Group preference aggregation based on ELECTRE methods for ERP

system selection

Suzana de França Dantas Daher

Adiel Teixeira de Almeida

Agenda

Introduction / Motivation

Model Proposed

Numerical application

Final Remarks

Motivation

Enterprise resource planning (ERP) improves operational efficiency by integrating business processes and providing better access to integrated data across the entire enterprise

Deciding which is the most suitable ERP solution is often a difficult task for many companies.

ERP not developed in-house

Motivation

Organizations select and implement ERP systems so as to obtain a variety of tangible and intangible benefits and for strategic reasons.

The evaluation process of ERP systems needs to take many criteria into account

organizational factors such as the complexity of the business;

dealing with change management,

cost drivers,

its functional requirements,

system flexibility and system scalability,

external factors

Motivation

There is a strong possibility that in several organizations, an ERP system will be selected by a group.

Multicriteria group decision making involves individuals who provide their preferences for a set of alternatives with respect to a set of attributes

Diverging opinions may arise

Introduction

This study deals with how support a group of individuals to achieve a collective decision when selecting an ERP system.

The methodology adopted considered that DMs act in accordance with their own interests and there is no information about their relative importance to each other

Introduction

Which decision making methodshould be used?

The selection of the most suitabledecision making method should bebased on the preference structures ofthe DMs

Additive Model

For instance, an additive model could be considered, such as in Daher, S F D ; Almeida, A T (2012) The Use of Ranking Veto Concept

to Mitigate the Compensatory Effects of Additive Aggregation in Group Decisions on a Water Utility Automation Investment. Group Decision and Negotiation Journal, v. 21, p. 185-204.

In this case the additive model is considered for

aggregating the multiple criteria, and

Aggregating the decision makers’ preferences

A veto concept is applied for the additive Aggregation of decision makers’ preferences

Outranking methods

An alternative approach is considered in this work:

Outranking methods

Other properties are assumed for the decision maker preference

Outranking methods

Outranking methods are particulary suitablefor decision-making through the notion ofweak preference and incomparability.

Outranking relations (S): “at least as good as”

aSb and not bSa a P b (a is strictly preferred to b)

bSa and not aSb b P a (b is strictly preferred to a)

aSb and bSa a I b (a is indifferent to b)

not aSb and not bSa a R b (a is incomparable to b)

Outranking methods

Construction of an outranking relation is based on 2 major concepts:

CONCORDANCE

For an outranking aSb to be validated, a sufficient majority of criteria should be in favor of the assertion “a is at least as goodas b”.

NON-DISCORDANCE

When the concordance condition holds, noneof the criteria in the minority should opposetoo strongly to the assertion aSb.

ELECTRE Family methods

ELECTRE ( Elimination Et Choix Traduisant laRealité / Elimination and Choice Expressingthe Reality)

This family seeks to obtain a set of Nalternatives that outrank those which do not belong to the subset N.

Choice problematic:

ELECTRE I, ELECTRE Iv, ELECTRE IS;

Ranking problematic:

ELECTRE II, ELECTRE III, ELECTRE IV

Sorting problematic:

ELECTRE A, ELECTRE TRI

Model Proposed

Based on a combination of two outrankingmethods: ELECTRE II and ELECTRE IV.

Assumptions:

The decision problem is well structured

Prior definition of a set n alternatives

Prior definition of a set k criteria

The model is organized in three steps

Alencar, L.H., Almeida, A. T., Morais, D. C.: A Multicriteria Group Decision Model Aggregating the Preferences of Decision-Makers Based on ELECTRE Methods.

Pesquisa Operacional 30, Issue 3, 687-702 (2010)

Model Proposed

Individual ranking of alternatives (ELECTRE II)

Matrix of global evaluation

Group ranking of alternatives (ELECTRE IV)

Model ProposedFirst step ELECTRE II

Generation of individual rankings of alternatives

For each Decision Maker:

Decision Matrix D

Concordance and discordance indices

Criteria weights (inter-criteria information)

Model ProposedFirst step A concordance index C(a,b):

represents the coalition of arguments in favor of the statement “a is at least as good as b ” or in other words “a outranks b ”

A discordance index D(a,b) :

is used to measure the arguments that may cast some doubt upon the latter statement.

No veto condition:

)}()(:{

),(

bgagj

j

jj

wbaC

)}()({max),()()(:

agbgbaD jjbgagj jj

Jjbgagvag jjj ),())(()(

Model ProposedFirst step These indices are used to construct two

pre-orders:

strong outranking relation (a SS b)

weak outranking relation (a Sw b)

baS S

baSW

)()(:)()(:

),(

),(

bgjagjj

j

bgjagjj

j ww

dbaD

cbaC

)()(:)()(:

),(

),(

bgjagjj

j

bgjagjj

j ww

dbaD

cbaC

iff

iff

Model ProposedSecond step Obtain a matrix of global evaluation

Analyst must collect all individual ranking and compile them in a global evaluationmatrix.

DMs are considered as criteria and theirrankings correpond to the evaluation of thealternatives (ranking position of thealternative)

Model ProposedThird step ELECTRE IV

Obtain a group ranking

ELECTRE IV is used in cases in which thereis a pseudo-criterion family and its mainfeature is the absence of a weightingrelated to the relative importance of thecriteria

Numerical application

Fictitious case study: ERP selection for a Brazilian airlines

The company has to deal with inefficient operational procedures and an IT/IS legacy system.

In order to improve its competitiveness, the company launch of several projects including, an ERP system and the reengineering of some business processes

Numerical application

Four decision makers:

the financial manager (DM1)

the IT/IS manager (DM2)

the operational manager (DM3)

the customer relation manager (DM4)

Analyst should conduct the decisionmaking process.

Number of alternatives: 4

Numerical application

Criteria selection

Based on: ISO/IEC 9126-1

a standard that addresses quality model definition and its use as framework for software evaluation

Group of criteria:

functional, portability, maintainability, efficiency, vendor, cost

Numerical application

Criteria adoptedCriteria Criteria group

C1 Completeness Functional

C2 Number of simultaneous users Functional

C3 DBMS Standards Portability

C4 Number of modules Maintainability

C5 Time behavior Efficiency

C6 Length of experience Vendor

C7 License cost Cost

C8Installation and implementation

costCost

Numerical application

Decision matrix

Concordance and Discordance coefficients

C1 C2 C3 C4 C5 C6 C7 C8

A1 4 5000 4 8 0.3 4 0.7 1.8

A2 5 3000 5 12 0.6 5 0.5 1.3

A3 3 4500 4 10 0.2 4 0.6 1.7

A4 5 4000 3 5 0.7 5 0.4 2.0

c+ c- d+ d-

DM1 0.8 0.6 0.4 0.5

DM2 0.7 0.5 0.3 0.4

DM3 0.8 0.5 0.3 0.4

DM4 0.9 0.6 0.3 0.5

Numerical application

Criteria weights

Table of preorders for each DM

C1 C2 C3 C4 C5 C6 C7 C8

DM1 0.175 0.175 0.082 0.221 0.043 0.117 0.093 0.094

DM2 0.081 0.101 0.086 0.333 0.005 0.081 0.188 0.125

DM3 0.102 0.004 0.005 0.200 0.136 0.149 0.370 0.034

DM4 0.035 0.035 0.198 0.167 0.056 0.232 0.211 0.066

Ranking DM1 DM2 DM3 DM4

1st A2 A2 A3,A4 A2

2nd A3 A1 A2 A3

3rd A1 A3 A1 A1,A4

4th A4 A4

Numerical application

Matrix of global evaluation

Global ranking of alternatives

DM1 DM2 DM3 DM4

A1 2 3 1 1

A2 4 4 2 3

A3 3 2 3 2

A4 1 1 3 1

Ranking Alternatives

1st A2

2nd A1, A3,A4

Final remarks

In general, decisions made in organizationsinvolve a group of people, from differentdepartments or sectors

Analyst should guarantee that client’s interestare as well represented as possible.

An approach to support a group of decisionmakers to select na ERP system

Different results could appear if ELECTRE IV be changed to another method such as Borda Count or Condorcet.

Thank you.

Suzana Daher

Federal University of Pernambuco

Brazil

sfdd@uol.com.br

suzanadaher@gmail.com