You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266079064

Multi-criteria decision methodology for selecting maintenance key


performance indicators

Article · September 2014

CITATION READS

1 181

1 author:

José Mendonça Dias


New University of Lisbon
18 PUBLICATIONS   127 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

OPTIMIZAÇÃO DO PERÍODO DE SUBSTITUIÇÃO PREVENTIVA DE COMPONENTES EM FUNÇÃO DOS CUSTOS View project

All content following this page was uploaded by José Mendonça Dias on 25 September 2014.

The user has requested enhancement of the downloaded file.


This article was downloaded by: [b-on: Biblioteca do conhecimento online UAlgarve]
On: 12 September 2014, At: 12:07
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Management Science and


Engineering Management
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tmse20

Multi-criteria decision methodology for selecting


maintenance key performance indicators
a b
César Duarte Freitas Gonçalves , José António Mendonça Dias & Virgílio António Cruz
b
Machado
a
Departamento de Engenharia Mecânica, Instituto Superior de Engenharia, Universidade
do Algarve, Campus da Penha, 8005-139 Faro, Portugal
b
Departamento de Engenharia Mecânica e Industrial, Faculdade de Ciências e
Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal
Published online: 10 Sep 2014.

To cite this article: César Duarte Freitas Gonçalves, José António Mendonça Dias & Virgílio António Cruz Machado (2014):
Multi-criteria decision methodology for selecting maintenance key performance indicators, International Journal of
Management Science and Engineering Management

To link to this article: http://dx.doi.org/10.1080/17509653.2014.954280

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
International Journal of Management Science and Engineering Management, 2014
http://dx.doi.org/10.1080/17509653.2014.954280

Multi-criteria decision methodology for selecting maintenance key performance indicators


César Duarte Freitas Gonc alvesa*, José António Mendonc a Diasb and Virgı́lio António Cruz Machadob
a
Departamento de Engenharia Mecânica, Instituto Superior de Engenharia, Universidade do Algarve, Campus da Penha, 8005-139
Faro, Portugal; bDepartamento de Engenharia Mecânica e Industrial, Faculdade de Ciências e Tecnologia, Universidade Nova de
Lisboa, 2829-516 Caparica, Portugal
(Received 11 April 2014; final version received 8 August 2014)

Many key performance indicators (KPIs) can be used for measuring maintenance performance. Companies must implement
a suitable set of KPIs to measure the performance of maintenance services and the influence of practices implemented in
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

their activities. Establishing a useful set of maintenance KPIs depends on a company’s maintenance objectives and is highly
related to specific business contexts, strategies, processes, and systems. Maintenance managers deal with the complex tasks
of finding the best performance indicators that can help them to achieve goals. This paper presents a new approach to
selecting relevant maintenance KPIs using a methodology based on the original ELECTRE I, which is a multi-criteria
decision making method. The proposed methodology, which involves decision maker’s preference information, determines
a ranking of possible alternatives following its evaluation according to important criteria. To validate the methodology, a
case study is presented. The results show that this methodology is an effective tool to aid maintenance managers in accurate
KPI selection tasks according to maintenance objectives and strategies. The proposed methodology makes the decision
process more explicit, rational, and efficient.
Keywords: Maintenance; KPI; multi-criteria decision making; MCDM; ELECTRE; SRF

JEL Classification: C44; C52; C88; L00; L25

1. Introduction problem. In this paper a methodology is proposed using


Maintenance management is an important activity in MCDM methods that deal with such issues in order to
companies; it has to do with ensuring the smooth operation assist maintenance managers.
of facilities, systems, and equipment, and that good ELECTRE is an MCDM method that draws on the
operating conditions with maximum availability are logic of the outranking model (Roy, 1991) and that has
achieved, all at an optimized overall cost. Companies been applied to various types of decision making
must strive to optimize all their activities to obtain problems. Based on the ELECTRE method, the proposed
satisfactory levels of productivity in order to stay methodology was developed to help decision makers
competitive in business. Maintenance objectives are resolve ranking problems, by evaluating several conflict-
highly related to the company’s specific business context, ing alternatives according to multiple criteria.
strategies, processes, and systems. Managers need This paper describes the methodology developed,
information about maintenance performance for planning which joins the original ELECTRE I, which is an MCDM
and controlling maintenance activities. method for choice type problems, and a complementary
Since measuring maintenance performance has analysis to allow the ranking of alternatives. For eliciting
become an essential element of strategic thinking, the weights of criteria, the SRF (Simos – Roy –Figueira)
maintenance managers need a good performance track software is proposed, which uses an algorithm of the
record vis-à-vis maintenance processes and results, and revised Simos procedure (Figueira & Roy, 2002).
this can be achieved through the development and The methodology is proposed for selecting relevant
implementation of a rigorously defined performance key performance indicators (KPIs) regarding maintenance.
measurement framework with relevant indicators that are Because the decision maker is typically not familiar with
able to assess important elements of maintenance function the mathematical procedures of MCDM methods, an
performance. informatics tool was conceived. Using the popular
Establishing a useable set of maintenance performance software MS Excelw, by means of Visual Basicw for
indicators depends mainly on maintenance objectives and Applications (VBA), the mathematical procedure of the
the company’s goals. Maintenance performance measure- methodology has been programmed to provide compu-
ment is, in fact, a multidisciplinary process taking into tational analysis of decision problems.
account multiple aspects of maintenance activities, and for In order to validate the proposed methodology, a case
this reason, the selection of the best maintenance study based on reality is presented. A number of possible
performance indicators is a complex task, which can be KPIs for measuring maintenance service quality in an
formulated as a Multi-Criteria Decision Making (MCDM) airport were evaluated by two expert maintenance decision
makers. The methodology ranked the alternatives, from
best to worst, and consistently selected KPIs with greater
*Corresponding author. Email: cgoncal@ualg.pt interest to the decision makers.
q 2014 International Society of Management Science and Engineering Management
2 C.D.F. Gonc alves et al.

The paper is organized into seven sections. After this large number of indicators to measure each maintenance
introduction, a maintenance KPIs approach is presented in aspect hinders comprehension and the work for which they
Section 2. A review of MCDM is made in Section 3, where are developed.
some studies are highlighted, relating mainly to mainten- Different categories of KPIs and distinct frameworks
ance. In Section 4 a methodology for selecting relevant have been discussed extensively and proposed in the
maintenance KPIs is proposed. The application of the literature to monitor and control maintenance activities.
proposed methodology is undertaken in Section 5. Section However, few publications propose methodologies for
6 reports results and discusses methodology validation to selecting relevant KPIs, especially in the field of
assist the maintenance managers in the KPIs selection maintenance. Muchiri et al. (2011) mention that the
process. Finally, Section 7 summarizes and makes final literature mainly proposes lists of KPIs but lacks
remarks. methodological approaches for selecting them. These
authors are also of the opinion that organizations should
resort to suitable models to identify relevant KPIs for a
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

maintenance function in a certain context. The selection of


2. Maintenance Key Performance Indicators (KPIs) KPIs has usually been derived from the needs and
Companies typically use KPIs to measure the success of experience of maintenance managers (Kumar et al., 2013b).
the activities in which they are engaged. Maintenance can
serve as a management system of appropriate indicators to
measure its own performance, that is, to measure both the
expected and actual results of its activities. 3. Multiple-criteria decision making
A performance indicator is defined by the European Multiple-Criteria Decision Making (MCDM) or Multi-
Standard EN 15341 (2007) as a measured characteristic (or Criteria Decision Analysis (MCDA) are different designa-
set of characteristics) of a phenomenon, according to a tions for the same concept. These well-known terms are
specific formula that evaluates its evolution. EN 15341 concerned with analysing and solving decision problems
provides three categories and three levels of maintenance involving multiple criteria, and are considered a sub-
performance indicators to appraise and improve efficiency discipline within operations research. Their purpose is to
and effectiveness in the maintenance of technical assets. aid decision makers when multiple alternatives are
This standard also presents some steps as a methodology evaluated by multiple criteria, which are often conflicting.
for the selection and use of maintenance performance Some authors refer to these matters by a third name,
indicators. In order to select relevant KPIs, EN 15341 Multiple-Criteria Decision Aiding (MCDA) (Figueira,
prescribes that the first step is to define the objectives to be Greco, Roy, & Slowinski, 2013) as a result.
reached. The next step is to find KPIs that allow the An MCDM approach refers to making preference
measurement of parameters or aspects related to the decisions over the available alternatives that are
objectives defined. Then, in the search for relevant KPIs, a characterized by multiple incompatible criteria. MCDM
first approach can be made, by choosing from among the problems usually involve discrete variables, which are a
list of existing indicators, which after analysis prove to be finite number of pre-specified alternatives to be evaluated
of interest for measuring the performance of maintenance against the criteria set. In turn, when the number of
of the intended aspects. Other KPIs can be developed for alternatives is infinite, a Multiple Objective Decision
specific measurement aspects. Making (MODM) approach is used. In the case of MODM,
A relevant indicator is a key element in the decision multiple objective programming models are employed to
making, which means that its evaluation and its data optimize a set of conflicting objectives subjected to a set of
should be related to the performance parameter to be mathematically defined constraints.
measured and in accordance with the defined objective. In the context of decision aid, Roy (1991) and Figueira
Some authors provide guidelines for choosing mainten- et al. (2013) note three main types of decision problems:
ance KPIs that seek to align maintenance objectives with ranking, sorting, and choice. Ranking problems call for
production objectives (Kumar, Soni, & Agnihotri, 2013b; rank ordering all alternatives from the best to the worst, i.e.
Muchiri, Pintelon, Gelders, & Martin, 2011) which are alternatives are ordered by decreasing preference; sorting
important for the success of performance measurement problems call for assigning each alternative to an
and for satisfying customers’ needs (Pacaiova, Nagyova, appropriate predefined class and according to the ordered
Namesanska, & Grencik, 2013). preference; choice problems call for selecting a single best
Regardless of the type of indicators that are used in alternative or a smallest subset of alternatives considered
measuring the performance of maintenance, there must be as the best or incomparable alternatives. Many authors
a procedure for their accurate identification, namely an favor the high potential of different MCDM methods.
algorithm that can be used to establish a group of There are numerous MCDM methods and the literature is
indicators that are representative in terms of an extensive in this field.
explanatory value of a given process (Pacaiova et al., The acronym ELECTRE stands for ‘ELimination Et
2013). The number of indicators used for each department Choix Traduisant la REalité [Elimination and Choice
in an organization should be limited through the Expressing Reality]’ and is a well-known MCDM method
identification of key features or key factors (Kumar, introduced by Bernard Roy in the late 1960s. The
Galar, Parida, Stenstrom, & Berges, 2013a). Having a ELECTRE approach uses a preference model based on
International Journal of Management Science and Engineering Management 3

the construction of ‘outranking relations’ by paired key performance indicators for a maintenance manage-
comparisons between the alternatives using the concept of ment framework.
concordance and discordance. ELECTRE I was the first There are several reasons why the ELECTRE I method
decision aid method using the concept of outranking is considered suitable for selecting relevant KPIs. First, the
relations (Roy, 1991) and was originally designed to lead to method has been successfully applied to other ranking
‘choice-type’ results. The ELECTRE approach has evolved problems, as mentioned above. The method is simple to
into a number of variants based on the same concept apply and many authors, such as Yazdani-Chamzini et al.
covering the different types of decision problems described (2013), consider that the ELECTRE logic is rational and
above, see Figueira et al. (2013). A complementary analysis the computation process is systematic and well organized.
to the ELECTRE I method that introduces the calculation of Furthermore, ELECTRE is the most commonly used
aggregate differential concordance and discordance values MCDM method for ranking alternatives (Zandi &
(net concordance and net discordance values), also exists Roghanian, 2013).
(van Delf & Nijkamp, 1976), which allows for the ranking Some authors discuss some advantages of the ELE-
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

of alternatives from best to worst. CTRE method when compared to the limitations of other
Some approaches in the literature dealing with the methods such as AHP and ANP. The ELECTRE family of
original ELECTRE I method deserved attention in the methods can deal with heterogeneous scales, i.e. different
current research. For instance, there are authors proposing scales for each criterion. Whatever the nature of the scale,
methodologies of decision making in supplier selection the procedure can run while preserving the original
(Sevkli, 2010; Hatami-Marbini & Tavana, 2011), evalu- performance of the alternatives on the criteria (Figueira
ation of transport sustainability (Bojković, Anić, & Pejčić- et al. 2013). Yazdani-Chamzini et al. (2013) observe that,
Tarle, 2010), bioinformatics investigation of human genes because there is no need to impose a rating scale, the
(Ermatita, Hartari, Wardoyo, & Harjoko, 2011), environ- alternatives keep their original concrete verbal meaning.
mental impact assessment (Kaya & Kahraman, 2011), risk In addition, the decision maker does not make any
assessment in tunnel construction (Yazdani-Chamzini, pairwise comparison, since all the comparisons are carried
Yakhchali, & Mahmoodian, 2013) and tourism manage- out by the procedure itself (Figueira et al. 2013). The
ment (Botti & Peypoch, 2013). In another approach, the ELECTRE method does not allow for compensation of
original ELECTRE I and the complementary analysis performances among criteria, i.e. the degradation of
through the net concordance and net discordance values performances on certain criteria cannot be compensated
were used for selecting appropriate materials (Shanian & for by improvements of performances on other criteria
Savadago, 2006). With regard to maintenance, there is also (Figueira et al. 2013).
a fuzzy model based on the original ELECTRE I method Saaty and Ozdemir (2003) call attention to Miller’s
for solving maintenance strategy selection problems theory (Miller, 1954), wherein seven plus or minus two
(Vahdani & Hadipour, 2011). represents the greatest amount of information that an
Two recent approaches to the application of MCDM observer can give about an object on the basis of
methods are also to be found, both using the Analytic judgement. Thus, an upper limit of seven plus or minus
Network Process (ANP), an extension of the Analytic two elements in AHP analyses is suggested. This is indeed
Hierarchy Process (AHP) method: one, for choosing the a limitation of the AHP, advising against its use in solving
most appropriate practices and KPIs in a supply chain decision problems with a large number of alternatives and
(Cabral, Grilo, & Cruz-Machado, 2012); and the other, for criteria.
helping the maintenance manager in the definition and The ELECTRE method, on the other hand, can deal
selection of the relevant KPIs (Horenbeek & Pintelon, with a large number of alternatives. Hatami-Marbini and
2014). However, these methods have some limitations. For Tavana (2011) mentioned that ELECTRE is the preferred
instance, Cabral et al. (2012) mention that the main method for problems with a large set of alternatives and
disadvantages of their proposed ANP model are the large few criteria.
number of comparisons needed and the inconsistency Within the ELECTRE family methods, the decision
problems. Horenbeek and Pintelon (2014) emphasize that, maker is asked to assign numerical values to the weights of
in the application of their ANP approach, sometimes the the criteria. The calculation of the priority vector of the
decision makers encountered difficulty in expressing AHP is mentioned by many authors as a method to be used
preferences using the defined ratio scale. Another remark for determining the weights of a set of activities or criteria
of these authors is that an ANP approach may require (Horenbeek & Pintelon, 2014; Kaya & Kahraman, 2011;
considerable effort and time from the decision makers. Yazdani-Chamzini et al., 2013). However, Bana e Costa
Decision making in the selection of indicators for and Vansnick (2008) address a critical problem regarding
maintenance performance measurement can be aided by the priority vector and the consistency of the ratio of AHP.
conducting MCDM ranking methods. The values of the priority vector do not always preserve
the order of the respective intensities of preferences,
although the ratio proposed to assess the consistency of
human judgments does not detect disturbances. In other
4. The proposed methodology words, the weights calculated by the priority vector of the
A methodology is proposed herein that makes use of the AHP method may not correctly reflect the intensity of
original ELECTRE I method to select the most important judgments established by the decision maker by compar-
4 C.D.F. Gonc alves et al.

ing pairs of criteria. For that reason, the use of AHP is also set the evaluation criteria according to the predefined
not advised for calculating the weights of the criteria. objectives to be reached by the maintenance department.
For assessing those weights, the proposal here is the It is also necessary to propose a battery of indicators
revised Simos procedure aided by SRF (Simos –Roy – (KPIs) focused on the maintenance aspect to be measured.
Figueira) software (Figueira & Roy, 2002). This method That cluster of KPIs is the set of competing alternatives
attributes an intrinsic weight to each criterion without which are evaluated by the decision maker against the set
dependence on either the range of the scale or on the of criteria.
encoding to express the score on this scale. On the other The evaluation of alternatives represents the convic-
hand, the decision maker has the possibility to express her/ tions of the decision maker, who can mirror her/his
his preferences spontaneously and achieve the outputs in a preferences by assigning a level of importance for each
coherent way without any application of weighted sum alternative. For defining the assessment scale, we followed
aggregation techniques. This means that the value of each the principles presented by Likert (1932) and Miller
criterion is fixed without taking into account the impact (1954). Likert (1932) indicates that the scales to capture
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

that other criteria weight can have. Augusto, Figueira, and the perception must present symmetrical options, and
Lisboa (2005) and Figueira et al. (2011) used SRF Miller (1954) complements these guidelines considering
software to determine the weights of the criteria for the that in addition to symmetry, the scales should provide five
resolution of different decision problems through different or nine response options. A Likert scale assessment of nine
ELECTRE methods. response options is proposed with which the decision
Ranking is very different from choice or sorting and maker can specify her/his level of agreement with each
therefore requires the use of specific methods. ELECTRE I alternative in relation to different criteria. The higher the
is a simple method for choice type problems (Roy, 1991). response level in a criterion, the more important is the KPI.
The ELECTRE method is applied in combination with
Nevertheless, their outputs can be worked upon to obtain a
the calculation procedures of the NC and ND values.
ranking of alternatives (Hatami-Marbini & Tavana, 2011;
Ordering the set of alternatives in descending order of the
Shanian & Savadogo, 2006; van Delf & Nijkamp, 1976;
NC values, a ranking list of all alternatives from the best to
Yoon & Hwang, 1995; Zandi & Roghanian, 2013).
the worst is obtained. The final selection should satisfy the
Toward that goal, calculation of the Net Concordance
condition that its NC value is at a maximum and its ND
(NC) and Net Discordance (ND) values is proposed. These
value at a minimum (Yoon & Hwang, 1995). The interest
methods make it possible to rank and select the most
is to find a set of maintenance performance indicators that
relevant maintenance KPIs following their evaluation by a
satisfy the conditions to be accepted, and that means
decision maker. having the highest NC values and the lowest ND values
Methodological procedures are given in Figure 1 in simultaneously.
order to illustrate the steps in the selection of indicators. A sensitivity analysis must be conducted in order to
The conceptual framework is intended to aid decision contrast the position achieved by the different KPIs,
makers using the methodology. considering their value of discordance. Moreover, the
For the selection of relevant indicators, measuring a decision maker must determine if some KPIs that were
particular aspect of maintenance is necessary in order to ranked as being preferred are still of interest and should be
selected, or if the measuring maintenance aspect is ensured
1 by another KPI already selected.
Definition of the maintenance
management objectives The procedures to obtain the weights of criteria by the
SRF software are described in the literature (Figueira &
Roy, 2002; Gonc alves, Dias, & Cruz-Machado, 2014). The
2
Specification of the criteria original ELECTRE I method and the calculation of NC
for evaluation of the KPIs and ND values can also be found in the literature
(Gonc alves et al., 2014; Shanian & Savadogo, 2006; van
3 Delf & Nijkamp, 1976; Yoon & Hwang, 1995). The
Determination of the weights of
criteria using the SRF software
mathematical procedures were programmed in ‘Excelw
VBA’ code to provide computational results. The main
goal is to provide a practical and effective informatics tool
4 for the decision making process with the proposed
Pre-selection of KPIs as
competing alternatives approach.

5
Evaluation of the KPIs
according to each criterion
5. Application – case study
Suppose a maintenance department of an important airport
6 realized that there should be a relevant set of indicators for
Ranking of KPIs using ELECTRE I
and calculation of NC and ND values maintenance service quality measurement of their
facilities. Two maintenance managers find it difficult to
Figure 1. Conceptual framework of the proposed methodology. establish strict preference judgements, due to the vast
International Journal of Management Science and Engineering Management 5

number of alternatives. The fundamental questions are: reality, the data interpretation, and the analysis of
‘Which KPIs must be selected to meet the maintenance results without compromising the effectiveness of
quality measurement needs?’ and ‘How many KPIs must maintenance management in decision making.
be selected to assist maintenance managers?’ The Sharing results (C5): to evaluate how the results of the
proposed methodology was applied to select an optimal KPI are of interest for being shared with other
set of KPIs that best measures the performance of the company sectors, employees, or customers; how the
maintenance sector with regard to the quality of their results of the KPI are of interest for informing and
service and achieving the goals of the organization. The motivating employees; to evaluate the common
decision process for resolving the problem is summarized interest between maintenance and other sectors of the
as follows. company in sharing the results of the KPI.
Step 3: Criteria weights determination. Maintenance
Step 1: Maintenance management objectives. It is managers were called upon to elicit the relative
necessary to understand the needs of the organization importance of criteria in an indirect way by using SRF
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

in order to answer the question of which indicators are software. The data required for the revised Simos
relevant to the business and related to maintenance. procedure used by the SRF software algorithm are
In general, the organization’s strategic goals are the given in Table 1. Through this software, the
efficient management of airport infrastructure, ensur- appropriate values for the weights of the five criteria
ing productivity levels and quality of service to were generated, which are also presented in Table 1 as
passengers and airlines, continuously improving the normalized weights.
ideal maintenance service, and lowering costs. Step 4: Pre-selection of KPIs. A set of competing
The main maintenance objectives were defined by alternatives, assuming that all are possible best
the maintenance department and can be summarized as choices, was pre-selected. The same indicators often
follows: provide continuous improvement of the appear repeatedly in the literature. Muchiri et al.
facilities; ensure quality of airport service; optimize (2011) provide a well organized structure of indicators
maintenance activities; allow proactive management identified as key elements to measure different features
and agility; achieve high availability and reliability of of maintenance management. Among them, the
physical assets; provide shared vision of maintenance authors and airport maintenance managers pre-selected
performance among employees; and provide a safe the ones important as alternatives. KPIs were also
work environment with motivated employees.
collected from the standard EN 15341. Airport
Step 2: Specification of criteria. The first way to
maintenance managers have used a pool of KPIs to
approach the problem is to generate appropriate
measure and manage maintenance activities, from
criteria as constraints in evaluating the set of
which some were also collected for the present study.
alternatives. The literature review revealed the
Additionally, the authors developed four KPIs. Table 2
implications in KPI selection. Three maintenance
presents a set of 20 KPIs having possible interest for
management experts were invited to identify the
measuring maintenance performance in terms of
important criteria. In a subsequent analysis with airport
service quality.
maintenance managers, the focus was on the main
Step 5: Evaluation of KPIs against criteria. The two
matters for determining the most important criteria in
maintenance managers of the airport evaluated the pre-
the context of maintenance service quality measure-
selected KPIs against the criteria together. They
ment. From all of this, five decision criteria were
settled on and are described as follows. discussed the importance of each KPI and for each
Maintenance quality (C1): to evaluate how the KPI criterion a Likert scale value was assigned. The
quantitatively measures the quality aspects of main- performance of the 20 alternative KPIs evaluated
tenance service; how it allows measuring and according to the five criteria is also shown in Table 2.
comparing maintenance performance, identifying the Step 6: Ranking of KPIs. Making use of the
strengths and weaknesses of maintenance service. informatics application built into VBA, 20 rankings
Maintenance objectives (C2): to evaluate how the KPI of alternatives were provided. The preference rank
measures the maintenance state in the fulfillment of
established objectives for maintenance and according Table 1. Weights of criteria generated by RSF software.
to company goals; how the KPI assists the manage- Criteria Criteria Blank Non-normalized Normalized
ment in setting goals in a process of continuous code rank cards weights weights
improvement.
KPI influence (C3): to evaluate how the result of the C5 1 – 1 224
– – 2 – –
KPI can influence decision making in maintenance; C2 2 – 8 1791
how the KPI helps to guide future decisions, to track C3 2 – 8 1791
progress and changes over time. – – 1 – –
Understanding and interpretation of KPI (C4): to C1 3 – 1267 2836
evaluate how the KPI expresses a measurement that – – 0 – –
C4 4 – 15 3358
easily translates an obtained or expected real result; to Total – 3 (Z value ¼ 15) 100
evaluate how easily the KPI gives a perception of
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

Table 2. Evaluation of the proposed KPIs against criteria.

Proposed KPIs (competing alternatives) Evaluation against criteria


N Description Calculation formula Source C1 C2 C3 C4 C5
1 Rate of incidents due to causes imputed to maintenance No. of incidents with causes imputed to maintenance/(No. of Airport 8 9 9 8 7
(per 1000 passengers) passengers/1000)
2 Average response time to emergency calls (incidents) Total response time to emergency calls/No. of emergency calls Airport 7 8 8 7 7
3 Average time to replacement of functionality Total replacement time/No. of incidents Airport 9 8 7 9 5
4 Availability related to maintenance Total operating time/(Total operating time þ downtime due to EN 15341 (2007) 8 7 8 8 8
maintenance)
5 Availability due to planned and scheduled maintenance Total operating time/(Total operating time þ downtime related to EN 15341 (2007) 9 8 7 7 7
the planned and scheduled maintenance)
6 Mean time between work orders that cause downtime Total operating time/No. of maintenance work orders causing EN 15341 (2007) 8 5 1 5 7
downtime
7 Percentage of systems covered by criticality analysis No. of systems covered by a critical analysis/Total no. of systems EN 15341 (2007) 7 4 4 7 6
8 Mean Time To Repair (MTTR) Total time to restore/No. of failures EN 15341 (2007) 8 6 8 7 8
9 Percentage of working time lost due to accidents Man– hours lost due to injuries of maintenance personnel/Total EN 15341 (2007) 5 7 8 7 9
(severity rate) man– hours worked by maintenance personnel
10 Fulfilment rate of work orders No. of work orders performed as scheduled/Total no. of scheduled EN 15341 (2007) 9 8 9 8 9
work orders
11 Percentage of work orders in backlog No. of overdue work orders/No. of received work orders Muchiri et al. (2011) 9 9 9 9 9
12 Percentage of available man– hours used for proactive Man– hours used for proactive work/Total man– hours available Muchiri et al. (2011) 7 8 7 4 7
work
C.D.F. Gonc alves et al.

13 Percentage of work orders requiring rework (quality of No. of work orders to rework/Total no. of work orders Muchiri et al. (2011) 9 9 9 8 9
execution)
14 Rate of planned work (planning intensity) Planned work/Total work done Muchiri et al. (2011) 8 3 4 8 8
15 Quality of planning No. of work orders requiring rework due to planning/Total no. of Muchiri et al. (2011) 9 8 8 8 9
work orders
16 Quality of scheduling No. of work orders with delayed execution due to material or Muchiri et al. (2011) 9 8 9 7 6
manpower/Total no. of work orders
17 Execution Index Hours of work orders planned and executed/Expected hours of Developed 7 9 8 7 7
work orders planned
18 Rate of customer complaints due to causes imputed to No. of complaints with cause imputed to maintenance/(No. of Developed 8 7 9 8 9
maintenance (per 1000 passengers) passengers/1000)
19 Maintenance Quality Index Total hours of maintenance/Rework hours Developed 8 9 8 5 8
20 Maintenance Performance Index Expected hours of maintenance/Hours used for maintenance Developed 3 7 6 5 5
International Journal of Management Science and Engineering Management 7

Table 3. Ranking of KPIs by concordance in descending order. work done to ensure that maintenance work is not
left to chance. Work orders are reworked as a result
Net values Ranking
Alternative, of low quality work or work that was initially poorly
N NC ND Rank C Rank D done.
. In third place was ranked the ‘fulfillment rate of
11 152,387 2190,000 1 1
13 122,165 2160,161 2 2 work orders’, which is identified by the standard EN
10 100,673 2128,535 3 3 15341 as an organizational indicator. This indicator
3 82,534 2101,294 4 4 provides the percentage of work executed within the
1 77,981 2 94,813 5 5 scheduled time and that also fulfills the explicit
15 77,390 2 91,803 6 6 tasks in the work orders.
16 47,389 2 50,338 7 8
. The ‘average time to replacement of functionality’
18 41,268 2 52,597 8 7
4 15,745 2 19,931 9 9 and the ‘number of incidents due to causes imputed
5 07,988 2 14,539 10 10 to maintenance per each thousand passengers’ were
17 2 23,510 2 11,079 11 11
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

the KPIs ranked in fourth and fifth places,


19 2 23,654 63,972 12 14 respectively. These indicators were already used in
14 2 34,403 88,398 13 15
8 2 40,222 47,772 14 13 the airport to monitor the efficiency and effective-
2 2 45,002 31,863 15 12 ness of the maintenance department in carrying out
9 2 74,403 103,287 16 16 its job. They can help to manage and optimize the
12 2 109,924 131,710 17 18 planning of maintenance activities, seeking con-
7 2 113,061 124,794 18 17 tinuous improvement of the airport facilities and
6 2 115,444 154,671 19 19
20 2 145,897 168,622 20 20 quality of service to passengers and airlines.
. In the sixth rank was placed the ‘quality of planning’,
which obtains the percentage of work orders requiring
based on the NC is given in Table 3, where the KPIs are rework due to failures in planning. Planning is
presented in descending order of the NC value. essential to maximize the efficiency of maintenance
and to ensure the necessary time to carry out the tasks
The proposed methodology allowed the ranking of according to the required procedures. Managers
KPIs according to the decision makers’ preferences. From report that sometimes teams have to repeat or add
the analysis of the computed results, taking into account tasks as a result of poor job planning. It also happens
the ranking by concordance versus discordance, a set of that there are tasks that have to be repeated owing to
relevant KPIs can be selected. insufficient time having been allocated to them.
According Muchiri et al. (2011), this is an important
indicator for measuring work planning.
6. Results and discussion Note that with only the six highest-ranked KPIs, some
The six top-ranked alternatives are classified with clear important aspects of maintenance quality will not be
evidence, where ‘Rank C’ equals ‘Rank D’. That is, they measured. Based on the experience of the authors and
achieved the same ranking, both in descending order of the airport maintenance managers, it was observed that the
NC value and in ascending order of the ND value. For two subsequent ranked KPIs should be selected consider-
subsequent ranked alternatives, although both classifi- ing the analysis as follows.
cations do not establish the same order, there are no very
significant differences. Thus, it establishes a balanced . The KPI ranked in seventh position is entitled
pattern of preferences in the selection of the main KPIs for ‘quality of scheduling’ and provides the percentage
measuring maintenance service quality at the airport. of delayed work orders due to lack of materiel or
The human brain can only handle four to eight manpower. The maintenance department must
measurements intended to quantify one aspect (Kumar ensure that all the necessary resources are available
et al., 2013a), which suggests that it is reasonable to use six before the work commences. Apart from ensuring
performance indicators for measuring each maintenance that planned activities are realized within the
aspect. Therefore, the six top KPIs were analysed as expected time, it is essential that they be scheduled
follows. in light of the availability of skilled manpower and
the necessary materials for work. According Muchiri
. The top two KPIs are identified by Muchiri et al. et al. (2011), this is a work scheduling indicator that
(2011) as indicators to measure the performance in should not be ignored.
work execution. The ‘percentage of work orders in . It was realized that there should be a measurement
backlog’ and ‘percentage of work orders requiring that relates potential customer dissatisfaction with
rework’ were validated to measure how scheduled maintenance performance. That information can be
activities are carried out within the allocated time given by the eighth ranked KPI, which was developed
and how maintenance work is done effectively. The by the authors in order to provide a rate of customers’
maintenance managers consider important the (passengers and airlines) complaints due to causes
measure of schedule compliance and the quality of imputed to maintenance per thousand passengers.
8 C.D.F. Gonc alves et al.

The maintenance managers are in agreement that the maintenance managers, and the results were well accepted.
first eight KPIs measure the main aspects of maintenance A ranking list of alternatives was generated, from the best
service quality in the airport and are also linked to the to the worst, which was considered acceptable by the
organization objectives. To ease understanding in maintenance managers of the airport. The KPIs ranked in
evaluating an aspect of maintenance, Kumar et al. the first eight positions meet the needs of intended
(2013a) hold that a number of indicators greater than measurement, covering the main aspects to be ensured for
eight should be avoided. This belief also derives from maintenance service quality in airports.
Miller’s theory (Miller, 1954). The proposed methodology requires little effort and
Therefore, the eight KPIs ranked highest with respect time from the decision maker. It provides decision support
to the concordance and in which discordance shows no and guidance for the selection of KPIs aligned with the
dispersion in the set were selected as measures. The organization’s objectives. Moreover, it reduces the level of
selected KPIs cover the main aspects for measuring subjectivity with which the decision maker reflects her/his
maintenance service quality, such as the management and preference of relevant KPIs to measure a particular aspect
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

execution of work orders, the quality of work executed, the of maintenance.


control of events (activities, rework, incidents), the quality The proposed approach shows advantages when
of planning and scheduling of maintenance activities, the seeking to rank a wide range of alternatives. The
use of time and agility of working teams, and the methodology developed is both practical and adequate
contribution of maintenance to customer satisfaction. for maintenance managers that find it difficult to establish
The methodology quantitatively compares the interest direct preference in the selection of KPIs facing a large
of KPIs for maintenance, through concordance values and number of alternatives. Furthermore, the proposed
also by checking the discordance, according to the approach avoids the limitations found in using AHP and
preference and experience of managers. This approach is ANP. The decision maker evaluates the KPIs directly in
not intended to group multi-criteria decisions, which relation to criteria. There is no need to draw comparisons
implies the consensus of opinion of maintenance managers between KPIs, which become ambiguous when the
in determining the weights of the criteria and the decision maker has many doubts about them. This makes
evaluation of KPIs. the decision process more explicit, rational, and efficient.
Facility and consistency were achieved in the The authors believe that the proposed methodology
procedures for the evaluation of the criteria and can support both maintenance managers in industry and
alternatives according to the preference information of academics interested in this strand of research.
the decision makers. The procedures of the methodology The methodology presented is an innovation in this
were well accepted by maintenance managers and both proposed application area. In further work, the method-
informatics tools allowed for the easy collection of data ology can be extended to the ranking and selection of KPIs
and also the quick achievement of results. for other maintenance aspects or in different business
contexts. Future work can compare the results obtained in
this study with those achieved through the use of different
MCDM ranking methods.
7. Conclusions
A new approach has been presented in order to aid
maintenance managers in the difficult task of selecting Funding
relevant KPIs for measuring the performance of This work is supported by a cooperation protocol between
maintenance services. Instituto Superior de Engenharia of Universidade do Algarve
In the literature, few applications are addressed for (ISE/UALG) and Faculdade de Ciências e Tecnologia of
selecting KPIs for maintenance. The majority of these Universidade Nova de Lisboa (FCT/UNL). The authors
acknowledge the support of UNIDEMI – R&D Unit in
contributions address problems with compensatory Mechanical and Industrial Engineering of FCT/UNL.
methods such as AHP and ANP, of which considerable
limitations were found.
Some approaches were found that dealt with the original
ELECTRE I, but none for the selection of KPIs. ELECTRE I References
is an outranking method for choice type problems that, Augusto, M., Figueira, J., & Lisboa, J. (2005). An application of
combined with net concordance and net discordance values multi-criteria approach to assessing the performance of
calculation, can provide ranking results. A simple ranking Portugal’s economic sectors – methodology, analysis and
MCDM methodology based on the original ELECTRE I implications. European Business Review, 17(2), 113– 132.
Bana e Costa, C. A., & Vansnick, J. -C. (2008). A critical analysis
method has been proposed that proved to be a suitable and of the eigenvalue method used to derive priorities in AHP.
efficient tool for dealing with the type of problems presented European Journal of Operational Research, 187, 1422– 1428.
in the case study. However, the methodology can be used Bojković, N., Anić, I., & Pejčić-Tarle, S. (2010). One solution for
even when dealing with ranking problems that are different cross-country transport-sustainability evaluation using a
from those addressed in this paper. modified ELECTRE method. Ecological Economics, 69(5),
1176– 1186.
In the airport case study used herein for exploratory Botti, L., & Peypoch, N. (2013). Multi-criteria ELECTRE
demonstration, the proposed methodology was applied to method and destination competitiveness. Tourism Manage-
ranking and selecting a set of KPIs with the preference of ment Perspectives, 6, 108– 113.
International Journal of Management Science and Engineering Management 9

Cabral, I., Grilo, A., & Cruz-Machado, V. (2012). A decision- national Journal of Research in Engineering and Technology,
making model for lean, agile, resilient and green supply 2(2), 136–142.
chain management. International Journal of Production Likert, R. (1932). Technique for measurement of attitudes.
Research, 50(17), 4830– 4845. Archives of Psychology, 140(1), 5 – 55.
EN 15341. (2007). Maintenance key performance indicators. Miller, G. (1954). The magical number seven, plus or minus two:
European Committee for Standardization (CEN). Brussels, Some limits on our capacity for processing information.
Belgium. Psychological Review, 101(2), 343– 352.
Ermatita, Z., Hartati, S., Wardoyo, R., & Harjoko, A. (2011). Muchiri, P., Pintelon, L., Gelders, L., & Martin, H. (2011).
ELECTRE methods in solving group decision support Development of maintenance function performance
system bioinformatics on gene mutation detection simu- measurement framework and indicators. International
lation. International Journal of Computer Science & Journal of Production Economics, 131(1), 295– 302.
Information Technology, 3(1), 40 – 52.
Pacaiova, H., Nagyova, A., Namesanska, J., & Grencik, J. (2013).
Figueira, J., Almeida-Dias, J., Matias, S., Roy, B., Carvalho, M.,
Systematic approach in maintenance management improve-
& Plancha, C. (2011). ELECTRE TRI-C, a multiple criteria
ment. International Journal of Strategic Engineering Asset
decision aiding sorting model applied to assisted reproduc-
tion. International Journal of Medical Informatics, 80, Management, 1(3), 228– 237.
Downloaded by [b-on: Biblioteca do conhecimento online UAlgarve] at 12:07 12 September 2014

262– 273. Roy, B. (1991). The outranking approach and the foundations of
Figueira, J., Greco, S., Roy, B., & Slowinski, R. (2013). An ELECTRE methods. Theory and Decisions, 31(1), 49 – 73.
overview of ELECTRE methods and their recent extensions. Saaty, T., & Ozdemir, M. (2003). Why the magic number seven
Journal of Multi-Criteria Decision Analysis, 20, 61 –85. plus or minus two? Mathematical and Computer Modelling,
Figueira, J., & Roy, B. (2002). Determining the weights of 38(3 – 4), 233– 244.
criteria in the ELECTRE type methods with a revised Simos’ Sevkli, M. (2010). An application of the fuzzy ELECTRE
procedure. European Journal of Operational Research, 139, method for supplier selection. International Journal of
317– 326. Production Research, 48(12), 3393– 3405.
Goncalves, C., Dias, J., & Cruz-Machado, V. (2014). Decision Shanian, A., Savadogo, O., & ELECTRE, I. (2006). decision
methodology for maintenance KPI selection – based on support model for material selection of bipolar plates for
ELECTRE I. 25 – 27 July 2014. In J. Xu, V. A. Cruz- polymer electrolyte fuel cells applications. Journal of New
Machado, B. Lev, & S. Nickel (Eds.), Proceedings of the Materials for Electrochemical Systems, 9(3), 191– 199.
Eighth International Conference on Management Science Vahdani, B., & Hadipour, H. (2011). Extension of the ELECTRE
and Engineering Management. Lisbon, Part V, Paper 87 method based on interval-valued fuzzy sets. Soft Computing,
(pp. 1001– 1014), Vol. 281 of Advances in Intelligent 15, 569– 579.
Systems and Computing.: Springer – Verlag Berlin Heidel- van Delf, A., & Nijkamp, P. (1976). A multi-objective decision
berg. doi:10.1007/978-3-642-55122-2_87. model for regional development, environmental quality
Hatami-Marbini, A., & Tavana, M. (2011). An extension of the control and industrial land use. Fifteenth European Congress
ELECTRE I method for group decision-making under a of the Regional Science Association. Papers of the Regional
fuzzy environment. Omega, 39(4), 373– 386. Science Association, 36(1), 35 – 57.
Horenbeek, A., & Pintelon, L. (2014). Development of a
Yazdani-Chamzini, A., Yakhchali, S., & Mahmoodian, M.
maintenance performance measurement framework – using
(2013). Risk ranking of tunnel construction projects by using
the analytic network process (ANP) for maintenance
performance indicator selection. Omega, 42, 33– 46. the ELECTRE technique under a fuzzy environment.
Kaya, T., & Kahraman, C. (2011). An integrated fuzzy AHP– International Journal of Management Science and Engin-
ELECTRE methodology for environmental impact assess- eering Management, 8(1), 1 – 14.
ment. Expert Systems with Applications, 38(7), 8553– 8562. Yoon, K. P., & Hwang, C. -L. (1995). Multiple attribute decision
Kumar, U., Galar, D., Parida, A., Stenstrom, C., & Berges, L. making: An introduction Sage University Paper Series on
(2013a). Maintenance performance metrics: A state-of-the- Quantitative Applications in the Social Sciences. Thousand
art review. Journal of Quality in Maintenance Engineering, Oaks, CA: Sage.
19(3), 233– 277. Zandi, A., & Roghanian, E. (2013). Extension of Fuzzy
Kumar, J., Soni, V., & Agnihotri, G. (2013b). Maintenance ELECTRE based on VIKOR method. Computers and
performance metrics for manufacturing industry. Inter- Industrial Engineering, 66(2), 258– 263.

View publication stats

You might also like