Modeling and Understanding Complex
Enterprise Behavior using a Hybrid
Simulation Approach
Presented By
Christopher Glazner
PhD Candidate
21 January 2009
Research Project Overview
Synopsis:
This research develops a methodology and
supporting theory for modeling complex enterprise
behaviors hybrid simulations based upon an
enterprise’s architecture.
Organization
Strategy
The
Enterprise
Architecture
Enterprise
Architecture
Simulation
http://lean.mit.edu
Product/
Service
Process
Inputs
Why:
The
Enterprise
Knowledge
Such simulations can be used to
find key levers across the
enterprise to affect change and
to evaluate future changes to the
enterprise architecture
Hypothesis:
Hybrid modeling of an
enterprise’s architecture can
allow a deeper understanding of
enterprise dynamics arising
from its structure and can
identify key control levers and
critical areas for alignment.
Policy
Information
Outputs
An executable architecture model
Method:
•Match modeling methodologies with views
of the enterprise architecture based on the
context of the view and the methodology.
•Simulate enterprise behavior arising from
the enterprise architecture using hybrid
(multi-agent, system dynamics, eventbased) modeling techniques.
•Combine sub-models into an integrated
hybrid model of enterprise behavior
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 2
Case Study: An Application of
Enterprise Architecture-Based
Simulation Modeling
ƒ “TechSys”
ƒ 2 Billion dollar multi-market “enterprise of
enterprises” in the aerospace/defense sector
ƒ 2004: New strategy developed to grow the
enterprise in the future
ƒ 2005: New organizational structures put in place
to support strategy
ƒ 2006:Work begun on developing new processes
to support strategy
ƒ 2007: Doubled in size through a combination of
internal growth and two major acquisitions that
complemented strategy
ƒ 2008: “Do we have the right enterprise
architecture to support our strategy?”
http://lean.mit.edu
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 3
Key Questions for TechSys
Area of Focus:
• TechSys’s pursuit and capture of new business
Key Questions:
• Can TechSys achieve its growth goals given its current enterprise
architecture with constrained resources available for growth?
•
How does the enterprise perform as resource allocations are
changed? Does the architecture favor a particular business
model?
•
What changes can be made to the enterprise architecture to
improve opportunities for growth given resource constraints?
http://lean.mit.edu
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 4
High-Level Conceptualization
The model is run several hundred times in a Monte Carlo fashion for
each combination of inputs to obtain a distribution for expected profits
over a three year period
http://lean.mit.edu
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 5
The Performance Landscape for
the current Enterprise Architecture
Performance is best
when :
Profit
s
1.the operating units
t
Join
e
Silo
d
http://lean.mit.edu
$
D
R&
p
Pro
ls
a
s
o
do not pursue joint
projects, and
2. the majority of
discretionary funds is
given to pursuing new
proposals, rather than
R&D
This is counter to the prevailing
mental models and does not
support TechSys’s strategy!
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 6
Analysis of Enterprise
Behavior
These surprising results in the model can be traced back to
two major areas within the enterprise architecture:
1.
2.
The process used for funding new proposals between Operating
Units (OUs) is biased against joint proposals;
Profits arising from joint proposals are not shared equally between
OUs, causing an additional bias against selecting them
These two factors appear to have a multiplicative effect
The strong bias toward investing discretionary dollars towards
pursuing new proposals can be attributed to:
1.
2.
The time horizon for the model may be to short to realize benefits
of R&D;
TechSys is not getting a good return on its R&D investment
http://lean.mit.edu
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 7
An Alternative Enterprise
Architecture: What If Analysis
Alternative Enterprise Architecture
Performance Landscape
100%
90%
70%
Prof
it
60%
50%
40%
30%
20%
10%
0%
Can the potential performance of the
enterprise be increased by
modifications to the architecture?
90%
100%
80%
70%
% to
40%
100%
90%
80%
&P
50%
40%
e
Syn
20%
60%
•Modified incentives againsty tjoint
oB
40%
70%
60%
50%
40%
30%
20%
10%
0%
30%
rgy
0%
%
http://lean.mit.edu
100%
80%
70%
B
to
y
r
na
o
i
t
cre
Dis
30%
20%
10%
50%
20%
40%
30%
60%
40%
e
Syn
remove
selection
bias against
synergistic
The Current
Enterprise
Architecture
proposals
Performance Landscape
% to
•Modified proposal selection process to
50%
90%
D
10%
%
0%
10%
0%
rgy
0%
20%
r
proposals with a changetioin
na the way
re
budgets are allocated
isc
10%
120%
110%
Changes to the Enterprise Architecture:
50%
140%
130%
Prof
it
80%
&P
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 8
Summary of TechSys Model
The simulation was able to:
• Evaluate TechSys’s architecture from a systems perspective
• Key dynamics were generated from the interaction of process, knowledge and
•
•
•
•
organization views of the architecture
No other currently used modeling approach could have given similar results
•
Clearly show that the current architecture would not support their
growth strategy
Suggest changes for an alternative architecture that was capable
of meeting growth goals
Indicate that TechSys needs to evaluate its discretionary budget
allocation strategy--it can clearly do better
• IRAD effectiveness also needs to be addressed. Perhaps develop a new process
for selecting and funding IRADs.
highlight unanticipated benefits owing to the process of model
development
There is great potential for this model to be further
developed.
http://lean.mit.edu
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 9
Observations and Future
Work
ƒ
This is the only known approach to enterprise architecture
modeling that can analyze an enterprise architecture from a
systems perspective and capture behaviors resulting from
interactions across the views
ƒ
The EA Framework provided tremendous guidance in structuring
the model, identifying data sources, and testing the model for
completeness
ƒ
This approach can be coupled with a specific EA framework for
fuller development into a standardized toolkit
ƒ
Libraries of architectures can be developed in a “toolbox” fashion
that could make future models easier to develop.
ƒ
Modeling would be made easier if EA frameworks could include
lists of common interfaces between views
http://lean.mit.edu
© 2009 Massachusetts Institute of Technology Glazner- 01/21/09 - 10