –Ain Shams University/Cairo, Egypt
Email: marwa.essam@gmail.com
2
Carnegie Mellon University-Qatar /Information Systems Program, Doha, Qatar
Email: selmal@qatar.cmu.edu

Abstract— A key element to the success of any organization is adaptation of the supporting IS.
the ability to continuously improve its business process Currently, most BPR projects depend mainly on an expert
performance. Efficient Business Process Redesign (BPR) to an organization. Experts/ Consultants may use some tools
methodologies are needed to allow organizations to face the for process modeling, business planning or process
changing business conditions. For a long time, practices for
prototyping [4]. However, there is currently no tool that
BPR were done case-by-case and were based on the insights
and knowledge of an expert to the organization. It can be supports the automatic redesign from the old business
argued that efficiency, however, can further be achieved with process to new innovative business processes.
the support of automatic process redesign tools which are few In recent years, process mining was introduced in the
at the moment. Process mining as a recent approach allows context of business process management [5]. Process Mining,
for the extraction of information from event logs recorded in similarly to data mining, allows for the extraction of
different information systems. In this paper we argue that information from event logs recorded in BPM-based systems.
results driven by process mining techniques can be used to Some of the possibilities offered by process mining results
capture the various types of inefficiencies in the organization are the discovery of new business process models, the
and hence propose efficient redesigns of its business model.
checking of the conformance to some prescriptive or
We first give an outline on the current directions towards
automatic BPR followed by a review on the different process descriptive models, or the extension of an initial model with
mining techniques and its usage in different applications. analysis data.
Then, a specific framework of a Software tool that uses process In this paper we argue that an evolutionary redesign to
mining to support automatic BPR is presented. business processes can be reached using results driven by
process mining techniques. The evolutionary redesign is
Index Terms— Process Mining, Business Process Redesign, based on the application of general best practices or heuristic
Business Process Management rules to an existing design. We think that realizing adaptations
to business process has become a difficult task to accomplish
I. INTRODUCTION due to the lack of knowledge to customize the process logic
A business process is a collection of related, structured at a sufficient level. However, using process mining, different
activities that produce a service or product that meets the models can be extracted from the reality logs and various
needs of a client. Business processes are critical to any types of inefficiencies in the organization can be captured by
organization as they generate revenue and often represent a analyzing these logs. These results can be used as an input
significant proportion of costs. Nowadays, many Business base for a tool that suggests efficient redesigns to the
Process Management (BPM) systems exist in the market (Ex.: business process, hence, providing consultants and experts
FileNet and Ultimas). BPM systems provide organizations with a vision on how to get from the old process to the new
with a broad range of facilities to design, enact, control and process.
analyze their business process [1]. A list of the some of the This paper is arranged as follows: Section II introduces
cross-industry BPM suites with their relative strengths can the notion of modeling a business process with an example
be found in [2]. on a credit application process. Section III discusses some
Despite its popularity and obvious pay-offs, the current the related work towards converting old process designs to
practices for monitoring and analyzing the execution of BPM new designs. In Section IV we give a review on the different
systems in the organizational reality still leaves a lot to be process mining algorithms and some of its application areas.
desired [3]. There is a vital need for BPM systems to In Section V, we discuss our vision on how process mining
satisfactory support Business Process Redesign (BPR). In can be used in a software tool that supports the automatic
many cases, the developed business functions do not redesign of business processes.
effectively reflect the actual business process. Many of the
implemented business functions are never used. Other II. BUSINESS PROCESS MODELING
business functions provide more functionality than actually A group of related tasks that together create value for a
needed. Another issue is related to the evolution of business customer is called a business process. Different modeling
processes and their variability. In many domains (ex: languages/techniques can be used to represent different
healthcare), frequent process changes requires the continuous
© 2012 ACEEE 23
DOI: 01.IJIT.02.01. 42


Figure 1. Credit Application Process Model
aspects of the business process. Control Flow modeling and is refined on the basis of what is called “redesign best
languages, like Petri-nets, represent the different activities in practices” [8]. A redesign best practice describes a well-tried
the process with constraints to control the execution between way to remove a particular problem from a process to improve
them (In what order activities are executed). Data models like its performance. An Example of a best practice is: Eliminating
Entity Relationship Diagrams represent the data organization unnecessary tasks from a process (the tasks with no added
in a process. Organizational models like organizational charts value for customers). In [8], an extensive literature survey
represent the structure in which the business process will has taken place to collect all best practices for evolutionary
execute (the entities/roles that can perform work for the process improvement.
enterprise). An Example of a credit application business Towards developing this “intelligent” tool and based on
process is shown in Fig.1. For space limitation we only show the aforementioned evolutionary approach for BPR, a new
the control flow of the process. The process is modeled using technique to find process design alternatives was proposed
Petri-Nets [6]. It begins with the recording of the application in [9]. In this technique, a business process first is put in a
where the client expresses an interest in acquiring credit. formal process definition defined by the authors called Proto
This stage includes the presentation of the application, and Net. A set of process measures are then calculated on the
the required documents to the organization for verification. process (Ex: Level-of-Control, which is the percentage of
This is followed by an analysis or study of the credit control/decision tasks). The authors specified 18 different
application to decide whether to accept/ reject the credit. process measures to be calculated on the process design
The client is notified in case of rejection. In case of acceptance, under study. The calculated measures are then compared
the credit is disbursed to the client by either a credit transfer against a set of condition statements that when evaluated to
to a bank account or by check. true a “redesign” best practice is selected to be applied on
the process model (Ex: Apply Task Elimination if level of
III. REVIEW OF TECHNIQUES TOWARDS AUTOMATIC BPR Control >0.2). Cutoff values for condition statements were
determined by the authors’ expertise in the field.
A BPR project starts when an organization is faced with
Although this technique suggests the use of some best
the need to change its business process to make
practices in the process model (the ones that their condition
improvements in its quality, cost, service, lead-times,
statements evaluated to true), it didn’t specify exactly how
outcomes or flexibility. Usually, BPR projects are carried out
these practices will be applied. As pointed out in [10], a
by setting up workshops within the organization to think of
redesign best practice just provides directions on how the
alternatives to the business process. Consultants, employers,
redesign should be performed. When we look at the
managers and specialties participate in these workshops to
parallelism best practice, for instance, it is suggested that the
make process redesigns. Some Software tools may be used
redesign should have more tasks in parallel than the original
within these workshops to aid the redesign process. However,
process. But it does not tell us to put tasks A, B and of
the identification of the problem areas and the opportunities
process X in parallel. In [10], four exact transformations were
of change are totally determined by the workshop group. As
suggested to be applied on selected process parts of the
a result to this manual approach, the outcome redesigns are
process model to produce different redesigns. The input
often subjective and non sustainable. This is because it is
process model is assumed to be a Petri-net with some
strongly influenced by the individual expertise of the
extensions like data dependencies and roles.
workshop group and may not suit the actual case of the
The proposed transformations are:
organization under study.
Unfolding of tasks, in which aggregated tasks (upper level
To make process redesigns that actually succeed on the
tasks that are modeled in a detailed sub-process) are split up
long term in improving the business process, the research in
into several smaller tasks.
BPR was recently directed towards making automatic
Parallel Transformation, in which tasks that do not have
business process redesigns. In [7], a research project was
data dependencies are executed in parallel.
proposed to address the lack of tools in industry with
Sequence Transformation, in which all parallel paths in a
“intelligent” capabilities to suggest favorable alternatives to
model are transformed into sequence path, provided that the
an existing process design. To develop such an “intelligent”
output sequence will have the lowest throughput time and
redesign tool, the project proposed the idea of making
will contain no errors related to data dependencies.
evolutionary, local updates to an existing workflow design to
Merging transformation, in which a task cluster executed by
gradually improve its performance. In this evolutionary
the same role is merged into one aggregated task.
approach, the existing process is taken as a starting point
© 2012 ACEEE 24
DOI: 01.IJIT.02.01.42


For each of the above transformations, the authors specified This is due to the lack of knowledge on what causes the low
some formal characteristics that the selected process part performance within the process. In the next sections, we will
should have in order to be eligible for the transformation (Ex: show how process mining can be used to enrich the business
A selected part for the parallel transformation must not have process model with the required knowledge to satisfactory
a selective route to prevent contradictions with the selection’s allow automatic process transformations.
original purpose). They specified also how exactly the trans-
formed part will be replaced in the Petri net model (Ex: Re- IV. BUSINESS PROCESS MINING
moving unnecessary edges/tasks). However, still their tech-
A. Introduction
nique doesn’t allow the automatic selection of the process
part to be altered or the automatic selection of the transfor- Process Mining allows for the extraction and the analysis
mation that produces the best result. of information from event logs recorded in BPM-based
In [11], an algorithm called the Boolean Verification Algo- systems. Example of event logs are: the audit trails of a
rithm (BVA) was presented for the optimization of workflows. workflow management system, the transaction logs of an
On the contrary to the techniques mentioned before, this enterprise resource planning system, or the electronic patient
approach does not focus on applying best practices to the records in a hospital. The two main components of a record
process design. It just focuses on finding the maximum in any event log are: the event (task/activity) that was
parallelization for a design to reduce the overall execution executed and an identification of the particular instance of
time of the process. For this purpose, BVA uses a method the process within which the activity was executed (case).
called the if-conversion. The main idea of this method is to More information can also be available in the log (for example,
assign Boolean activation conditions to workflow tasks based the timestamp of the event, or the performer of the event, the
on their control flow dependence. While scanning the pro- data elements recorded with the event, etc...).
cess model from the start task , BVA assigns Boolean control To illustrate the concept of process mining, consider the
parameters on different branches and choice nodes (C1, log information recorded in Table I. This log contains
C2,…Cn) and forms a Boolean activation condition for each information on three execution cases (1, 2 and 3). Executed
task (Ex: !C1 & C2 v C1). Tasks are then checked for their events are represented as tasks and are assumed to be
control flow dependencies by analyzing their activations recorded in order. When scanning the log, we can detect four
conditions and tasks with no control or data dependencies different tasks in the process (Tasks A, B, C and D). One can
are parallelized. see also that all cases starts by task A and all cases ends by
Although this algorithm promises to ensure a full task D. In two cases (1 and 3), task B follows task A in the
parallelization of a business process, it still doesn’t relate the execution. In case 3, task C follows task A. From this
reality to the design. In other words, it focuses only on the information, we can simply draw, using Petri Nets, the process
parallelization of tasks while in some cases, a sequential pro- model that corresponds to this log as in fig. 2.
cess may be perceived as a simpler process by employees
and clients. Since the order of the tasks is fixed in sequential
constructs, the execution of the process is done in the most
logical way hence reducing errors. Furthermore, the synchro-
nization that is required after the execution of tasks in parallel
is not necessary in sequential processes.
Let us note now that all the techniques that we men-
tioned above don’t satisfactorily support the automatic pro-
duction of process redesigns. Although they provide guide-
lines on how to apply different transformations on the pro- Figure 2. A process model corresponding to the log
cess model, none of them supports either the automatic se-
lection of the process part to redesign or the best transfor- B. Process Mining Algorithms
mation to apply on it. Over the last decade, many process mining algorithms
has been developed to find process models that successfully
TABLE I. A PROCESS LOG
mimics the behavior registered in the logs. In [12], a control
flow mining algorithm called the alpha algorithm was
presented. The alpha algorithm assumes that there is no noise
(logs containing exceptions) in the data and that the log
contains sufficient information about the workflow (no paths
exist with low probability that prevents them from being
detected). The alpha algorithm scans the log and looks for
causal relations ( a relation between two tasks A and B such
that B is directly followed by A in a log trace and A is never
followed by B). These causal relations are then represented
in a Petri-net model describing the output process.
© 2012 ACEEE 25
DOI: 01.IJIT.02.01. 42


To address the issue of noise existing in the log, an example, execute some tests at earlier phases).
algorithm called the Heuristic Miner was presented in [13]. Healthcare is another famous application domain for
The Heuristic miner follows the alpha algorithm in finding process mining. The applicability of process mining in
causal relations between tasks. However, it calculates healthcare was demonstrated using a real case of a
frequencies for the occurrence of each task and for the relation gynecological oncology process in the AMC hospital in the
between tasks. Cutoff values are then used to eliminate tasks/ Netherlands [18]. The log data contained information about
relations that are a result of noise/ un-complete executions. a representative group of 627 gynecological oncology
Because in some application areas, the resulting models after patients. The goal of using process mining was to discover
process mining become very difficult to understand the care paths followed by individual patients and whether
(Spaghetti-Like), many process mining algorithms based on certain procedures are followed or not. After applying process
clustering was introduced to eliminate unnecessary mining techniques, many useful results became visible to the
information from the output models. The Fuzzy miner [14] for people at the hospital. For Example, it was found that patients
example starts by drawing a very complicated graph with all who undergo several chemotherapy sessions often need to
causal relations found in the log. It then simplifies the graph visit the dietician. This was not immediately clear to everyone
by creating clusters of nodes. Clusters are created by and illustrates the value of creating transparency using
aggregating low significant nodes with their highly correlated process mining.
neighbor nodes. Different significance and correlation The above two mentioned projects were implemented with
matrices can be used for different application purposes. The the process mining tool named ProM [19]. ProM contains
lion’s share of efforts in process mining has been for more than 250 plug-ins that implement different process
discovering control flow models resulting in many other mining algorithms. However, it is not clear how to use ProM
models other than the ones mentioned above. However, in process redesign projects. In the above two projects, the
process mining can also be used to mine much other useful authors used different plug-ins but viewed each plug-in result
information. As mentioned before, the event logs not only alone. Although ProM allows the results from some algorithms
record information about the different cases and the different to be integrated in a Colored Petri Net (CPN) that support
tasks. It also records information on the role who executed analysis and simulation, there was no guidance from ProM
these tasks, the input and output attribute values to and on how to improve the business processes. Instead, the
from each task and the execution start and end time of tasks. researchers concluded the redesign ideas from viewing the
This information can be used to mine the relations between simulated models. i.e. It is hard to make process redesign
the different roles creating an organizational model. It can using process mining a repeatable service.
also be used to analyze the information flow between the
different roles, the interactions between the co-workers, the V. PROPOSED SOFTWARE FRAMEWORK FOR AUTOMATIC BPR
decision points in the models and the performance of
A. Introduction
executions [15], [16].
Based on the discussion in the previous sections, we will
C. Process Mining Application in BPR now focus on two phases in BPR that, up to now, are done
Process mining has been applied in a variety of manually by the designers:
organizations covering many application domains. In [17], - The designer of the new process manually selects
process mining was used to analyze the test process in the process part to be redesigned from the old process model.
ASML. ASML makes so-called wafer scanners that are used He also decides what change to be made on the selected part.
to manufacture processors in devices ranging from mobile - When using process mining for BPR, experts/
phones to desktop computers. Wafer scanners are really researchers in process mining determine which process
complex machines that use a photographic process to image mining algorithm to use and results after modeling do not
nanometric circuit patterns onto a silicon wafer. The testing suggest redesign ideas.
of the manufactured wafer scanners is a time-consuming For the above two points, we present our view on a
process. So, the goal of the analysis was to reduce the testing software that automatically outputs a specific redesign to a
time. business process using its recorded log information as input.
Each wafer scanner in the ASML factory produces a log In Fig. 3, the framework for this software is presented. The
of the software tests that are executed on it. Process mining proposed software framework is composed of two main parts,
was used to visualize the actual flow of the test process and a business process miner and a redesign engine. A Business
confront this visualization with the idealized view of the tests Process miner is responsible simply for applying process
according to engineers. It was found that as soon as one test mining techniques on the input process log to gather
fails, a fix is made to the scanner and all other tests are put on information that will aid the redesign process. The Redesign
hold (idle time) and often after the fix is made, some tests are engine is responsible of generating new redesigns based on
re-executed again. Visualizing this loop-backs caused by some both process mining results and knowledge of the redesign
tests gave the engineers a useful view on what was causing best practices. Next, we illustrate both components in more
the time loss in the test process. Hence, allowed them to detail.
make changes to the test process to reduce the time (for
© 2012 ACEEE 26
DOI: 01.IJIT.02.01.42


Figure 3. A Software Framework for Automatic Business Process Redesign

B. The Business Process Miner Component redesign best practices. Therefore, the proposed software
In the business process miner, a proper selection of must contain a database of redesign best practices that
process mining algorithms is applied on the event logs to contains, for each best practice, an execution rule that shows:
produce different model types. All the results gathered are - The conditions that a process part must match in
then stored in a process mining results database. The choice order to be eligible to apply the best practice on,
of which process mining algorithms to apply on the log will - And, the process transformation that will be applied
be determined in a log inspection phase using some parametric if these conditions are true.
characteristics from the log (for example, a log which contains Matching the selected process part against the conditions in
a large no of tasks such that mining all these tasks will produce the best practices data base allows the redesign engine to
a spaghetti like model will be mined using the fuzzy miner). If find what transformations to apply on it. The selected part is
role information is available in the log, the log will be mined then transformed and integrated with the process model in a
to produce organizational and social network models. This redesign generator element. Since different combinations of
role analysis is important since in some cases, the malfunction transformations may applied to the same process part (more
in a process design is mainly because of bad organizational than one best practice rule evaluate to true), different
or social structures. The log will be mined also to produce redesigns to the process model may be generated. Hence,
performance data like the throughput time of cases, the they are saved in a redesigns database. Moreover, if there
slowest tasks, the delays before tasks execution, the are more process parts that need to be changed, a redesign
resources utilizations,…etc. loopback starts again by selecting another process part to
change. The generated redesigns are then evaluated for the
C. The Redesign Engine Component selection of the best redesign. The evaluation will be based
Using results gathered from the business process miner, on simulating the different redesigns using data from the
the redesign engine starts by determining the malfunction logs (for example, the arrival time of different cases, the routing
part in the process design (a certain component in the process probabilities of different paths in the model, the response
where the mining results show that it somehow causes low time from certain roles,… etc).
performance). This malfunction part selector is the key
element in this proposed software as it integrates the results VI. CONCLUSIONS
from all the process mining algorithms to come out with
conclusions on what causes the low performance of the In this paper we focused on how to automatically redesign
process (For example, a specific path in the control flow model, business processes in order to increase its performance. We
a bad organization structure, etc…) . It is important to note showed that current redesign methodologies let the designer
also that, to select the problem areas in the business process, choose the process part to be redesigned and also choose
the targets for the redesign must be specifically input to the the proper change to apply on it. To allow automatic process
software in order. Targets can be for example, lowering cost, redesign, process mining can be used. We briefly presented
increasing quality, increasing flexibility, etc... The order of the concept of process mining and we showed that although
targets is important because, in some times targets contradict. a lot of process mining algorithms exist and some were already
For example, increasing the quality in some processes may used in redesign projects, it is still not clear how to make
result in increasing the cost and so on. To find the changes process redesign using process mining a repeatable service.
to apply on the selected process part, we suggest the use of For this reason we presented a framework of a software that
© 2012 ACEEE 27
DOI: 01.IJIT.02.01. 42


automatically produces a redesigned business process model [6] W.M.P. van der Aalst: “The Application of Petri Nets to
using event logs from the old system execution. The basic Workow Management” The Journal of Circuits, Systems and
idea of this software is to integrate the results after using a Computers, 1998.
proper selection of process mining algorithms, and then use [7] M. Netjes , I. Vanderfeesten, , and H. Reijers: “Intelligent
tools for workflow process redesign: A research agenda”. Vol. 3812
these results to select a malfunction process part and
of Lecture Notes in Computer Science, pp. 444–453. Springer Verlag,
transform it using redesign best practices. Berlin, 2005
The proposed framework provides guidelines on how to build [8] H. Reijers, S. L. Mansar: “Best practices in business process
the redesign software. To actually build this software, we redesign: An overview and qualitative evaluation of successful
plan to work on the following points: redesign heuristics” Omega: The Int. Journal of Management Science,
- The different parameters and parameter values that 33(4):283–306, 2005
determine the process mining algorithms to use. [9] M. Netjes, S. L. Mansar, H.A. Reijers, W.M.P. van der Aalst:”
- How to integrate the different results from process An Evolutionary Approach for Business Process Redesign: Towards
mining to find the process part to redesign. an Intelligent System”, Proceedings of the 9th Int. Conference on
Enterprise Information Systems. pp. 484–494. INSTICC, Setubal
- The rules that guide the use of a specific best
(2007)
practice (transformation) on a process part. [10] M. Netjes, H.A. Reijers, and W.M.P. Aaslt: “On the Formal
- The log analysis parameters that will be used for Generation of Process Redesigns”, First Int. Workshop on Model
redesigns simulation. Driven Engineering for Business Process Management, pp. 49–60,
2008,
ACKNOWLEDGMENT [11] A. A. Kalenkova: “Application of If-Conversion to
Verification and Optimization of Workflows”, Programming and
The authors would like to acknowledge that the work for Computer Software, vol. 36, no. 5, pp. 276–288, 2010.
this paper was partly funded by the Qatar Foundation for [12] W.M.P. van der Aalst, A.J.M.M. Weijters, L. Maruster.
Education, Science and Community Development. The “Workflow Mining: Discovering Process Models from Event Logs”,
statements made herein are solely the responsibility of the IEEE Transactions on Knowledge and Data Engineering, 16(9):1128–
authors and do not reflect any official position by the Qatar 1142, 2004.
Foundation or Carnegie Mellon University. [13] A.J.M.M. Weijters, W.M.P. van der Aalst. “Rediscovering
Workflow Models from Event-Based Data using Little Thumb”,
Integrated Computer-Aided Engineering, Vol. 10 (2003)
REFERENCES
[14] C.W. Günther, W.M.P. van der Aalst, “Fuzzy mining - adaptive
[1] W.M.P. van der Aalst , A. H. M. Ter Hofstede , M. Weske : process simplification based on multi-perspective metrics”.
“Business Process Management: A Survey”, Proceedings of the 1st Proceedings of the 5th Int. Conference on Business Process
Int. Conference on Business Process Management, Springer Verlag, Management, Brisbane, Australia, 2007
2003. [15] M. Song,, W.M.P. van der Aalst. “ Towards Comprehensive
[2] J. Sinur, J. B. Hill: “ Magic Quadrant for Business Process Support for Organizational Mining.” Decision Support Systems
Management Suites”, Gartner Research, Pub-Date 18-Oct 2010, 46(1), pp. 300-317 ,2008.
ID Number: G00205212 [16] A. Rozinat, W.M.P. van der Aalst. “Decision Mining in
[3] B. Mutschler, M. Reichert, J. Bumiller: “Unleashing the Business Processes”. BPMCenter.org Report, 2006.
Effectiveness of Process-Oriented Information Systems: Problem [17] A. Rozinat, I. S. M. de Jong, C. W. G¨unther, W. M. P. van der
Analysis, Critical Success Factors, and Implications”, IEEE Trans. Aalst, “Process Mining Applied to the Test Process of Wafer
on Systems, Man., and Cybernetics, vol.38, no.3, pp. 280 -291, Scanners in ASML,”, IEEE Trans. on: Systems, Man, and
2008 Cybernetics, vol.:39, no.:4, pp: 474 - 479, 2009
[4] M. Attaran: “Exploring the Relationship between Information [18] R.S. Mans, M.H. Schonenberg, M. Song, W.M.P. van der
Technology and Business Process Reengineering. Information & Aalst, P.J.M. Bakker, “Application of Process Mining in Healthcare
Management”, 41(5), pp. 585-596, 2004. – A Case Study in a Dutch Hospital”, Communications in Computer
[5] W.M.P. van der Aalst , A.J.M.M. Weijters: “Process Mining, and Information Science, 25, pp. 425-438). Berlin: Springer,2008.
a Research Agenda”, Special Issue of Computers in Industry, vol.53, [19] The ProM tool, http://prom.win.tue.nl/tools/prom/
no.3, Elsevier Science Publishers, Amsterdam, 2004.

© 2012 ACEEE 28
DOI: 01.IJIT.02.01.42

Towards a Software Framework for Automatic Business Process Redesign

More Related Content

What's hot (20)

Viewers also liked (9)

Similar to Towards a Software Framework for Automatic Business Process Redesign (20)

More from IDES Editor (20)

Recently uploaded (20)

Towards a Software Framework for Automatic Business Process Redesign