with Bayesian Belief Networks

lettuceescargatoireAI and Robotics

Nov 7, 2013 (4 years and 5 days ago)

83 views

Enhancing the Prognostic Power of

IT Balanced Scorecards

with Bayesian Belief Networks

Stefan A. Blumenberg

Daniel J. Hinz

HICSS 39








8. November 2013

Summary


Balanced Scorecard (BSC) is a widely used performance
measurement system. Causal relationships are an integral part,
however it is neither thoroughly introduced in theory nor applied in
practice


Bayesian Networks (BBNs) can be employed to improve Balanced
Scorecard methodology


Balanced Scorecards and Bayesian Networks can be combined due to
structural similarities


A combination of both approaches allows for an
a priori

validation of
causalities with reduced effort in validity maintenance and better
prediction of value chain figures


A sample IT scorecard shows the applicability of both methods

HICSS 39








8. November 2013

Balanced Scorecard is widely adopted,
also in IT management


BSC is a widely adopted performance measurement method that
provides „a fast but comprehensive views of all businesses“

[Kaplan, R.S., Norton, D.P., 1992]
, that


Combines classical financial figures and „soft“ factors


Reduces information complexity


Consists of 4 perspectives in its generic design



BSC is also a common tool for IT management (e.g.
IT governance
or IT contribution to firm performance
)
[Graeser et al., 1998; Van der Zee,
J.T.M., De Jong, B., 1999]


Vision &

Strategy

Financial

Perspective

Internal

Business

Processes

Customers

Perspective

Innovation &

Learning

Perspective

HICSS 39








8. November 2013


Thoroughly extracted cause
-
and
-
effect chains provide a good basis for a
BSC to succeed
[Kaplan, R.S., Norton, D.P., 1996]

and allow


Prediction of value chain performance measures
[Kaplan, R.S., Norton, D.P., 2001]


Communication and realization of the corporate strategy
[Kaplan, R.S., Norton, D.P., 2001]


Incentives based actions
[Malina, M.A., Selto, F.H., 2004]









89% of surveyed companies agree to better understand the value of
intangible assets with modeled cause
-
and
-
effect chains
[Marr, B., 2004]

Causal relationships are the core and
major strength of BSC

IT support of

internal

communication

Customer

satisfaction

Net income

Cause
-
and
-
effect chain example

HICSS 39








8. November 2013

However, most firms do not implement
these powerful causal relationships

23%
54%
[Marr, B., 2004]

Not modelled

Modelled

[Ittner et al., 2003]

46%

77%

Explanation for these findings


The theoretical description of causalities by Kaplan and Norton within
BSCs is only vague
[Norreklit, H., 2000]


From a practical point of view, extraction of corporate causalities and
their reliability maintenance is complex and tedious
[Malina, M.A., Selto, F.H.,
2001; Grey, C., 2001]

Causal relationships within BSC

Empirical findings

HICSS 39








8. November 2013

Bayesian Belief Networks offer a
promising method to substantiate BSCs…

I

P

C

F

Bayesian Belief Network

F

P

C

I

Balanced Scorecard

BBN Principles


A BBN models the cause
-
effect relationships of its
nodes


Each node is a conditional
probability distribution
describing the effect of the
parent nodes on itself
Prob(node|parent nodes)


Network allows for simulation
in both directions


Given a change in a parent
node, how do the target
figures change?


Given a desired target value,
which values do the parent
nodes have to have?


…substantiated by …

HICSS 39








8. November 2013

… because of the structural similarities of
both approaches

Balanced Scorecard (BSC)


Consists of entities (called
figures), grouped within
perspectives


Directed edges indicate
causal relationships




Loops are allowed, but
should be omitted to be
compatible with BBNs

Bayesian Belief Networks (BBN)


Consists of entities (called
nodes), may be grouped
graphically


Directed edges describe causal
relationships and are used to
calculate conditional
probabilities


Loops are not allowed (graphs
has to be directed and acyclic)

HICSS 39








8. November 2013

Example: A sample BSC from literature…

Source:

Journal of Management Information Systems,



Van der Zee, J.T.M.; De Jong, B., 1999)

Learning and Growth

Perspective

Learn employees to use

productivity tools

Improve awareness about IT

opportunities in the context

of business activities

Increase employee productivity

through ubiquitous capabilities

Financial

Perspective

Customer

Perspective

Internal

Perspective

Start up call centers and

automated electronic distribution

channels for Internet

Reduce costs by designing

IT enabled processes

Make products more

accessible through new and old

distribution channels

Operating costs down

Improve „low cost“ perception of

potential & current customers

Standard sales growth

Net income growth


BSC from
Journal
of Management
Information
Systems


Implemented by a
European bank


Boxes contain
partial strategies
(not already
chosen figures)

HICSS 39








8. November 2013

… can be transformed into a BBN

Transformation process

The simple steps…


Transform every figure
of BSC into a BBN
node


Transform every
dependency arrow into
a directed edge


Eliminate recursive
loops (n/a in this
example)


… and the challenge


Discretize node values


Add probability tables
for each node





Some dependencies are easy to determine (e.g. impact of
sales growth on net income growth), others are not


Each assumption of the BSC has to be thoroughly revisited

HICSS 39








8. November 2013

More detail can be added by breaking
down partial strategies

Learning
and
Growth
Perspective
Improve awareness about
IT
opportunities
in
the
context
of
business activities
Likert scale
IT
trainings
Percentage
of
taught
employees
Business
workshops
Percentage
of
taught
employees
Learning
and
Growth
Perspective
Improve awareness about
IT
opportunities
in
the
context
of
business activities
Likert scale
Improve awareness about
IT
opportunities
in
the
context
of
business activities
Likert scale
IT
trainings
Percentage
of
taught
employees
IT
trainings
Percentage
of
taught
employees
Business
workshops
Percentage
of
taught
employees
Business
workshops
Percentage
of
taught
employees
Simulation 1


20% of all employees have yearly
business and IT trainings


80% only every trhee years


Awareness increases by

0.9 points


Simulation 2


80% of all employees have yearly
business and IT trainings


20% only every trhee years


Awareness increases by

2.1 points


HICSS 39








8. November 2013

Limitations and further research


BSC does allow loops, BBNs do not


Integration of time component is difficult in BSC
as well as BBN

Limitations

Further research


Empiricial validation can be done in two areas


Companies in the process of implementing a
BSC:

To demonstrate effectiveness in scorecard ramp
-
up



Companies with a working BSC:

To
test improvement capabilities of the proposed
approach against traditional double loop learning

HICSS 39








8. November 2013

Contact


Stefan A. Blumenberg


blumenberg@wiwi.uni
-
frankfurt.de



Daniel J. Hinz


dhinz@wiwi.uni
-
frankfurt.de