Mathematical Problems in Engineering

Volume 2016, Article ID 7850960, 7 pages

http://dx.doi.org/10.1155/2016/7850960

## A Model for Selecting a Strategic Information System Using the FITradeoff

Departamento de Engenharia de Produção, Universidade Federal de Pernambuco, P.O. Box 7462, 50722-970 Recife, PE, Brazil

Received 17 January 2016; Accepted 8 June 2016

Academic Editor: Juan C. Leyva

Copyright © 2016 Ana Paula Henriques de Gusmão and Cristina Pereira Medeiros. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

This paper arose from the perceived need to make a contribution towards assessing a strategic information system by using a new method for eliciting the weights of criteria. This is considered one of the most complex and important stages in multicriteria models. Multicriteria models have been proposed to support decisions in the context of information systems given that problems in this field deal with many conflicting criteria. The new procedure for eliciting the weights of the criteria has the advantage of requiring less effort from the decision-maker and, thus, the risk of inconsistent answers is minimized. Therefore, a model based on this new procedure is proposed and applied using data from a glass packaging factory that needs to select a single information system from a set of systems previously identified as relevant. The results obtained are consistent both with the performance of alternatives and with the additive model used to evaluate the alternatives.

#### 1. Introduction

As observed by Arvidsson et al. [1], information systems (IS) have a strategic role within organizations since they are used to bring about strategic intent. Industrial production managers, service providers, and business administrators invest in IS with a view to obtaining greater efficiency, agility, and security in their operations. Although the market offers a variety of IS which serve different areas within organizations, it is not possible, in most organizations, to invest simultaneously in all the systems that they require. Usually, the resources (whether these have to do with financial resources, time, workforce, infrastructure, and so on) required for these investments are scarce.

Due to the relevance of this subject, academic researchers have devoted special attention to selecting IS, which is one of the three strands within IS strategy research, identified by Chen et al. [2]:(i)IS and business strategy alignment: where models to provide a strategic alignment between IS and business strategy [3] and models to analyze the positive impact of this alignment [4] are developed.(ii)Strategic IS planning to identify portfolios of systems: where the aim is to define a set of systems that jointly contribute to the organization [5, 6]. In some models, depending on the objective and particularities of the organization, the aim is not to define a portfolio but simply to select or to rank the systems [7].(iii)To use a specific system and assess its contribution towards adding to competitive advantage [1, 8].

Regarding the second strand, a large number of intuitive and analytical models have evolved over the last twenty or so years to assist decision-makers (DMs) in evaluating IS projects [9]. According to Zandi and Tavana [9], the main methodologies used for selecting and prioritizing IS projects can be divided into single-criterion cost benefit analysis, multicriteria scoring models and ranking methods, and portfolio methods. A subjective committee is also identified by Chen [10] as a class of methodologies used for IS project selection. The single-criterion cost benefit analysis is illustrated in [11], where Research and Development (R&D) projects are evaluated, and in [12], where R&D projects are selected by using an expected utility approach. The use of multicriteria scoring models and ranking methods for IS project selection is demonstrated in [13], which uses a multiobjective decision model, and in [14], where an analytic network process and goal programming are applied. Currently, the use of multicriteria methods designed to handle portfolio problems has gained special attention within the context of IS. de Almeida and Vetschera [15] propose a method to correct a problem on scale transformations in the PROMETHEE V method, which can be used to support portfolio decision problems. Other papers demonstrate the use of a portfolio method such as [16]. A fuzzy set approach for R&D portfolio selection is proposed by [17].

Initially models that evaluate the financial return on investment in IS were proposed to assist the selection of IS projects. However, the process for doing so requires several conflicting objectives to be analyzed but these cannot be measured monetarily. Thus, multicriteria decision-making/aiding (MCDM/A) methods are gaining in importance due to their inherent ability to judge different alternatives and to rank them.

The aim of this paper is to put forward a contribution in the field of strategic IS selection by using the FITradeoff (Flexible and Interactive Tradeoff) method, proposed by de Almeida et al. [18]. The application presented here was conducted using data from IS investments in a factory of one of the largest manufacturers of glass packaging in the world. These data were also used in an application presented by de Gusmão et al. [19] when the ELECTRE IV method was used by different DMs to rank the alternatives and, at the end of the model, the ranks were aggregated. Here, the alternatives are ranked according to an additive model in MAVT (Multiattribute Value Theory) scope, using the FITradeoff method for eliciting scaling constants or weights of criteria. The results of the two applications cannot be compared since the first application is conducted using a noncompensatory method while the value function of MAVT is compensatory. However, in both cases, the issue of defining the weights of criteria is undertaken (automatically by the method used) so as to minimize the effort that DMs need to make with regard to this.

This paper is organized as follows: Section 2 discusses the importance of multicriteria models in many instances of decision-making and presents an overview of the FITradeoff procedures. The model proposed is briefly presented in Section 3 and applied in Section 4, which discusses details of the steps of the model. The results area is also analyzed in Section 4, thereby allowing the DM to test the robustness of the model. Finally some conclusions and insights that may prompt future research are considered.

#### 2. Background

##### 2.1. Multicriteria Models

When planning investments in IS, not only the financial implications of such investment, but also other objectives such as competitive advantage, market share, and future growth need to be assessed. Thus, MCDM/A methods are gaining in importance since they allow different alternatives to be evaluated and ranked or one subset to be selected [20, 21]. These methods consider several points of view, characterized as conflicting issues, thereby enabling an integrated assessment of the problem in question to be made [22].

According to Vincke [23], MCDM/A can be considered a set of methods developed to support decision problems faced by organizations and individuals. The problems can include selecting suppliers [24], planning maintenance [25, 26], IS selection [20], and evaluating critical technology for generating energy [27].

According to de Almeida et al. [18], defining the weight of criteria, in multicriteria decision, mainly in additive models, is considered a problem either because the DM does not understand the meaning of weight or because the DM does not have sufficient knowledge and information to define the weights. Due to some major difficulties and challenges in procedures for eliciting weights that have been identified and considering that the tradeoff elicitation procedure—the procedure most used in additive models—has a strong axiomatic foundation besides some inconsistencies, de Almeida et al. [18] propose the FITradeoff method.

Unlike traditional procedures, in the flexible elicitation procedure, the DM does not need to provide imprecise or incomplete information* a priori*. In the flexible elicitation process, there is an attempt to minimize the effort that the DM must make, when compared with considering the traditional procedures. Thus it is expected that less inconsistency occurs during the process [18].

The search for ways to make it easier to determine the parameters required in decision-making, including the weights of the criteria, is not something new. The ELECTRE TRI Assistant method, proposed by Mousseau et al. [28], for example, requires from the DM much less cognitive effort. In this case, the parameters are defined indirectly using holistic information given by the DM through assignment examples, which are alternatives assigned by the DM to categories according to his/her comprehensive preferences. The use of assignment examples makes sense since the ELECTRE TRI is a multicriteria model whose goal is to assign each alternative to one of the categories which are predefined. Also, Mousseau and Dias [29] propose a slight adaptation of the valued outranking relation used in ELECTRE III and ELECTRE TRI. Although the modified outranking relation keeps the complexity of inferring the weights and cutting level, the veto thresholds are inferred easily.

The last two examples refer to methods of outranking. Regarding additive models, Melo Brito et al. [30], for example, present the application of a multicriteria methodology to support the selection of repair contracts in a context where information is imprecise, that is, when it is not possible to assign precise values to importance parameters of the criteria used for contract selection, but unlike the FITradeoff there is no elicitation procedure. A decision support system (DSS) was proposed by de Almeida et al. [31] to support a DM to establish the weights of criteria in a multicriteria decision problem by using a flexible elicitation procedure.

##### 2.2. Overview of the FITradeoff Procedure

The usual notations adopted in elicitation of weights processes are used [32, 33]:(i) represents the vector of consequences of an alternative, considering all criteria.(ii) represents the weight for the criteria .(iii) represents the value function of the consequences for the criteria.Thus, according to Keeney and Raiffa [32] and Keeney [34], aggregates the value functions asAssuming that it is well known that, in additive models, the most appropriate denomination for is as a scaling constant and not as a weight, but, as in [18], in this paper the term weight will be used for the sake of simplification.

The procedure for applying the FITradeoff, like that for the traditional model, is divided into two parts:(i)Obtaining the orders of the weights , using the preference .(ii)Obtaining the values of , using the indifference relation .The first part allows an -dimension weight space () to be defined which is given byIn this notation, and from now on, it was assumed, as assumed in de Almeida et al. [18], that the criteria are ordered from the most relevant to the least relevant.

The second part is now begun and it is in this part that the difference between the procedure for the traditional model and the FITradeoff is seen. In the FITradeoff, it is not necessary for the DM to define an exact value (), which would denote the outcome of criterion for which the indifference is obtained between consequences, whereas the traditional method requires this. In the FITradeoff, this procedure requires the DM to specify a range from to that represents, respectively, the upper and lower limit that can assume. Thus, given any criterion , the following relations can be established:Thus, as shown in [18], as a result of the second part, a new weight space ( may be obtained, which is a subspace of (3), in which all the valid relations of type (4) are considered:For more details on the definition of , see Keeney and Raiffa [32] and Keeney [34].

The FITradeoff is operationalized by a DSS, which includes the following stages [18]:(1)Evaluating the intracriteria.(2)Ranking the weights of the criteria.(3)Attempting to solve the problem using the available set of weights.(4)Evaluating the DM’s preferences.At the end of Stage , a check is made on whether or not a unique solution has been obtained, that is, if an optimal alternative has been identified. The DSS classifies the alternatives in three situations: potentially optimal, dominated, or optimal. If a single solution is not found, the DM goes on to the next stage, namely, that of evaluating the DM’s preferences which can be divided into four steps:(4.1)Setting values for testing the distribution of weights.(4.2)Asking the DM to state his/her preferences.(4.3)Computing LPP.(4.4)Finalization.These four steps constitute the main stage of the FITradeoff [18]. The aim of using the heuristic presented in Stage (4.1) is to compute the value of , thereby minimizing the number of questions to the DM. The output of this step, and the input for Stage (4.2), is a new set of values for and . Based on this new set of values, the DM, in Stage (4.2), has three options: to see partial results; not to proceed in the system; or to proceed with a view to making a choice. This choice consists of defining if there is a preference or an indifference relation between two consequences. If there is a preference relation, the DM has to signal the preferred consequence. Depending on the consequence chosen, either or assumes the value. In the case of indifference, assumes the value.

Having made this choice, Stage (4.3) is started and a Linear Program Problem (LPP) model is run [18]. It is important to note that this LPP model is also applied in Step assuming and = 0 for all criteria . This LPP has the following objective function:Thus, the aim is to find an alternative , from the set of alternatives, that has the maximum value given in (1) in accordance with the weight of criteria space given by (5). Therefore, it is necessary to consider some constraints in the LPP. The relations (4) are introduced as constraints on how strict inequality is avoided.

Also, what should be considered to solve the problem (the optimal alternative) is that the maximum value of the alternative should be greater than (or equal to) any other alternative in the subset. Thus, the following constraint has to be considered:This LPP runs until an optimal alternative is found. If this does not happen, the dominated alternatives are eliminated and the process is started again, from Step (4.1). Now, just the alternatives identified as potentially optimal are considered in the subsequent steps.

#### 3. Model Proposed

As already explained, the aim of this paper is to put forward a contribution in the field of strategic IS selection by using FITradeoff and based on the data presented in [19]. Thus, based on the framework presented in [19] and on the general procedure of the FITradeoff, the model proposed can be structured as shown in Figure 1.