Models to Aid Decision Making in Enterprises
Suman Roychoudhury, Asha Rajbhoj, Vinay Kulkarni and Deepali Kholkar
Tata Research Development and Design Centre, Tata Consultancy Services, 54b Hadapsar Industrial Estate,
Pune, 411013, India
Keywords: Modeling, Enterprise Adaptation, Simulation.
Abstract: Enterprises are complex heterogeneous entities consisting of multiple stakeholders with each performing a
particular role to meet the desired overall objective. With increased dynamics that enterprises are
witnessing, it is becoming progressively difficult to maintain synchrony within the enterprise for it to
function effectively. Current practice is to rely on human expertise which is time-, cost-, and effort-wise
expensive and also lacks in certainty. Use of machine-manipulable models that can aid in pro-active
decision-making could be an alternative. In this paper, we describe such a prescriptive decision making
facility that makes use of different modeling techniques and illustrate the same with an industrial case study.
1 INTRODUCTION
Globalization forces and increased connectedness
have led to rise in business dynamics and shortened
time-to-market window for business opportunities.
Modern enterprises are subject to several change
drivers such as opportunities in a new market,
technology advance and/or obsolescence, regulatory
compliance etc. Current practice is to rely solely on
human expertise, which is largely a synthesis of past
experience, in order to arrive at a suitable response
to a change in the operating environment. This is an
effort-, time- and cost-intensive endeavour and is
also error-prone. Arriving at a response involves
addressing issues like: many a time it is not clear
which of the available options is the best option for a
given evaluation criterion, what would be the ripple
effect of taking that option and what is the best way
of implementing that option. As the cost of taking a
potentially incorrect decision is prohibitively high, it
is highly desirable to have aids that can support pro-
active (semi-) automated decision making, where it
would be possible to play out various what-if (and
if-what) scenarios to arrive at the right response,
feasibility of the response, and ROI of the response
(Kulkarni et al., 2013).
Typically, enterprises can be viewed as large-
scale distributed systems characterized by high
complexity, heterogeneity and intense dynamism
leading to complex interactions among humans,
business processes, IT systems and IT infrastructure.
Therefore the key idea is to model an enterprise
across various planes (see Fig. 1) namely,
Infrastructure plane concerning hardware
infrastructure and firmware managing it, Systems
plane concerning IT systems and their inter-
relationships and Business plane concerning
organization’s vision-mission-goals, structure and
operational processes (Kulkarni et al., 2013).
Furthermore, each plane of the enterprise is
amenable for specification in terms of various kinds
of models. For instance, intentional model for
specifying enterprise objectives and goals; business
process and/or event based models for specifying
workflows; various UML models for specifying
business applications; system dynamical model for
specifying the stocks of interest, their flows and
variables influencing the flows etc. These models
need to be relatable to each other so as to ensure
consistency and completeness within a plane and
alignment across adjoining planes. Therefore, we
believe a holistic model-centric approach will enable
organizations improve agility leading to better
adaptive responsiveness.
The rest of the paper is organized as follows.
Section 2 of the paper presents a motivating
example. Section 3 explains our modeling approach
in the light of the motivating example. Finally, we
discuss some of the key issues and present the
related work in Section 4 before concluding in
section 5.
465
Roychoudhury S., Rajbhoj A., Kulkarni V. and Kholkar D..
Models to Aid Decision Making in Enterprises.
DOI: 10.5220/0004966504650471
In Proceedings of the 16th International Conference on Enterprise Information Systems (ICEIS-2014), pages 465-471
ISBN: 978-989-758-029-1
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
Figure 1: Modeling Approach.
2 MOTIVATING EXAMPLE
In this section we introduce a motivating example
that sets the context for the rest of the paper. Let us
consider a large financial company (FinCom) whose
earnings model is based on the number of financial
products (e.g., loans, mutual funds, insurance etc.) to
be sold to potential customers. To sell its products,
the company needs to acquire a large customer base.
Customer acquisition is an expensive process that
involves investment in advertisement, promotions,
publicity etc. However, instead of incurring such
financial expenditure, FinCom intends to maintain a
minimal customer acquisition cost by partnering
with large retail chains selling consumer durable
products such a televisions, refrigerators etc.,to a
sizeable customer base. FinCom targets this
customer base by providing attractive loans at 0%
interest for 1-2 years duration. In this manner,
FinCom acquires new customers, then cross-sells
other financial products to them. FinCom works on a
very thin margin for individual clients; however,
their overall profitability remains high due to high
volume sell to a substantial customer base. FinCom
business model works for the retailers too who can
offer their clientele attractive third party credit
facilities.
The business model of FinCom seemed to have
worked very well. The company is able to hook
increasingly large number of customers at a faster
growth rate of 20% Quarter_on_Quarter. They are
also able to convert a healthy chunk of prospects to
customers with minimal selling cost. Cost of
servicing customers is also quite low. Thus, overall
their business is growing at a pretty fast pace.
However, the company has identified new
challenges that are vital for their future growth and
expansion into new markets. The company aims to
scale up revenues by a factor of 10 without having to
increase the associated cost. In other words, FinCom
would like to have a non-linear revenue growth. As
the company ventures into emerging markets, most
of the IT intensive business processes need to scale
and seamlessly integrate with newer systems.
Currently the IT operations are managed by FinCom
itself but the company is finding it increasingly
difficult as managing IT is not their primary forte.
Instead they would like to concentrate on developing
new financial products, perform various market
analyses and focus on diverse data-centric analytics
on its existing customer base.
FinCom is looking for able IT service providers
who can manage their end-to-end IT operations
including guidance towards future IT expansion. For
example, to remain competitive, FinCom would like
their gadget savvy customers to avail new channels
like smartphones, tablets and other ubiquitous
devices for making payments towards their loans or
mortgage products. Similarly the company would
like to evaluate whether some of their IT services
could be moved to a cloud-based infrastructure
without compromising any security issues or
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
466
Figure 2: Intentional Model to capture Stakeholder goals.
Figure 3: Customer Acquisition Business Process Flow.
degradation of quality of service (QoS). However,
their internal IT is not well equipped to provide
these additional pervasive channels or move to a
cloud-based infrastructure at a rapid pace. Therefore
FinCom would like to outsource their non-core
business to external IT service providers.
3 MODELING APPROACH
In the context of the above business case, Fig. 1
presents our holistic modeling approach aimed at
pro-active data-driven decision making. The goal is
to capture various facets of an enterprise belonging
to each of its plane (i.e., Business, System and the
Infrastructure plane) with precise modeling
techniques and finally analyzing the result by
simulating them in concert. For example, in Fig. 1,
the goal model captures the business objectives of
the enterprise, while the organization structure
model describes the people/role aspect of the
enterprise along with vision, mission, local policies
etc. Similarly, the IT System model describes the
overall IT need of an enterprise from a system
perspective i.e. which steps of operational processes
are being automated using which application
services and what data needs to be monitored. Apart
from the structural aspects of the enterprise, the
general behaviour of the enterprise is realized in the
form of a behavioural model. Currently this is
realized using Business Process Modeling Language
(Scheer, 1996) but in future will be extended to a
ModelstoAidDecisionMakinginEnterprises
467
Figure 4: Stock-n-Flow model to analyze peak season load.
more general modeling language such as event based
modeling (Clark et al., 2011).
Once the structural and behavioural aspects of
an enterprise across multiple planes are captured,
different what-if scenarios are played out and the
results so obtained are provided as feedback to fine
tune the models in each of these planes. For
example, analysis of workflow can lead to
restructuring of business process and/or task
assignment resulting in improved resource
utilization. Similarly one can validate if none of the
critical business goals are compromised or the IT
system model is in conformance with the perceived
IT needs of the enterprise. The analysis models are
chosen such that they are best suited for what-if
analysis, for instance, stock-n-flow, agent-based,
petri-nets and event-based models (Forrester, 1958;
Reisig, 1991; Bresciani, 2004). In the following
section, we demonstrate use of various models and
their relationships with an objective of automated
analysis. We start with an intentional model that
captures the overall goals of the enterprise (Yu et al.,
2006).
3.1 Modeling Goals with Intentional
Model
In Fig. 2, the ovals represent the strategic rationales
(SR) for the two primary stakeholders (i.e., FinCom
and SP). Each SR is further decomposed into goals,
soft-goals and tasks that are means by which the
goals can be reached. For example, the root-level
goal ‘Manage Business’ can be achieved by defining
business strategies, acquiring customers, making
profit, and scaling up business. The links between
the stakeholders represent strategic dependencies.
For example, FinCom is dependent on SP to
automate some of their business processes, while SP
is dependent on FinCom to define appropriate SLAs.
The highlighted part in Fig. 2 shows specifically
the customer acquisition process that is one of our
primary point of interest. Some of the sub-tasks of
Acquire Customers are Credit History Check, Cross-
Sell products etc., while Market Analysis helps to
acquire new customers. Although Intentional model
can capture goals, tasks and their dependencies, it
does not capture the sequence of events that truly
describes the process.
Therefore to capture such event-oriented process
behaviour, we use standard BPM language as
depicted in Fig. 3. The process model shows two
primary stakeholders, namely the consumer durable
Retailer who offers customer with various products
along with attractive financial schemes and the
Service Provider who facilitates the customer loan
request process. Fig. 3 describes the process, i.e., the
customer fills the loan request form, which is then
scanned and send to SP for credit check. Loan is
approved if the customer has a good credit score,
otherwise it is denied. However there are SLAs that
guarantees that the entire approval process from loan
application to approval/rejection should not take
more than 3 minutes. Moreover, during peak festive
season, there is a sudden increase in number of
customers to be serviced. Keeping this concern in
mind, the business process must be able to adapt to
changing business scenario without any significant
increase in cost or deviation in SLAs. Since it is not
possible to analyze such a scenario using standard
BPM time, cost and resource analysis techniques
(Scheer, 1996; IBM RSA, 2014), we use stock-n-
flow model which provides quantitative analytical
modeling abilities to play out various what if-
scenarios, the results of which are used by the
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
468
process models for further optimization. The next
section demonstrates use of stock-n-flow model for
this purpose.
3.2 Analyzing Peak Season Load
Stock-n-Flow Model
Stock-n-flow models (Forrester, 1958) typically
capture the temporal behaviour of an enterprise. For
example, during peak season (e.g., festive holidays),
there is a steady rush in sale of consumer durable
products. As a result the number of consumers
applying for loan increases. To ensure there is no
bottleneck, FinCom would like to know whether
they can still manage their loan approval business
process with existing manpower without
compromising on QoS. Fig. 4 shows the stock-n-
flow model to analyze peak season load. The stocks
are typically the Incoming Customers, Customers
Applying Loan etc., while the flows are Application
rate, Verification rate, Customer flow etc. Some of
the key variables that are used to analyze the peak
season load are Available FTEs, wait time in queue,
no. of concurrent users etc. Using stock-n-flow
model one can parameterize and then refine the
values of the variables to play out various what-if
scenarios for pro-active decision making. For
example, the model can simulate the impact on
verification rate or loan application rate during peak
season (i.e., when customer flow increases).
Consequently, one can reason about the number of
FTEs required at the retailer side or IT side or both.
Similarly, additional questions like how many
concurrent requests can still be handled without
significant degradation in QoS (e.g., wait time in
queue). All such questions and scenarios can be
played out in advance and thereby help both the
service provider and the enterprise to arrive at more
informed decisions.
Although stock-n-flow models helped us to
simulate various what-if scenarios, however, one
needs to refine the values of the stocks, flows and
variables to arrive at an optimum solution. That is, to
carry out the simulation, an initial set of valid input
values from the sample space is required. This is a
manual time consuming process because one needs
to keep all the dependency constraints in mind while
assigning values to variables. For example, FTE
productivity can be increased with improved training
and additional incentives like increase in salary.
However, this increases FTE cost. FTE cost can also
increase with increase in total number of FTE, which
in turn increases administration cost and as well
impacts on Service Provider profitability. Similarly,
Retailer and FinCom profit increases with increase
in total number of customers. Thus one can observe
that there is a dependency relationship among
various parameters and considerable manual effort is
required to assign right values to all the parameters
without breaking any of the pre-defined
organizational policy constraints. Also, there are
pre- and post-condition constraints from the business
process model and business rule constraints that
need to be considered. In order to remove such
manual intervention, we have introduced model
checking (Merz, 2001), by which we automatically
obtain values that satisfy the given constraints in the
stock-n-flow model.
Once a valid set of input values from the model
checker is automatically obtained, the stock-n-flow
model is then simulated and the results so obtained
are used as feedback to other models in the three
planes. For example, using stock-n-flow we obtained
the optimal number of FTEs for managing peak
season load and feed these results back to
infrastructure and organization structure model as
the new utilization plan (see Fig. 1).
3.3 Discussion and Future Work
So far we have seen how each of the models
belonging to Business, System and Infrastructure
planes of Fig. 1 captures a specific problem of the
enterprise. However, it is important to relate the
analysis results and percolate them across different
planes to get a more holistic view of the enterprise.
For example, using intentional models we captured
business objectives or tasks. But these tasks were not
ordered or sequenced. Therefore, by using BPM
language we were able to describe the sequence or
ordering of events as well as associate cost, time and
resources to these events. However, since BPM
models were not sufficient to play out certain
analyses, other suitable modeling techniques were
employed. For instance, stock-n-flow models were
used for analyzing peak customer load. Moreover
we related the key variables used in the stock-n-flow
models with the “data” variables available from the
BPM and System models. Thus, we were able to
establish an initial relationship between these
disparate models.
4 RELATED WORK
Enterprise Architecture modeling is prevalent for a
number of years (Lankhorst, 2005). There are a quite
a few EA frameworks like FEA (FEA, 2006),
ModelstoAidDecisionMakinginEnterprises
469
Zachman (Zachman, 1987) TOGAF (TOGAF,
1995), Archimate (Archimate, 2012), that provide
holistic blueprints for the organizational and
architectural models. However, a key aspect that is
missing is machine processability analyzability,
which is the core contribution of this paper. MEMO
(Frank, 2002) provides a method to support the
development of enterprise models. Abstractions for
various interrelated aspects like corporate strategy,
business processes, organizational structure and
information models are provided, but, with limited
support for automated analysis. Other key topics like
Business-IT alignment, landscape mapping etc, are
covered in detail over the past (Schekkerman, 2006),
however the focus of this paper is more on
automated machine-dependent (i.e., minimum
human dependency) decision making using a variety
of appropriate modeling techniques. From a tooling
perspective, various tools exist for enterprise
architecture and business process modeling (Scheer,
1996; IBM RSA, 2014; iGrafx, 2014; MEGA,
2014), however analysis support is limited to
simulation of business processes so as to identify
process bottlenecks and suggest optimization in
terms of resources, time and cost. These tools do not
provide support for taking forward analysis results
of one model onto another. Moreover, analysis
capability of these tools is limited to business
process models only. Existing literature on
enterprise modeling research (Schekkerman, 2006)
also does not include evidence of use of multiple
modeling techniques in conjunction, or of model
checking to verify multiple modeling paradigms. To
this respect, our previous work on mapping
Intentional models with System Dynamic models in
the context of EA (Sunkle et al., 2013) was an early
start. In this paper, we have extended that work by
introducing the concept of modeling across various
layers of the enterprise with suitable techniques that
are appropriate for that layer and finally we propose
to orchestrate them in concert to get a holistic view
of the enterprise.
5 CONCLUSIONS
In this paper, we discussed a model-centric approach
to enable enterprises improve their agility and
prepare them for better adaptive responsiveness. We
proposed a layered architecture for modeling
enterprises wherein the adjoining layers have a well-
defined relationship and each layer addresses a set of
coherent concerns as seen from the perspectives of a
set of stakeholders. The key idea is to specify each
layer in terms of a model which can be viewed as a
set of relatable models each constituting an intuitive
and closer-to-problem-domain specification of a
concern – as advocated by separation of concerns
principle. We argued the case for these models to be
relatable, analyzable and simulatable. We illustrated
the rationale behind the proposed model-centric
approach through a motivating example. We
described several modeling techniques (e.g.,
intentional, stock-n-flow, agent-based) that best
match an underlying problem scenario. We
described how each one of the models caters to
specific goals and how they relate to and
complement each other. We further described how
our proposed solution percolates analysis results
from one model to another model either in the same
or in a different enterprise layer. Until now, we have
found very little evidence of such an approach in the
existing literature and believe that the enterprise
engineering community can largely benefit from the
investigations and position taken in this paper.
REFERENCES
ArchiMate 2. 0 Specification 2012. Van Haren Publishing
Series. Bernan Assoc.
Bresciani, P., Giorgini, P., Giunchiglia, F., Mylopoulos, J.,
Perini, A., 2004. Tropos: An Agent-Oriented Software
Development Methodology,In Autonomous Agents
and Multi-Agent Systems, Vol. 8, No. 3, pp. 203–236,
Springer.
Clark, T., Barn, B., 2011. Event driven architecture
modelling and simulation, In SOSE’11, pp. 43-54,
IEEE Computer Society.
FEA 2006. Federal Enterprise Architecture Program EA
Assessment Framework 2.1.
Frank, U., 2002. Multi-perspective enterprise modeling
(MEMO) conceptual framework and modeling
languages," In HICSS., pp.1258-1267. IEEE
Computer Society.
Forrester, J., 1958. Industrial Dynamics--A Major
Breakthrough for Decision Makers, In Harvard
Business Review, Vol. 36, No. 4, pp. 37–66.
IBM RSA 2014. http://www-03.ibm.com/software/
products/us/en/ratisystarchfami.
iGrafx 2014. http://www.igrafx.com.
Kulkarni, V., Roychoudhury, S., Sunkle, S., Clark, T.,
Barn, B., 2013. Modeling and Enterprises - The Past,
the Present, and the Future, In MODELSWARD’13,
SCITEPRESS.
Lankhorst, M., 2005. Enterprise Architecture at Work:
Modeling, Communication and Analysis, Springer.
MEGA EA Tools, 2014. http://www.mega.com/en/
solution/enterprise-architecture.
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
470
Merz, S., 2001. Model Checking: A Tutorial Overview”,
In Proc of the 4th Summer School on Modeling and
Verification of Parallel Processes, pp. 3-38.
Scheer, A., 1996. ARIS-Toolset:Von Forschungs-
Prototypen zum Produkt" Informatik-Spektrum 19:
71–78, Springer-Verlag.
Schekkerman, J., 2006. How to Survive in the Jungle of
Enterprise Architecture Frameworks: Creating or
Choosing an Enterprise Architecture Framework,
ISBN-13: 978-1412016070.
Sunkle, S., Roychoudhury, S., Kulkarni, V., 2013. Using
Intentional and System Dynamics Modeling to
Address WHYs in Enterprise Architecture, In
ICSOFT-EA, SCITEPRESS.
TOGAF Version 9.1, 1995. http://pubs.opengroup.org/
architecture/togaf9-doc/arch/
Reisig, W., 1991. Petri Nets and Algebraic Specifications.
Theoretical Computer Science Vol. 80, No. 1, pp. 1-
34, Elsevier.
Yu, E., Strohmaier, M., Deng, X., 2006. Exploring
Intentional Modeling and Analysis for Enterprise
Architecture, In EDOC Workshop on Trends in
Enterprise Architecture Research.
Zachman, J. A., 1987. A Framework for Information
Systems Architecture. In IBM Systems Journal, Vol.
26, No. 3.
ModelstoAidDecisionMakinginEnterprises
471
