You are on page 1of 30

Executive Program on

Data & Decision science Consulting


06 Oct.-09 Dec. 2018 (on Weekends)

DEMATEL:
Decision making trial and evaluation laboratory  

Prof. Surya Prakash Singh


PhD (IIT Kanpur), PDF (NUS Singapore, MIT USA)
Associate Professor
Department of Management Studies
Indian Institute of Technology Delhi
E-Mail: surya.singh@gmail.com
Step2: Normalize the direct-influence matrix:

Once the direct influence matrix has been obtained, it can be


normalized by using Eqs. (1) and (2) together initial normalized direct-
influence matrix “N”.
Step 3: Obtain total Influence Matrix:

Once the normalized direct relation matrix “N” is obtained, the total
relation matrix “T” can be acquired by using Eq. (3).
The sum of rows is denoted by D and of columns is denoted by R through Eqs.
(4, 5, 6). Addition of D and R reveals the relative importance of each criterion.
And the subtraction of D and R divides the criteria into two groups: cause
group and effect group. When (D-R) is positive, that particular criteria belongs
to the cause group and when the (D-R) is negative, that criteria belongs to the
effect group.

Therefore, the causal diagram can be obtained by mapping the data set of the
(D+R, D-R).
DEMATEL’s Causal Analysis
Case of Mobile APP
• This case applied DEMATEL to analyze Mobile APP
quality characteristics from the perspectives of
Developers and Users.

• According to the results, users regard security as the


most important quality characteristic, while functional
suitability, usability and reliability are regarded as the
most important quality characteristics by developers.
Overview
• Mobile App is the software on smartphones, tablets and other mobile
devices.

• ABI Research(2013). The Mobile App Market will be Worth $27


Billion USD($168 Billion RMB) in 2013 as Tablet Revenue Grows.

• Despite many mobile Apps are creative, the negligence of


the key factor, namely product quality, has led to usage
problems and negative reviews.
Introduction

The purposes of the case are the following :

1) To explore the Mobile App software product quality


characteristics to clearly grasp the facts about Mobile App quality
and provide mobile App developers with a clear quality vision for
planned control.
2) To explore the cognitive differences regarding quality
characteristics from the perspectives of developers and users.
The developers can find out whether their products meet the
quality expectation of the users, and carry out the proper control
and quality improvement plan under limited resources.
ISO 25010 Quality Model
DEMATEL
Analysis
Quality Characteristics D R D+R D-R

Functional Suitability 14.583 15.573 30.156 -0.990

Reliability 14.143 14.799 28.943 -0.656

Result Usability 14.301 15.144 29.445 -0.843

Performance efficiency 14.011 13.764 27.775 0.246


The direct & indirect
effects of the Quality Portability 14.072 14.186 28.258 -0.114
Characteristics under
Maintainability 14.325 13.282 27.607 1.043
user's perspective.
Compatibility 14.512 13.612 28.124 0.900

Security 15.282 14.868 30.150 0.414

AVG 28.807 0
Result
The Causal
diagram
unde user’s
perspective.
Quality Characteristics D R D+R D-R

Result Functional Suitability 8.926

8.810
8.498 17.424 0.428

Reliability 7.969 16.779 0.841

The direct and Usability 8.770 8.346 17.116 0.424

indirect effects Performance efficiency 8.000 7.799 15.798 0.201


of the Quality
Portability 6.777 8.130 14.907 -1.353
Characteristics
under developer's Maintainability 7.424 8.648 16.072 -1.224
perspective
Compatibility 7.321 7.553 14.874 -0.233

Security 8.486 7.570 16.055 0.916

AVG 16.128 0
Result
The Causal
diagram
under
developer's
perspective.
Conclusion
•Functional Suitability, Usability and Reliability are
quality characteristics perceived as important by
both the developer and the user.

•Portability is the only quality characteristic that the


user and the developer commonly agree.

•Both the developer and the user regard


Performance efficiency and Security can affect other
quality characteristics.
Thanks
Questions Please…
Executive Program on
Data & Decision science Consulting
06 Oct.-09 Dec. 2018 (on Weekends)

Interpretive Ranking Process (IRP)

Prof. Surya Prakash Singh


PhD (IIT Kanpur), PDF (NUS Singapore, MIT USA)
Associate Professor
Department of Management Studies
Indian Institute of Technology Delhi 17
E-Mail: surya.singh@gmail.com
Interpretiveness in Decision-
making

Mental Models (Senge, 1990).


Organizational Culture (Schein, 1992)
Sense making (Weick, 1979, 1995)
Managerial frames (Schrivastava and Mitroff, 1983)
Garbage can model (March and Olson, 1976)
Theories in use (Argyris et al., 1985)
Critical thinking and argument mapping (Gelder, 2007; Gelder
and Lewis, 2006)
Patterns in management (Mintzberg, 1978, 1989)
Why IRP when we have AHP?

For “n” criteria and “m” alternatives, in AHP we need “m+1”


matrices to be constructed through questionnaires. And, each
matrix need to be checked for consistency to finally get
rankings.
Whereas

For “n” criteria and “m” alternatives, in IRP we need only one
matrix to be constructed through questionnaires and there is no
need for consistency check.

19
Four matrices in AHP for n=4 & m=3
  C1 C2 C3 C4
Step 1:

C1  --      
C2   --     
C3     --   
C4       --  Single matrix in IRP for n=4 & m=3
 C1 A1 A2 A3
A1  --       A1 A2 A3
A2   --   
A3     --  A1  --    
 C2 A1 A2 A3 A2   --   
Step 2:

A1  --    
A3     -- 
A2   --   
A3     -- 
 C3 A1 A2 A3 Criteria (C1, C2,C3,C4) to
A1 --      be written in each cell
A2   --   
A3     -- 
 C4 A1 A2 A3
A1  --    
A2   --   
20
A3     -- 
Steps of the Basic Process
254
Table: Problem data having 18 alternatives and 6 criteria

Case
Example

23
Table: Pair-wise comparison matrix for IRP
(derived from questionnaires used to get inputs for AHP)

24
25
Table: Net-Dominance matrix for IRP

26
Strengths
It is easy to compare by the impact of interactions rather
than the variables in abstract sense.

It is comparatively easy to judge the dominance of one


interaction over the other rather than the extent of
dominance.

It is based on the strength of paired comparison as it


does not create any cognitive overload.

It is not necessarily dependent on weightage of criteria,


which is a debatable issue in MCDM methodologies.
Limitations
It is based on interpretive and judgmental processes and at
times may be highly subjective.

It usually treats all the criteria equally ignoring their relative


importance, as given in the illustration. However, this limitation
can be overcome by assigning ordinal weights to various
criteria and carrying out sensitivity analysis.

It is difficult to be validated in terms of objective validation


tests.

It is difficult to interpret a matrix of size beyond 10x10 as the


number of paired comparisons would exponentially increase,
and thus only modest sized problems can be effectively
implemented with this process.
Thanks
Questions Please…

You might also like