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

net/publication/350772670

Automated two-stage continuous decision support model using exploratory

factor analysis-MACBETH-SMART: an application of contractor selection in
public sector construction

Article in Journal of Ambient Intelligence and Humanized Computing · April 2021

DOI: 10.1007/s12652-021-03186-w

CITATIONS READS

7 291

5 authors, including:

Ali Raza Khoso Aminah Md Yusof

Mehran University of Engineering and Technology Universiti Teknologi Malaysia
77 PUBLICATIONS 382 CITATIONS 53 PUBLICATIONS 621 CITATIONS

SEE PROFILE SEE PROFILE

Shabir Hussain Khahro Nur IzieAdiana Abidin

Mehran University of Engineering and Technology Universiti Teknologi Malaysia
127 PUBLICATIONS 1,067 CITATIONS 42 PUBLICATIONS 267 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ali Raza Khoso on 03 June 2021.

The user has requested enhancement of the downloaded file.

Journal of Ambient Intelligence and Humanized Computing
https://doi.org/10.1007/s12652-021-03186-w

ORIGINAL RESEARCH

Automated two‑stage continuous decision support model using

exploratory factor analysis‑MACBETH‑SMART: an application
of contractor selection in public sector construction
Ali Raza Khoso1,2 · Aminah Md Yusof1 · Shabir Hussain Khahro3 · Nur Izie Adiana Binti Abidin1 ·
Nafees Ahmed Memon2

Received: 24 August 2020 / Accepted: 25 March 2021

© The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2021

Abstract
Public sector client marks contractor selection decisions on technical and financial bid considerations where efficient use of
public resources is never unheeded. A plethora of past studies has developed two-stage models; however, continuous assess-
ment of contractors is disregarded, and the models compromise on the discontinuous progression that partially recognizes
the prominence of the technical stage in the selection process. This research aims to develop a novel automated two-stage
continuous decision model for contractors’ assessment and selection where each contractor would be assessed on correspond-
ing performance assessment grading levels. Exploratory Factor Analysis (EFA) assimilated with MACBETH (Measuring
Attractiveness by a Categorical Based Evaluation Technique) employed to assess the model criteria, whereas, criteria assess-
ment stage is developed using a novel hybrid combination of SMART (Simple Multi-Attribute Rating Technique), which
in turn entails the EFA-MACBETH-SMART triplet-combination. The model encompasses extensive model criteria; thus,
76 model criteria were investigated and evaluated. Final selection of a contractor is proposed on technical bid/financial bid
ratio mechanisms based on performance levels such as ­RT/F: 80/20; 75/25; 70/30; 65/35; and 60/40. A hypothetical case is
encompassed to portray the operational mechanism of the automated assessment system. Findings from the model unveil that
continuous progression of technical assessment stage in final selection make justice with the highly qualified contractors,
and the likelihood of project success increases. The developed model further conclude that technically highest bidders may
be awarded the contract if additionally offers a feasible bid. The developed model preserves the concept of efficient use of
public resources alongside supporting the technically highest bidders.

Keywords Contractors · SMART​· MACBETH · Hybrid system · Multi-criteria decision making · Automation

1 Introduction government agencies practice their specific public pro-

curement processes to upkeep the domestic industries and
The public sector accounts for symbolic benefaction in projects (Abdelrahman et al. 2008). A public sector is gen-
the economic and social augmentation around the globe. erally considered a larger sector that undertakes mega pro-
This advancement in public works is indispensable; thus, jects. This sector trails government’s acts and legitimate

1
* Ali Raza Khoso Department of Structure and Materials, School of Civil
[email protected] Engineering, Faculty of Engineering, Universiti Teknologi
Malaysia, Johor Bahru, Malaysia
Aminah Md Yusof
2
[email protected] Department of Civil Engineering, Mehran University
of Engineering and Technology, Jamshoro, Pakistan
Shabir Hussain Khahro
3
[email protected] College of Engineering, Prince Sultan University, Riyadh,
Saudi Arabia
Nur Izie Adiana Binti Abidin
[email protected]
Nafees Ahmed Memon
[email protected]

13
Vol.:(0123456789)
 A. R. Khoso et al.

boundaries; thus, public tendering is somewhat multifarious. meta-heuristic algorithm (Ghadimi et al. 2018), neural net-
Kog (2014) proclaim that comparing to the private sector; works (Gao et al. 2019), fuzzy decision-making approach
the public sector strives the most owing to several formali- (Khodaei et al. 2018), and information gap decision theory
ties and legitimate boundaries. In general, private sector cli- (Bagal et al. 2018) etc. Several studies have developed deci-
ents are unenthusiastic and trail their own tendering process; sion support models and systems to overcome the problem
however, in the public sector, owing to public accountability, of capable contractor selection by considering the technical
the project bid price is a foremost apprehension. Accord- and financial bids. Rashvand et al. (2015) devised a model
ingly, most often in the public sector around the globe, the for selecting the contractors where the model focused a sole
award offers on the lowest bid (Awwad and Ammoury 2019; parameter of management capabilities. Likewise, Zhao et al.
Cheaitou et al. 2019). This lowest bid award is the most (2019) developed an efficiency-based system to rank and
prevailing practice in a competitive bidding system and select the contractor, but regrettably, the study deliberated
apparently accountable for efficient use of public resources. quite a few model criteria for the system on which the selec-
Nonetheless, it creates imperfect competition in the mar- tion is rather questionable. Several similar cases were found
ket (Brunjes 2020). Persisting many loopholes in the lowest where studies have focused quite a few model criteria such
bid price tendering, Brook (2017) critiques this method and as (Cheng and Li 2004; Jr. et al. 2005; Darvish et al. 2009;
propose that tendering should never be situated on the low- Watt et al. 2010; Lam and Yu 2011; El-abbasy et al. 2013;
est price alone. Awwad Ammoury (2019) also claimed that Jie et al. 2016; Birjandi et al. 2019; Marcarelli and Nappi,
no doubt the method is the most prevailing, but it does not 2019).
necessarily fallouts in favour of projects. In persistence to Padhi (2010) devised a system in a single-stage mode
this, many developed countries have already progressed to where the final award was subjected to insufficient crite-
the multi-criteria selection process. ria besides the inclusion of bid price in the same stage.
The theory of multi-criteria selection is the most preva- Similarly, a few other single-stage models were proposed
lent and has profound roots in several selection problems. by (Anagnostopoulos and Vavatsikos 2006; Vahdani et al.
Hashemi et al. (2018) advocate that the multi-criteria deci- 2013; Wang et al. 2013; Yang et al. 2016; Semaan and Salem
sion making (MCDM) approach is an appropriate technique 2017) where the bid price is deliberated during the technical
expressly for the contractor selection. A primary element assessment which is in contrast to the public sector procure-
of decision-making is the right choice of MCDM. Since ment procedures around the globe. Other decision models
the selection of a contractor is not a tranquil task; hence, were designed without considering the bid price emulating
thoughtful attention is always required (Khoso and Md sole quality based selection which is not pertinent in the pub-
Yusof 2020). In recent past, researchers has focused over lic sector, for instance, (Bendaña et al. 2008; Nieto-Morote
various MCDM techniques in contractor selection and and Ruz-Vila 2012; Taylan et al. 2017; Hashemi et al. 2018;
other relevant models such as in case of contractor selec- Tomczak and Jaśkowski 2018). Apart from this, Taylan et al.
tion; Analytical Hierarchy Process (AHP) (Abudayyeh (2017) applied the theory of big data in contractor selection
et al. 2007; Yang et al. 2012; Marcarelli and Nappi 2019; which in turn entails enormous analytical expertise to assess
Zhao et al. 2019; Gurgun and Koc 2020), Data Envelop- the contractors, and hence, the model transforms into more
ment Analysis (DEA) (Yang et al. 2016; Zhao et al. 2019), burdensome and besides encompasses complex calculations.
Analytical Network Process (ANP) (Cheng and Li 2004; Likewise, Hashemi et al. (2018) came up with a system of
Ebrahimnejad et al. 2012; Rashvand et al. 2015; Hasnain a multifaceted model with enormous calculations; also, the
et al. 2017), fuzzy set theory (Nieto-Morote and Ruz-Vila introduction of a grey number with fuzzy turned into an
2012; Afshar et al. 2017; Rao and Rathish 2018; Tomc- extra vague and uncertain environment.
zak and Jaśkowski, 2018), TOPSIS (Technique for Order In addition to the above systems and models, various
of Preference by Similarity to Ideal Solution) (Orkun other attempts have been made to devise a two-stage model
Alptekin et al. 2017), ELECTRE (ELimination Et Choice where the earlier phase assesses the technical performance
Translating Reality) (Marzouk, 2010), PROMETHEE (Pref- among the competitors and the later stage accountable for a
erence Ranking Organization METHod for Enrichment of financial assessment. However, such models are subjected
Evaluations) (Semaan and Salem 2017), Analytical Neu- to dissimilar concepts, and researchers are not agreed on a
ral Network (ANN) (Safa et al. 2015), VIKOR (Opricovic single suitable solution, for instance, San Cristóbal (2012)
and Tzeng 2004; Ebrahimnejad et al. 2012; San Cristóbal developed a two-stage system where a technical assess-
2012; Hashemi et al. 2018), and grey theory (Zavadskas ment was carried out initially, and later the final award was
et al. 2016). A few algorithms were also presented in past based on project completion time and bid price. Likewise,
models with similar approaches in different applications Liu et al. (2017) designed a two-stage system built on par-
such as hybrid robust-stochastic approach (Abedinia et al. tial least square where the final award was based on health,
2019), robust optimization approach (Saeedi et al. 2019), safety, and environment, technology, and bid price basis.

13
Automated two‑stage continuous decision support model using exploratory factor…

In contrast, Cheaitou et al. (2019) in their two-stage model Furthermore, the continuity of the technical stage up to the
selected the final contractor based on risk parameters along- final selection stage and the continuous assessment is a novel
side with bid. Marcarelli Nappi (2019) developed another research gap and desperately entails further investigation.
two-stage model constructed on AHP, wherein after the The literature is flooded with a plethora of studies in
technical qualification assessment, the final award subjects assessment and selection of contractor realm; however, none
to the least completion time along with the lowest bid. Zhao of the studies has until focused on the continuous assessment
et al. (2019) applied efficiency method to prequalify the con- model that can subsidy the technical capabilities of competi-
tractors initially, and later those contractors were allowed to tive contractors in the final selection process while consid-
offer any financial bid, and final award is subjected to the ering the standard public tendering procedure. This paper
consent of Decision Makers (DMs) which is centred on the aims to develop a two-stage model wherein during the first
bid price. In above-discussed models, the case of technical stage (technical stage) the contractors’ assessment process is
assessment is crystal clear as it encompasses computing and carried out via assessing each contractor through extensive
assigning the weights (or weightages) to assessment criteria, model criteria. The EFA is assimilated with MACBETH to
however, the concept of bid price and final award is dissimi- overcome the shortcoming of MACBETH and applied to
larly dispensed and also in contrast to the standard procedure rank the model criteria. MACBETH assists in computing
of public sector (see Sect. 2 for further details). the model criteria weight using M-MACBETH software.
Several previous attempts have been made in the estate Later, the contractors’ performance levels were measured
of decision support models, but none of the models admin- with the aid of SMART. This EFA-MACBETH-SMART
istrated a practical solution based on the current complexity triplet-combination is unique and has several fundamental
of the construction sector, especially, in recent development advantages and is being applied for the first time. In the sec-
in the public sector. Most of the models are situated on the ond stage, the technically qualified contractors are allowed to
weak foundations of model criteria. A considerable number quote a bid price wherein the submitted bids are subjected to
of studies are unaligned with the adopted public sector pro- strict assessment based on public accountability. This would
cedures around the globe. Several models are overburden preserve the idea of efficient use of public resources and
with multifarious calculations and also involved arduous at the same time avoids the non-feasible bids and supports
procedures whose application in existent circumstance is technically highest bidders. These two stages are intercon-
still doubtful. Apart from the aforementioned problems in nected and based on the novel idea of a continuous assess-
past studies, a major point for the research interest is still ment model. The entire model is based on an automated
unexplored and overlooked in the two-stage selection model assessment process developed in MS Excel spreadsheet. The
since the inception of multi-criteria selection. During the automated system computes the assessment results of con-
initial scrutiny of contractors (i.e., called technical assess- tractors and can assign the performance level accordingly.
ment stage), clients set the weightage/marks as a threshold The system identifies the qualified and dis-qualified contrac-
value to either qualify or disqualify the contractors where, tors and later computes the bid price score and Final Sum
subsequently, all qualified contractors are considered equal. Score (FSS) according to the computed assessment levels.
A contractor with the highest attained marks in the techni-
cal stage is not provided with any leverage, and this pro- 1.1 Novelty and contribution
cess terminates, and the final award is again centred on the
minimum offered bid. This contemporary process is based This research investigates the novel two-stage continuous
on a discontinuous progression that partially recognizes the decision support model for contractor’s selection. A plethora
prominence of the technical phase. All qualified contractors of past models have been developed, but the idea of con-
even which is at the threshold would stand in the same queue tinuous assessment is still unexplored. The study proposes
competing for contract award, and this does injustice with a novel combination of MACBETH with SMART (Fig. 1)
the highest-ranked contractors, which is also highlighted by owing to inherent problems in both techniques when applied
Rao Rathish (2018). Nevertheless, this part has not been individually. The novelty and contribution of this study is
widely addressed, and the decision models are overlapping briefly presented below and discussed in details in Sect. 10.
with a similar concept. Moreover, studies are not focusing on
the applications of models for the public sector on the real 1. Extensive model criteria are investigated under three
ground, and rather their focus is on complex and exceed- novel categories such as Critical Criteria, Value-Added
ingly hybrid models. Such models can have applications in Criteria, and Desirable Criteria.
academia only, however, the industrial applications are over- 2. Simple but a novel hybrid system with a triplet com-
looked in those models and their adaptability and applicabil- bination of Exploratory Factor Analysis-MACBETH-
ity are debatable. This needs further investigation in terms of SMART is introduced, which is entirely a unique con-
precise contractor assessment, especially in the public sector. cept that has fundamental advantages.

13
 A. R. Khoso et al.

Fig. 1  A novel MACBETH-SMART integration

3. The proposed model is based on two distinctive but con- related works, and Sect. 11 describes the conclusion of
tinuous stages which support the continuity of technical the study.
stage in the final selection system.
4. Concept of value for money is retained with high prior-
ity to the technical stage that serves the major purpose 2 Bid evaluation methods
of public sector client.
5. The automated assessment system is an additional con- Project cost is curious to clients, especially when dealing
tribution for easy and efficient assessment in case of a with public funds. In addition to this, the existence of a
larger pool of contractors. larger number of contractors induces higher competition.
Thus, public clients often call for tenders on the lowest bid
The remaining paper is organized as follows. Section 2 price amid colossal competition. To break this competition,
highlights the bid evaluation methods developed in the past the lowest responsive bid is typically the possible solution
for financial assessment of contractors. Section 3 presents among the public client. However, several other attempts
the insight on MACBETH method and its advantages and were made in literature in the last few decades to find out a
shortcomings; similarly, Sect. 4 covers the background of more appropriate way of dealing with price criteria to main-
SMART. Research Method is explained in Sect. 5, which tain the quality outcome. In this quest, following several
also highlights the development of novel MACBETH- models and indices have been worked out to evaluate the
SMART model. Section 6 encompasses data collection bid price in recent past.
and preliminary analysis, such as primary tests of EFA. Topcu (2004) developed an extensive model and proposed
Analysis and Results are covered in three sub-sections a system of dealing with the bid price, according to the
under Sect. 7, and model development stages are explained study, threshold bid values (upper and lower) can be deter-
in Sect. 8. A hypothetical case is tested for implementa- mined using sum and difference of average bid and by con-
tion of the model is presented in Sect. 9, whereas, Sect. 10 sidering the standard deviation where all bids beyond those
highlights the comparison of the proposed model with past values were discarded. Further, the bid price scores were

13
Automated two‑stage continuous decision support model using exploratory factor…

computed using linear normalization (i.e., lowest bid price/ applied by San Cristóbal (2012) where cost parameters were
price under consideration), the most commonly employed optimized with other resources using TOPSIS and VIKOR
formula for bid evaluation as claimed by (Ballesteros-Pérez method.
et al. 2013). A few studies proposed that when the selection El-Abbasy et al. (2013) developed a simulation-based
is based on a bid price basis, a contractor who submits the model wherein the Monte Carlo simulation method was
following bid must be selected i.e. average bid price (Rocha devised to compute optimum index amongst qualification
de Gouveia 2002), below the average bid price (Ioannou and criteria and iteratively bid price weightage. The proposed
Awwad 2010) or using a truncated method (below average iterative process was based on a large sum of historical data
bid) after excluding outliers (Waara and Bröchner 2006). which may not results in reasonable decisions and can be
In contrast, (Albano et al. 2006) found that the average bid more problematic to investigate. A similar iterative kind of
method has numerous drawbacks. Besides, Watt et al. (2010) model was proposed and designed by Safa et al. (2016) using
recommended a straightforward way of dealing with the bid Pareto front optimization where the decision model was
price where any bid 10% below and 10% above the aver- trained using several constraints and objectives including;
age bid is deemed as minimum acceptable and the best bid cost, time and other evaluation criteria. However, the model
respectively. Regrettably, no rationalization is provided for yields a more significant number of solutions, and the final
their proposed method. selection was subject to human judgments that could create
Teixeira De Almeida (2007) applied the idea of numeri- more shakiness in the justified decision. Awwad Ammoury
cal modelling in a bid price solution via a hybrid combi- (2019) employed agent-based modelling to determine the
nation of PROMETHEE and MAUT (Multi-Attribute Util- best bid amongst the second-lowest bid, average bid, below
ity Theory) along with ELECTRE to compute probability average bid, above-average bid, and truncated bid (closet and
function. The model figured out the best alternative among below the average). The simulation process found that the
six different bids by computing the criteria weightages. The second-lowest bid price was in favour of the client. In this
proposed model was somewhat equivocal and encompassed approach, the concept of the efficient use of public resources
an extreme hybrid combination of different techniques that was not considered. Similarly, a competitive bidding model
mark this rather problematic to apply in a real scenario. is devised by Semaan Salem (2017) founded on optimized
Moreover, appropriate directions to apply the case on dif- bid solutions keeping in view time, cost, safety, and qual-
ferent studies were also not addressed. A similar complex ity as selection criteria. The submitted bids were treated in
model was also developed by Marzouk (2008) based on percentage differences from the maximum submitted bid and
superiority and inferiority ranking via utilizing SAW (Sim- minimization of bid and time and maximization of safety
ple Additive Weighting), TOPSIS, AHP, and MAUT for and quality were kept under consideration.
computing the contractor’s final ranking. The cost param- Liu et al. (2017) designed a two-stage contractor selec-
eter was handled using the superiority and inferiority model tion model wherein during the bid evaluation phase, the
based on SAW and TOPSIS. Later on, another study is highest and lowest bids were disregarded in the beginning.
presented by Marzouk (2010) but this time author utilized Final contractor selection was based on bid price alongside
ELECTRE III for contractor selection case. The weightage technology and health, safety, and environment parameter
of the bid was estimated as 25%, and the values to each correlations. The bid price was treated on a benchmark of
bid were estimated using credibility score via ELECTRE standard bid value obtained through a formula, i.e. (consid-
III. Both studies embroil exhilarating calculation and entail ered bid-mean value bid)/mean value bid*100). The highest
efforts in dealing with the case of bid price criteria. In con- value goes to any bid closer to this estimate. The problem
trast, Padhi (2010) worked out a system based on the opti- with approach can be a) no justification of removing the
mization of the bid price, resources, time of project comple- lowest and highest, and b) the mean value itself can be too
tion, and maintenance period. The system auto-generates the high or too low and would be on the mercy of other bids.
ranking of contractors based on the mentioned criteria. The Similarly, Lai et al. (2004) introduced a bid evaluation index
system, unfortunately, did not devise a separate mechanism that calculates a benchmark value, and any value closer to
of dealing with bid evaluation. The bid price is considered as that benchmark would be provided higher weightage with a
an inverse function of resources; thus, the system would take maximum of 90 marks. The benchmark value can be calcu-
the lowest price as an optimized one. A similar approach lated as shown in Eq. 1

bid price submitted by bidders − baseline bid given by owner (1)

Say, valid bid =
base bid

13
 A. R. Khoso et al.

Benchmark bid = 0.4 ∗ baseline bid + 0.6 ∗ average of valid bids studies has devised a mechanism to benefit the technically
high ranked contractors in the final award.
The specified approach was quite systematic and beyond
the abnormally bid range in terms of baseline bid specified
by the owner; however, the problem occurs in the ceiling 3 MACBETH
price. The formula is valid for any value extremely higher
than the ceiling price that is not acceptable to public depart- C.A. Bana e Costa and J.-C. Vansnick developed MAC-
ments. Furthermore, the use of the average bid is again BETH and later modernized and restructured it in 2004.
questionable to some extent, as this can increase the project This method is primarily based on linear programming,
price. Likewise, Ballesteros-Pérez et al. (2015) found that wherein each element of a set assigned an absolute value
the problem of computing the bid price weightage in the past say A (Bana 1994). The MACBETH method operates on a
can handle with the following equations (Eq. 2, 3). qualitative judgment, unlike the classical approaches in AHP
( ) and ANP and their families. The final judgment is decided
submitted bid based on the formulation of an additive value model which
bid price score = 1 −
ceiling price (estimation by client) prioritizes the alternatives with the aid of criteria weights
(2) (further details in the form of preliminaries of MACBETH
max. bid − submitted bid method is described in Appendix A). The method produces
bid score = (3)
max. bid − min. bid impartial and constructive outcomes by considering the
fuzziness of the judgment through the seven-point semantic
Equation 2 can be used when the ceiling price is known,
scale of judgment. The semantic scale inherently based on
and the applicability of Eq. 3 is under non-availability of
the fuzziness in responses which is a common occurrence in
information. However, in both cases, the true representa-
any decision process. Owing to its constructive, and interac-
tion of the efficient use of public resources is not reflected.
tive outcome with a property of fuzziness, the qualitative
It is because a contractor can quote any bid lower than ceil-
judgments from the path of ordinal data transpired into car-
ing price in case of Eq. 3, and there are no upper or lower
dinal preferences.
limits in Eq. 2, hence, would not be resulted in the feasible
MACBETH offers several fundamental edges over the
solutions.
classical MCDM methods such as AHP. Its non-numerical
Looking at the past bid evaluation studies, none of the
pairwise scale converts the method into a simple process
aforementioned studies has properly resolved the problem
(Bana e Costa and Chagas 2004; Cox et al. 2013). The addi-
of bid evaluation through a rational and simple mechanism.
tional perks are catered by its qualitative scale that offers
The researches worked out the various disparate solutions of
bounteous opportunity to resolve the judgments, and assists
the bid price, which in many cases challenging to compute
in eluding the forceful decisions from DMs (Ertay et al.
and implement in the public sector. Moreover, few predictive
2013), and further conveys precise information (Joerin et al.
models required historical data on submitted bids which is
2010). The method operates on the principle of transpir-
challenging to collect, and such models entail the bid data to
ing the qualitative judgment to quantitative judgment that
workout optimum bid with respect to other selection crite-
is rather smooth and easier to understand by DMs while
ria. Such mechanisms are hardly workable in public tender-
responding judgements. Unlike AHP, the responses are
ing where qualification assessment takes place at an early
assembled on a qualitative seven-point semantic scale of
stage before bidding. In the aforementioned methods, the
differences, as shown in Table 1. Moreover, the Saaty scale
weight to bid price and other evaluation criteria are not fully
in AHP does not offer the fuzziness, whereas, this trait
addressed and not covered until now. Moreover, none of the
is inherently added in the semantic scale that also offers

Table 1  Semantic scale of Semantic scale of differences Quantitative Qualitative equivalent

differences scale

Null 0 No difference
Very weak 1 Very weak difference of attractiveness over another
Weak 2 Weak difference of attractiveness over another
Moderate 3 Moderate difference of attractiveness over another
Strong 4 Strong difference of attractiveness over another
Very strong 5 Very strong difference of attractiveness over another
Extremely strong 6 Extremely strong difference of attractiveness over another

13
Automated two‑stage continuous decision support model using exploratory factor…

additional flexibility in the form of intermediary judgment preliminaries of SMART is presented in Appendix B).
opportunities to the DMs. Besides the prime perks of scale SMART offers a straightforward process of computing the
and flexibility in MACBETH, a major apprehension in the grades using a simple formula. Although various types
classical MCDM method is the consistency of judgments of utility functions are available, i.e. linear, non-linear,
which is often problematic to the researcher. But thanks to and exponential, but the linear utility function is recom-
its auto-consistency check and instantaneous validity via a mended in case of independent group judgments (Rayno
built-in function in M-MACBETH software that eradicates et al. 1998). There are several shreds of evidence that the
the encumbrance of inconsistency. linear function is relatively healthy and more comfortable
The present case of contractor assessment is established to be interpreted and elicited and also a close approxima-
on extensive model criteria. The accurate and appropriate tion (Gómez-Limón and Martínez 2006; André and Riesgo
assessment of a large number of criteria is often a problem 2007). The linear function has got another advantage of
in many MCDM techniques such as in PROMETHEE, AHP, operating without group DMs and also computes simi-
and ANP, etc. The MACBETH has leverage in such circum- lar results as of non-linear and exponential (Konidari and
stances and still delivers fair and precise results irrespec- Mavrakis 2007).
tive of a larger number of model elements (Madeira et al. Marler Arora (2010) asserted that to know the prefer-
2012). Apart from the several aforementioned dominances ences of DMs, SMART is the most superior technique
of MACBETH, it is also affected by a few shortcomings that in decision making and also comfortable in the applica-
are still not unveiled on a larger scale. Unlike AHP and ANP, tion. Moreover, they believe that the cognitive complexity
the weightage computation in MACBETH is a function of level in SMART is much lower even from the simple AHP.
criteria order which is on the benevolence of DMs. C. A. Brugha (2004) and Konidari Mavrakis (2007) suggested
Bana et al. (1994) affirm that in MACBETH, the ranking that SMART is a comprehensive tool for quantitative
of attributes is conceivable with the support of DMs either evaluation and entails less computational efforts compar-
in the shape of pairwise judgmental information (swing ing to AHP. Chou Chang (2008) linked the popularity of
weights) or via direct consultations with DMs. The swing the method with its wide range incorporating quantitative
weight and direct rating approaches in the past have been and qualitative criteria. The additional perks of SMART
under tremendous criticism. Owing to several inherent draw- include its powerful assessment method comparing to
backs, researchers believe that these methods have no scien- AHP (Brugha 2004). The additive value model in SMART
tific approach. While comparing the swing weight method, has numerous advantages such as it represents the true
various studies criticized the method and claimed that this aggregate utility function even in a case if the additive
method is rather complex, challenging to apply, less intui- utility independence does not hold precisely (Duarte and
tive and has a higher chance of errors (Monat 2009), intri- Reis 2006). It reduces complexity in the process (Kwak
cate in the application (Dabrowski 2014; Barfod and Salling et al. 2001), and provides more robust outcomes during
2015), and subjected to variations in outcomes (Winterfeldt sensitivity analysis as compared to other functions (Kumar
and Edwards 1986; Edwards and Barron 1994). In contrast, and Alappat 2005).
(Bana e Costa and Chagas 2004; Konidari and Mavrakis Apart from the several advantages of SMART, the basis
2007) criticized the direct rating method, and asserted that of computing weightages in SMART is highly criticized in
the method is less precise compared to other methods in the recent past. To come up with this problem, Konidari
similar family. Mavrakis (2007) utilized the AHP-SMART hybrid option
where criteria weightage were calculated using AHP, and
SMART was employed to assign performance grades.
4 SMART​ The study found that no doubt the SMART technique is
extensive and involves lesser efforts but the process of
SMART is fundamentally derived from MAUT and con- weightage determination in SMART is not irrational and
sidered as its simple version (Brugha 2004). The SMART not acceptable in case of complicated problems. Since
aims to rank the alternatives in a subjective order and the direct weightage method is involved in SMART and
offers ratings in performances using an appropriate numer- depends on the direct judgment of DMs, this creates a
ical grading. Besides, SMART computes the performance problem as their judgments are more subjective (Konidari
of any function in the form of distinctive grading levels. It and Mavrakis 2007). Thus, the weightage assessment
is based on a linear additive model likewise MACBETH, process in SMART is shaky and less confident. Owing
wherein swing weights or direct weighting systems are to discussed fundamental challenges and shortcomings in
applied which has fundamental drawbacks as claimed by SMART, the weightage computation process is preferred
(Bana et al. 2004). The process of grading assessment in from MACBETH analysis, whereas, the SMART would
SMART is based on its utility function (Furthermore, the

13
 A. R. Khoso et al.

compute the performance grading levels. This MAC- sensitivity analysis. Five different levels of performance
BETH-SMART integration is explained in Sect. 5. grades were designed based on the technical weightage
of each contractor. The assessment of each contractor
was decided on the utility values computed using the
5 Research method SMART technique, which in turn form a novel combina-
tion of MACBETH-SMART. In the recent past, several
Scientific research trails a systematic and structured hybrid combinations of MCDM techniques were employed
method to achieve a research goal. The choice of con- because the single technique is incapable of resolving the
tractor in construction, especially in the public sector is intricate challenges in a few cases. Recently, SMART and
stimulating and arguable that entails a more sophisticated MACBETH techniques have been integrated with other
research method. Holt (2010) found that the problem of methods to get optimum results. For instance, MAC-
contractor selection in the public sector is worsening day BETH-fuzzy AHP (Ertay et al. 2013) in case of renew-
to day and no distinctive actions have been taken to resolve able energy, MACBETH-MAUT (Hurson et al. 2012) for
the problem, rather the studies are getting intricate and no portfolio selection, MACBETH- COPRAS (Kundakcı and
appropriate system developed for the public sector. For a Işık 2016) for air compressor selection, MACBETH-EDAS
long time, a massive number of studies have emerged in (Kundakcı 2019) in the application of small and medium-
resolving the problems of the private and public sectors; sized enterprises (SME). Further, Konidari Mavrakis
nevertheless, the quest for well-organized and systematic (2007) utilized the AHP-SMART combination, but none of
research in the domain of the public sector is still endur- the studies has ever employed MACBETH-SMART inte-
ing. Also, Khoso Md Yusof (2020) confirmed that the gration as a hybrid technique (Fig. 1). The MACBETH can
topic of contractor selection had been raised for the last serve as an alternative in SMART to evaluate value func-
three decades; nonetheless, there are still more avenues of tion. The integration of MACBETH-SMART is simple
research in this field. and straightforward as both techniques have compatibility
The present investigation focused on extensive model because of their same origin from MAUT. Moreover, the
criteria, the building blocks of a model. To come up SMART uses criteria weightage to compute the overall
with more valuable and extensive model criteria, vari- utility value, and MACBETH can efficiently compute the
ous prominent databases were explored. Published litera- weightage.
ture followed by interviews with experts laid exhaustive In the first stage of the model, each contractor, after
discussion. A novel and an extensive set of criteria were qualifying the screening would be technically assessed on
listed out considering the complexity in today’s public most critical model criteria identified from EFA analysis.
sector projects. Appropriate classification of model cri- This stage results in five distinct levels of contractors i.e.
teria alongside the suitability of criteria as per the public not acceptable (L0), hardly acceptable (L1), acceptable
project need was a top priority. With the experts’ con- (L2), highly acceptable (L3), and outstanding (L4). The
sultation, 76 model criteria structured into three primary second stage of the model scrutinized the submitted bid
classifications namely; the Critical Criteria (CC), Value- from (L1–L4) groups of contractors. The bids that are not
Added Criteria (VAC), and Desirable Criteria (DC) were meeting the purpose of efficient use of public resources
investigated and evaluated. The data on the level of sig- (i.e. either too low or higher than government’s estimation)
nificance of model criteria were gathered with the aid of a would be called non-feasible bids, so, discarded. The final
questionnaire. SPSS software tool applied to analyze the selection of a contractor is a continuous model i.e., techni-
significance of model criteria using EFA in the form of cal stage would not be obsolete and the benefits of higher
Principal Component Analysis (PCA) where the rotation technical weightages would be provided to contractors in
of factors produces the criteria in terms of their signifi- the final stage of the financial assessment. This continuous
cance. At this stage, EFA substitutes the first condition model would assign the weightage to technical and finan-
of MACBETH, i.e. ratings of criteria owing to discussed cial bids exclusively based on their respective performance
inherent problems in MACBETH (see Sect. 3). levels. This continuous assessment approach indicates that
The ranking of model criteria leads to the design of the no two groups (L1–L4) would be treated uniformly, and
second questionnaire based on the pairwise semantic scale the higher compensation in bid price is provided to the
in MACBETH. Top hierarchy experts from the public sec- one who ranked highest in technical assessment tier. This
tor called for their judgment input on a semantic scale of computation is based on an automatic system that initially
differences (as per Table 1). The experts’ judgments were computes the technical weightages from provided informa-
analyzed in a registered M-MACBETH software (pur- tion through strict scrutiny by a team of DMs. Later, based
chased online from http://m-​m acbe​t h.​c om), and model on assigned performance grading levels, the FSS would be
criteria were weighted, modified, and verified through calculated that can decide the contract award. This novel

13
Automated two‑stage continuous decision support model using exploratory factor…

Fig. 2  Functional research flow

based on an employed hybrid
system

13
 A. R. Khoso et al.

research design is organized with the aim of achieving fair correlation among the variables using the Chi square test.
and rational results. Figure 2 demonstrates the functional A significant correlation was found for all three categories
research methodology based on hybrid system carried out of criteria (i.e., 0.000 < 0.05). Later, the KMO analysis was
in this work. conducted to confirm the data adequacy for EFA. The KMO
value for three criteria categories found as 0.710, 0.809, and
0.809 for CC, VAC, and DC categories, respectively. A value
6 Data collection and preliminary analysis of 0.6 is set as a benchmark for KMO analysis as per the
suggestions of (Jeremy et al. 2006).
Data collection is a fundamental and essential part of sci-
entific research that represents a process of gathering infor-
mation to answer the research questions, whereas, the data 7 Analysis and results
analysis is a process of transforming the data into useful
information to support decision making. The most impera- 7.1 Model criteria
tive part of this process is to make sure that rich and reliable
information was collected. King et al. (1994) asserted that Criteria ranking is a prime step in analyzing the criteria
scientific research must follow codified, explicitly, and popu- weightage in MACBETH. PCA analysis was computed in
lar methods to collect and analyze the data. For this present SPSS where factor rotation (FR) produces the factor loading
research, two exhaustive questionnaires were designed to (FL) of each sub-criteria. The process of FR does not alter
collect the data sample. A questionnaire survey is a common the solutions rather a fair and simple structure of variables
and appropriate tool to gather data from the respondents emerged as an outcome in the form of FL. The greater the
for an empirical study (Wang et al. 2019). A pilot survey value of FL, the higher the significance of variables and a
was conducted in the form of a pre-expert survey with a value of 0.5 is suggested as a cutoff for measuring this sig-
few experts. This is conducted to verify the viability of nificance (Phogat and Gupta 2019). Since several varieties
the study before the actual data collection process on large of FR methods are available and its correct choice subjected
sample size. The pilot survey facilitated in final instrument to the variable correlation; however, for this case, a varimax
design and later, experts’ survey was conducted to collect method is employed which is more systematic and has a
the data on a larger sample size. To target the larger sample, tendency to produce fair results (Phogat and Gupta 2019).
an expert sampling technique of purposive sampling method Each category of criteria was subjected to PCA indepen-
was adopted. This is a non-probabilistic sampling approach dently and later rotated to produce the significance variables
that is based on the population characteristics and targets (model criteria). With this analysis, 73 model criteria (i.e.,
the objectives. This sampling method is generally conducted sub-criteria) were identified as most influential out of 76
from renowned personnel of relevant fields. Besides, data whose FL values were greater than the minimum cutoff (i.e.,
sample for this work was collected from highly qualified 0.5) see Figs. 3,4,5.
practitioners from the client, consultants, contractors, and Figure 3 displays the FL results of the CC criteria cat-
other organizations within Pakistan having rich experience egory, where 32 sub-criteria were analyzed, and 29 were
and expertise in public tendering works. The second ques- found as the most significant. These sub-criteria were dis-
tionnaire is based on one–one interaction with highly expe- tributed into eight major criteria, as shown in Fig. 3. Simi-
rienced personnel, called here DMs. In total, 15 DMs were larly, Fig. 4, 5 demonstrate the FL results of VAC and DC
targeted to acquire their judgements. categories, respectively. In these categories, none of the
To validate the quality and quantity of data, various sub-criteria was omitted. Besides, the FL, another analysis
screening tests were conducted. In case of EFA analysis, in the form of Factor Score (FS) was performed to rank the
the sample size was confirmed from (Kline 1994; Bryman, major model criteria. The FS computation is quite simple
A. and Cramer, 1997), according to them, 100 sample size and therefore attracted many researchers in recent past such
is sufficient for conducting EFA. Whereas, the quality of as (Madeira et al. 2012; Benson et al. 2016; Jiang and Zhang
data was confirmed by measuring the internal consistency 2016). Distefano et al. (2009) defined the FS as dividing
of data. The analysis unveils that the Cronbach’s alpha value the highly loaded subsets from the addition of FL in each
is 0.872, 0.904, and 0.902 for CC, VAC, and DC categories, group. In other words, FS is an average of FL for a particular
respectively. The Cronbach’s alpha values are greater than group. The ranking of criteria and sub-criteria are compiled
the minimum cutoff of 0.7, as suggested by (Phogat and and demonstrated in descending order, as shown in Fig. 6.
Gupta 2019). Further, the data sample authentication for
EFA was examined using two different and the most popular
data analysis methods, i.e., Bartlett’s Sphericity Test (BST)
and Kaiser-Mayer-Olkin (KMO) analysis. BST examines the

13
Automated two‑stage continuous decision support model using exploratory factor…

Fig. 3  Factor loading plotting for sub-criteria of CC category

Fig. 4  Factor loading plotting for sub-criteria of VAC category

Fig. 5  Factor loading plotting for sub-criteria of DC category

7.2 Criteria weightage M-MACBETH. According to C. Bana e Costa et al.

(2003), the weightage computation stage converts the ordi-
Weightage computation process in MACBETH requires nal data from 1st Condition into cardinal data using the
data on a pairwise semantic scale of differences. The 2nd Condition of MACBETH’s linear programs. The dif-
essential data from DMs on the second question- ferences of attractiveness on each criterion and sub-criteria
naire was gathered and later analyzed in software, i.e., were recorded in SPSS and analyzed using median values.

13
 A. R. Khoso et al.

Fig. 6  Classification and ranking of contractor assessment criteria

The completed judgment matrix was formulated after criteria are assigned a relative weightage of 100, and the
compiling the results of individual DM, as suggested by remaining criteria are weighted according to the judg-
(Mateus et al. 2017). MACBETH offers excellent collec- ments given by DMs. In addition to the scale weight,
tions of analyzing the criteria weightages among those a M-MACBETH also computes a self-normalization matrix
top-down hierarchical method was embraced for analysis. where the criteria weightage are auto-normalized, and the
Consistency validity is appalling in MCDM methods; normalized weightages can be easily computed. The nor-
however, with the aid of M-MACBETH, the issue does not malized weightage of criteria is a function of scale weight
persist. Thanks to the software’s real-time consistency test and can be modified accordingly. Since the scale weight
and self-adjusting option. When the value-judgment matrix is independent and flexible, therefore, if it requires can be
or any of its judgment is inconsistent, the software auto- modified by the users. The normalized weightage changes
warns the illogical judgments and the matrix can no longer according to the variations performed in the scale weight.
be analyzed until the consistency problem is resolved. The This process is carried out whenever the weightage of any
powerful M-MACBETH auto-suggests different likely pat- specific criteria is either too large so that the weightage of
terns to modify the judgments, and after approving, the remaining criteria reduces illogically or exceedingly too
matrix can be validated and ready for further analysis. Fig- small to evaluate. This transpires, while making judgments
ure 7 illustrates the judgment insertion process and auto- from DMs, henceforth, the true essence of judgments on
inconsistency judgment detection, and Fig. 8 clears how the any criteria does not reflect on such cases. Therefore, scale
inconsistencies are auto-adjusted. weights need modifications that otherwise create problems
Once the auto-consistency validation is performed, during accurate assessments (Bana E Costa et al. 2008).
M-MACBETH produces a linear scale showing the criteria This inconsistent variation is a result of fluxes in judg-
weightages on a 0–100 point scale where the top-ranked ments from DMs. In such cases, M-MACBETH offers a

13
Automated two‑stage continuous decision support model using exploratory factor…

Fig. 7  Auto detection of inconsistency

powerful and self-adjusting method where users can easily The attained weightage of major criteria and sub-criteria
modify the scale weight and at the same time, check the are later distributed according to their hierarchical distribu-
criteria weightage. This process of adjusting and modify- tion such as from categories to major criteria and then to
ing the weightages by observing the effect of one criterion sub-criteria. The distribution of weightage is supported by
on others is called sensitivity analysis. The advantage of the aid of the distribution factor (i.e. weightage of parent
this process is in the form of unfaltering the judgment node). Distribution factors of each parent node weightage
value matrix. were initially calculated and applied to their children nodes.
The process of modifying the scale weight with unfal- The weightage distribution from parent nodes to children
tering the judgment value matrix is only possible to a cer- nodes is illustrated in the weightage computation model in
tain extent, and the judgment value matrix would be unal- Fig. 11a–c.
tered within the prescribed limits. Thus, M-MACBETH
auto offers upper and lower adjusting levels within which 7.3 Criteria assessment
the modifications are tolerable, see Figs. 9, 10. The auto-
validation approach curtails the human efforts of validating The concept of the criteria assessment was accomplished
the results from each DM. This process of computing the by employing the SMART technique. The grading assess-
weightages was applied in the form of top-down hierarchi- ment was computed using Eq.15 (Appendix B). In the first
cal order, i.e. beginning from the categories of criteria, later stage, the grading assessment in the case of each sub-criteria
the major criteria in each category individually and their are generated in the form of rubrics, and later, the gener-
sub-criteria. Each time, the judgment value is inserted, auto- ated levels are assessed using the basic linear concept of
checked for consistency, auto-suggested, and validated. This SMART. Distinct grading levels are set for different sub-cri-
leads to initial scale weight and later, the modified criteria terion depending upon their nature. The maximum grading
weightage. The computed initial and modified weight of levels are five ranges from 4 to 0 in their decreasing worth.
each attribute (categories, major criteria, and sub-criteria) Furthermore, the minimum designed grading are kept up
are demonstrated in Table 2. to two levels depending upon the nature of the information

13
 A. R. Khoso et al.

Fig. 8  Auto-suggestions on modifying the judgement in M-MACBETH inconsistence judgements

contained by the sub-criteria. Table 3 exemplifies the values each contractor would be assigned accordingly as described
in case of different grading levels. in Table 4.
In the present model, there are four distinctive grading According to Table 4 five distinctive performance levels
levels assigned to sub-criteria, i.e., 4–0, 3–0, 2–0, and 1–0 as can be assigned depending upon the acquired ­TBS value.
mentioned in Table 3. The zero is assumed by default mini- For achieving any level (L), a threshold was set, i.e., mini-
mum value, either the value is assigned or not to any attrib- mum ­TBS must be 70 (out of 100). The contractor with the
ute, therefore, ∆min is the minimum scale value indicating lowest performance level, i.e., the poor performer would be
the zero. Furthermore, ∆αβ, in this case, exemplifies the disregarded from further competition. Except for the poor
“considered grading level” that is assigned by the evaluation performer, all remaining contractors would be allowed to
team to each contractor depending upon their performance participate in the final stage of contractor selection, i.e., the
levels. Once the grading assessment is devised for each sub- financial assessment stage.
criterion, the next level is to compute their weightages using
Eq.15 (Appendix B). The entire process of calculating the
weightages is mentioned in the supplementary data file. The 8 Model development stages
purpose of grading assessment weightage is to distribute the
parent node weightages to the achieved performance levels. 8.1 Screening process
At this stage, SMART is assimilated with the MACBETH
technique. Through the SMART assessment levels, each An exhaustive screening process is followed to verify the
sub-criteria weightage is distributed and the DMs would eligibility of competitors before entering the competition.
assign the achieved levels in each sub-criteria, and corre- In this regard, elementary information is gathered from each
sponding weightages would be calculated. Depending upon contractor in the form of eligibility criteria. These criteria
the total technical bid score ­(TBS), the performance levels of vary according to government policies and regulations, few

13
Automated two‑stage continuous decision support model using exploratory factor…

Fig. 9  Initial scale weight and

criteria weight

examples of such criteria are; tax return proofs, a license of 8.3 Second stage‑financial assessment stage
work, registration with professional bodies, proof of non-
blacklisting, etc. Once the eligibility of the contractor meets, Financial assessment in public tendering has a predominant
the further process of performance assessment instigates. role which is accountable for bid assessment and evalua-
The successful candidates offered to apply for the technical tion. The successful bidders who qualified the technical
assessment stage, and those who could not meet any single stage would compete further for the final stage. At this stage,
eligibility criterion would be out of the competition; there- all the contractors from the performance level category of
fore this is served as knock out stage. L1–L4 would be entertained, and level L0 would be rejected
for further assessment.
8.2 First stage‑technical assessment stage In the bid assessment stage, the DMs are responsible for
evaluating bid price according to the following classifica-
All the eligible contractors can contest for the technical tion, i.e., Type A bid -feasible bid; Type B bid-abnormally
phase of tendering. At this stage, extensive technical crite- lowest bid, and Type C bid; above ceiling price bid. Type A
ria are required to meet in order to qualify for the bid stage. bid is subjected to further analysis, and the final decision is
These technical criteria are divided into three distinguished centred on the multi-criteria decision, i.e., the combination
categories, i.e., CC, VAC, and DC. Each distinguished set of technical and financial bid score. Type B bid is consist-
of the category was further classified into major criteria and ing of all the marginally low bids, i.e., sufficiently below
finally the sub-criteria. The assessment process of each con- the engineer’s estimation and Type C bid comprises of all
tractor is carried out via the attained grading assessment overestimation bids as these are not accountable for limited
values, as presented in Table 4. public resources and would be disregarded from the com-
petitive process. The formula for calculating the financial
bid score is designed in Eq. 4.

13
 A. R. Khoso et al.

Fig. 10  Modified scale weight

and criteria weight

Ap ∗ 100 final selection. Additionally, to anticipate the concept of

FBS = (4) continuous assessment and providing the benefits of higher
As
technical weightage into bid price in the form of compensa-
where; tion, five distinct continuous ratio mechanisms (technical
Ap = Amount of the lowest submitted bid bid/financial bid, R
­ T/F) are suggested as presented in Table 5.
As = Amount of a bid under consideration According to the final contractor’s selection mechanism
According to Eq. 4, the contractors would get lesser mentioned in Table 5, the benefits of technical parameters
weightage if their bid is above the lowest bid. This would are reducing with dropping in the performance level. This
serve two major purposes; firstly, there are no restrictions in suggests that higher technical weightages must be assigned
quoting any amount of bid as long as the bid is feasible, and to a contractor(s) that has/have higher chances of producing
secondly, the lowest bid contractor is not always the winner, quality performance. Moreover, owing to higher chances of
a common issue in the construction sector for decades. quality performance, the contractor(s) must be given a cer-
tain exemption in quoting the least bid price as a discount
8.4 Final selection Computing‑FSS (but must be feasible bidder). Similarly, the contractor(s)
with less performance grading level must quote a bid price
The final selection of a contractor is based on technical and closer to the least bid. These contractors would have a like-
financial assessment combined weightages. Since during the lihood of gaining lesser profits as their bid would have the
financial assessment all the contractors who have submitted lesser quoted price comparing to higher performance grad-
the overestimated bids are disqualified, so the client is not ing contractors (e.g. L4). The formula of computing the FSS
anxious about the efficient use of public resources as none is followed by Eq. 5.
of the contractors is receiving any benefit of higher profit
from public funds. This would augment the satisfaction level FSS = TBS *TBW + FBS *FBW (5)
of clients as the right value of public funds can be utilized.
where;
The final selection of a contractor is proposed distinctly,
FSS = Final sum score
i.e., there must be continuity of the technical phase in the

13
Automated two‑stage continuous decision support model using exploratory factor…

Table 2  Initial and modified Attribute Initial Scale Initial Criteria Modified Scale Modified
weight of attributes Weight Weight Weight Criteria
Weight

Categories of Criteria CC 100 58.33 100 50

VAC 57.14 33.33 60 30
DC 14.29 8.34 40 20
Major Criteria (CC) MPC 100 23.36 100 20
QAS 86.83 20.33 89 18
FPA 74.43 17.29 77 15
FE 60 14.02 71 14
CFS 46.67 10.98 59 12
FTR 33.33 7.71 50 10
PE 20 4.67 30 6
MP 6.67 1.67 24 5
Major Criteria (VAC) RM 100 30.30 100 25
SC 75 22.73 86 22
SWC 65 19.70 79 20
VECP 55 16.67 71 18
TP 30 9.09 38 10
AT 5 1.52 20 5
Major Criteria (DC) RS 100 36.67 100 30
OS 72.73 26.67 84 25
CS 54.55 20 68 20
EQ 36.36 13.33 52 15
ITC 9.09 3.33 34 10
Sub-Criteria (MPC) MPC2 100 36 100 30
MPC1 77.78 28 82 25
MPC3 55.56 20 64 20
MPC5 33.33 12 49 15
MPC4 11.11 4 33 10
Sub-Criteria (QAS) QAS2 100 46.16 100 33
QAS4 66.67 30.77 80 28
QAS1 33.33 15.38 60 22
QAS3 16.67 7.69 50 17
Sub-Criteria (FPA) FPA5 100 55.56 100 47
FPA2 60 33.33 71 33
FPA3 20 11.11 43 20
Sub-Criteria (FE) FE2 100 43.75 100 35
FE1 71.43 31.25 83 29
FE3 42.86 18.75 63 22
FE5 14.29 6.25 40 14
Sub-Criteria (CFS) CFS4 100 43.75 100 34
CFS3 71.43 31.25 85 29
CFS1 42.86 18.75 73 25
CFS2 14.29 6.25 35 12
Sub-Criteria (FTR) FTR3 100 55.56 100 60
FTR1 60 33.33 50 30
FTR2 20 11.11 17 10
Sub-Criteria (PE) PE1 100 100 100 100

13
 A. R. Khoso et al.

Table 2  (continued) Attribute Initial Scale Initial Criteria Modified Scale Modified
Weight Weight Weight Criteria
Weight

Sub-Criteria (MP) MP2 100 36 100 30

MP5 77.78 28 93 28
MP3 55.56 20 67 20
MP4 33.33 12 40 12
MP1 11.11 4 33 10
Sub-Criteria (RM) RM2 100 43.75 100 40
RM3 71.43 31.25 71 28
RM4 42.86 18.75 50 20
RM1 14.29 6.25 31 12
Sub-Criteria (SC) SC3 100 46.67 100 41
SC2 50 23.33 54 22
SC4 35.71 16.67 45 18
SC5 21.43 10 35 14
SC1 7.14 3.33 11 5
Sub-Criteria (SWC) SWC4 100 45 100 40
SWC3 66.67 30 75 30
SWC1 44.44 20 51 20
SWC2 11.11 5 25 10
Sub-Criteria (VECP) VECP2 100 36 100 33
VECP5 77.78 28 84 28
VECP1 55.56 20 67 22
VECP3 33.33 12 33 11
VECP4 11.11 4 18 6
Sub-Criteria (TP) TP2 100 36 100 100
TP1 77.78 28 84 84
TP3 55.56 20 67 67
TP5 33.33 12 50 50
TP4 11.11 4 34 34
Sub-Criteria (AT) AT3 100 36.37 100 40
AT2 75 27.27 51 20
AT1 58.33 21.21 51 20
AT5 33.33 12.12 25 10
AT4 8.33 3.03 25 10
Sub-Criteria (RS) RS2 100 55.56 100 50
RS1 60 33.33 68 34
RS3 20 11.11 33 16
Sub-Criteria (OS) OS1 100 55.56 100 52
OS3 60 33.33 69 36
OS2 20 11.11 22 12
Sub-Criteria (CS) CS3 100 43.75 100 40
CS4 71.43 31.25 75 30
CS2 42.86 18.75 50 20
CS1 14.29 6.25 25 10
Sub-Criteria (EQ) EQ1 100 55.56 100 53
EQ3 60 33.33 62 33
EQ2 20 11.11 26 14
Sub-Criteria (ITC) ITC3 100 55.56 100 50
ITC2 60 33.33 60 30
ITC1 20 11.11 40 20

13
Automated two‑stage continuous decision support model using exploratory factor…

Fig. 11  a Weightage computa-

Technical Phase
tion model for critical criteria
category. b Weightage computa-
50
tion model. c: Weightage com- Critical Criteria (CC)
putation model for value-added 0.5
Man Power Capabilities (MPC)
criteria category for desirable 6 Quantity of technical staff (MPC2) 3
criteria category 5 Experience of technical personnel
2.5
(MPC1)
4
20 Qualification of technical staff (MPC3) 2
3 Quantity of skilled and non-skilled
1.5
2 labors (MPC5)
Availability of HSE supervisor (MPC4) 1
0.5
Quality Assurance/Control System (QAS)
6
Quality management system (QAS2) 3
5
Monitoring & controlling policy (QAS4) 2.5
18 4 Compliance with quality and
2
3 standard (QAS1)
Quality control policy (QAS3) 1.5
0.5
Firm’s Past Achievement (FPA)
7 Firm’s reputation (FPA5) 3.5
5
15 Quality achieved in past projects (FPA2) 2.5
3 Completed projects in stipulated time/
1.5
cost (FPA3)
0.5
Firm’s Experience (FE)
5 Experience in similar nature work or 2.5
similar contract size (FE2)
4 Amount of similar nature works in the 2
14 past (5yrs.) (FE1)
3
Experience as the main contractor (FE3) 1.5
2 Similar nature and complexity projects
1.5
in hand (FE5)
0.5
Current Financial Strength (CFS)
4
Current liabilities (CFS4) 2
3.5
Current assets (CFS3) 1.75
12 3
Net worth (CFS1) 1.5
1.5
Financial references (CFS2) 0.75
0.5
Financial Track Record (FTR)
6 Proof of credit line from a reputable 3
3 bank (FTR3)
Average annual income in the last 5
10 1.5
1 yrs. (FTR2)
Submission of Financial Statement
(FTR1) 0.5
0.5
Plant & Equipment (PE)
6 Sufficient no. of required equipment as
6 desired in the project and their sources 3
(PE1)
0.5
Management Potential (MP)
1.5
Quality of managerial staff (MP2) 0.75
1.5
HSE policy (MP5) 0.75
5 1
Project management system (MP3) 0.5
0.5 Firms’ planning for completing the
0.25
0.5 work (MP4)
Project Manager qualification & 0.25
experience (MP1)

(a)

13
 A. R. Khoso et al.

Fig. 11  (continued)

(b) (c)

TBS = Technical bid score of screening is responsible for the verification of eligibility
FBS = Financial bid score of contractors. The first major stage is consisting of techni-
TBW = Technical bid weightage cal assessment, and in the second stage, the qualified con-
FBW = Financial bid weightage tractors are treated according to their performance grading
Following the aforementioned stages, a two-stage con- levels in decreasing order of priority. The financial bids of
tinuous model is developed, as shown in Fig. 12, having a all qualified contractors are scrutinized where only a feasi-
screening stage alongside two major steps. The initial stage ble bid contractor is further entertained. The final selection

13
Automated two‑stage continuous decision support model using exploratory factor…

Table 3  Grading assessment values in different cases using SMART​ 9 Automated two‑ stages continuous model
Type of cases Considered Grading assessment values assessment system‑ a hypothetical case
grading Δ𝛼𝛽−Δmin
Vk = Δmax−Δmin for implementation
level

5 grading levels (max 4 (4–0)/(4–0) = 1 Recent developments in the construction sector along the
value = 4) 3 (3–0)/(4–0) = 0.75 globe have transpired this industry into a vibrant and multi-
2 (2–0)/(4–0) = 0.5 faceted industry (Abdelmegid et al. 2020). Owing to rapid
1 (1–0)/(4–0) = 0.25 industrialization, the construction sector entails intricate
0 (0–0)/(4–0) = 0 chores, enormous technological diversity, and multi-oper-
4 grading levels (max 3 (3–0)/(3–0) = 1 ation. However, the public sector in many developed and
value = 3) 2 (2–0)/(3–0) = 0.67 developing country, especially the task of contractor selec-
1 (1–0)/(3–0) = 0.33 tion is still at the embryonic stage and still not being ulti-
0 (0–0)/(3–0) = 0 mately benefited from the automated computerized systems.
3 grading levels (max 2 (2–0)/(2–0) = 1 This automated two-stage continuous model is designed
value = 2) 1 (1–0)/(2–0) = 0.5 with the aim of considering the simplicity and efficient use.
0 (0–0)/(2–0) = 0 With this aim, a system is developed in MS-Excel (“if” and
2 grading levels (max 1 (1–0)/(1–0) = 1 “AND” statements) that can automatically calculate the tech-
value = 1) 0 (0–0)/(1–0) = 0 nical weightages of each contractor and their corresponding
grading levels. It can efficiently deal with the larger pool
of contractors and decides their performance levels and
ranking.
Table 4  Performance levels measurement criteria DMs requires to extract the information from each con-
Level Technical bid Score ­(TBS) Performance grad- tractor, and after verifying the eligibility, the system starts
ing Assessment levels operations. The qualitative data from each contractor is
(PGAL) firstly converted into quantitative and later inserted (see sup-
L4 TBS = 96–100 Outstanding plementary data). The system calculates the total technical
L3 TBS = 90–95 Very good bid score in the first phase, and the performance grading
L2 TBS = 81–90 Good level would be assigned accordingly. Besides, the auto sys-
L1 TBS = 70–80 Hardly accepted tem can filter the passing and failure contractors based on
L0 TBS > 70 Poor performer attained technical weightage and by comparing with thresh-
old values. In the second stage, only passing contractors are
called for their bid proposals. The system can calculate the
financials bid score, and according to their assigned PGAL
stage comprises computing the FSS score that is a combined level, the FSS can be computed. The operational flow of the
weightage of technical and financial bid score based on cer- entire system is illustrated in Fig. 13.
tain percentages associated with each level of contractor’s In order to understand the applicability of the model, a
performance. The contractor with the highest FSS score hypothetical example of four contractors is tested. Further-
would win the competition. more, for exemplifying the calculation process of technical
assessment, a single contractor is evaluated presently. None-
theless, with a similar process, as many as contractors can be
evaluated quickly without limitation. In the beginning, con-
tractors are already verified for their eligibility and therefore

Table 5  Contractor’s final selection score mechanisms

Performance Lev- RT/F (80/20) RT/F (75/25) RT/F (70/30) RT/F (65/35) RT/F (60/40)
els of contractors
TBW (%) FBW (%) TBW (%) FBW (%) TBW (%) FBW (%) TBW (%) FBW (%) TBW (%) FBW (%)

L4 95 5 90 10 85 15 80 20 75 25
L3 90 10 85 15 80 20 75 25 70 30
L2 85 15 80 20 75 25 70 30 65 35
L1 80 20 75 25 70 30 65 35 60 40

13
 A. R. Khoso et al.

Fig. 12  A two-stage continuous decision support model for contractors’ assessment and selection based on performance grading levels

promoted for the technical assessment. Each contractor was possible cases are evaluated to come up with possible solu-
evaluated on 73 sub-criteria divided into three major classi- tions. The technical and financial assessments are performed
fications, as illustrated in Fig. 6. Information on each assess- based on an automatic two-stage continuous model assess-
ment criterion was inserted in the system where based on the ment system. Table 6 demonstrates the process of possible
performance criteria, the assigned level was decided. Let us cases assessed for the award of a project. Each contractor
assume, the hypothetical contractor has “N–3” or below the is assessed based on five distinct technical bid/financial bid
number of technical staff available (“N” will be defined by ratios. The hypothetical case assumes that contractors that
the client as per project need). In this hypothetical case, the fall in the L4 category can never be a minimum bidder and
assigned level will be “1” and the corresponding distribu- the highest bidder among all. The assumption is only made
tion factor will be applied automatically, i.e. 0.75. Similarly, to simplify the process as if the L4 contractor offers the
the contractor would be assessed on each technical crite- minimum bid; then this would be a direct winner in any ­RT/F.
ria, and the assigned level would be inserted, and an auto- The following findings are obtained from hypothetical cases.
matic distribution factor would be assigned. Following this In the 80/20 ratio, the L4 contractor is the winner in all
obtained technical score of the contractor would be added, cases. In the 75/25 ratio, the L3 contractor is a winner only
and the corresponding performance level will be assigned if the L4 contractor quotes a bid of at least 20% above from
(for details see supplementary data file). the minimum bid and at the same time must be a minimum
The next stage determines the financial bid score. For this bidder. While applying 70/30 ratio; L3 contractor can win
hypothetical case, let us assume “σ” represents the minimum the competition even not being the lowest bidder only if, L4
proposed bid in (million USD) by any contractor. Further- contractor quotes 20% above from the minimum bid, and L3
more, let the remaining contractors proposed their bid by a should be the second-lowest. Also, L3 contractor can win the
certain percentage increment say “σ + 5–20%” for instance; contract, if L4 quotes at least 15% above the minimum bid
∅ = variations from 5 to 10% (5%, 7.5%, 10% respectively when at the same time L3 should be the minimum bidder. If
by each contractors), ∇ = variations from 5 to 15% (5%, the client applies 65/35 ratio then; L3 contractor only wins
10%, 15% respectively by each contractors, and ϑ = varia- if L4 contractor quotes at least 20% above from the mini-
tions from 5 to 20% (5%, 10%, 20% respectively by each mum bid when the L1 contractor must be the lowest and if
contractors). To clarify this further, see Table 6, where 90 L4 contractor quotes at least 15% above from the minimum

13
Automated two‑stage continuous decision support model using exploratory factor…

Fig. 13  Automated two- stages continuous model assessment system operational flow

bid when L3 contractor is the lowest. In last, if 60/40 ratio 10 Model’s comparative assessment
is chosen; L3 contractor only wins if L4 contractor quotes
at least 15% above from the minimum bid when L3 contrac- The comparison of the developed model with the past
tor is at 2nd lowest and the L1 contractor is a minimum models propagates impressive comparative outcomes in
bidder and if L4 contractor quotes at least 20% above from five primary directions (Fig. 14).
the minimum bid when L3 contractor is at second-lowest. The critical understanding from the analytical ability
In case if L3 contractor is the lowest bidder, it must win the drives to compare the appropriate directives for the devel-
competition. oped model. The reviewed past models have various short-
The above findings from hypothetical cases exemplify comings in the light of the primary element of a contractor
that the developed system supports a technically highest bid- selection model, i.e. model criteria. The critical understand-
der in the majority of cases. However, other than L4 contrac- ing of this essential element found that the model criteria
tors can also win the project if those contractors compensate except for some cases lack in various directions. Since the
in the financial bids and offer higher benefits to the client. model criteria are key pillars for a robust model, henceforth,
Various ­RT/F are suggested, and those offer almost similar their appropriateness can never be ignored. The contempo-
results and support the technical side of the competition rary models possess a few limitations such as ‘limited model
when the amount of submitted bid is within the estimation criteria’ observed in the models of (Cheng and Li 2004; Jr.
price of the client. However, the client may choose an appro- et al. 2005; Darvish et al. 2009; Watt et al. 2010; Lam and
priate ­RT/F based on the project requisite. Yu 2011; El-abbasy et al. 2013; Jie et al. 2016; Birjandi
et al. 2019; Marcarelli and Nappi, 2019). Similarly, several
ambiguous criteria were considered in the studies of (Ebra-
himi et al. 2016; Jie et al. 2016; Semaan and Salem, 2017;
Birjandi et al. 2019; Cheaitou et al. 2019; Zhao et al. 2019).

13
 A. R. Khoso et al.

Table 6  Hypothetical case

80/20 ratio Case 1 (minimum bidder is L1)
testing for possible rankings of
contractors A (L2 is the 2nd lowest bidder) B (L3 is 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L4
2nd L3 L3 L3 L3 L3 L3
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 2 (minimum bidder is L2)
A (L1 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L4
2nd L3 L3 L3 L3 L3 L3
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 3 (minimum bidder is L3)
A (L1 is the 2nd lowest bidder) B (L2 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L4
2nd L3 L3 L3 L3 L3 L3
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
75/25 ratio Case 1 (minimum bidder is L1)
A (L2 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L4
2nd L3 L3 L3 L3 L3 L3
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 2 (minimum bidder is L2)
A (L1 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L4
2nd L3 L3 L3 L3 L3 L3
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 3 (minimum bidder is L3)
A (L1 is the 2nd lowest bidder) B (L2 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L3 L4 L4 L3
2nd L3 L3 L4 L3 L3 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
70/30 ratio Case 1 (minimum bidder is L1)
A (L2 is the 2nd lowest B (L3 is the 2nd lowest bidder)
bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L3
2nd L3 L3 L3 L3 L3 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1

13
Automated two‑stage continuous decision support model using exploratory factor…

Table 6  (continued)
Case 2 (minimum bidder is L2)
A (L1 is the 2nd lowest B (L3 is the 2nd lowest bidder)
bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L3
2nd L3 L3 L3 L3 L3 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 3 (minimum bidder is L3)
A (L1 is the 2nd lowest B (L2 is the 2nd lowest bidder)
bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L3 L4 L3 L3
2nd L3 L3 L4 L3 L4 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
65/35 ratio Case 1 (minimum bidder is L1)
A (L2 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L4 L3
2nd L3 L3 L3 L3 L3 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 2 (minimum bidder is L2)
A (L1 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L3 L3
2nd L3 L3 L3 L3 L4 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 3 (min minimum bidder is L3)
A (L1 is the 2nd lowest bidder) B (L2 is the lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L3 L3 L3 L3 L3 L3
2nd L4 L4 L4 L4 L4 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
60/40 ratio Case 1 (minimum bidder is L1)
A (L2 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L4 L4 L3 L3
2nd L3 L3 L3 L3 L4 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1
Case 2 (minimum bidder is L2)
A (L1 is the 2nd lowest bidder) B (L3 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L4 L4 L3 L4 L3 L3
2nd L3 L3 L4 L3 L4 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1

13
 A. R. Khoso et al.

Table 6  (continued)
Case 3 (minimum bidder is L3)
A (L1 is the 2nd lowest bidder) B (L2 is the 2nd lowest bidder)
∅ ∇ ϑ ∅ ∇ ϑ
Ranking 1st L3 L3 L3 L3 L3 L3
2nd L4 L4 L4 L4 L4 L4
3rd L2 L2 L2 L2 L2 L2
4th L1 L1 L1 L1 L1 L1

Fig. 14  Comparative assessment of model

In addition to the model criteria, the second chief pillar of and Salem, 2017; Rao and Rathish 2018; Birjandi et al.
the contractor selection model is the employed methods to 2019; Marcarelli and Nappi, 2019). This problem is wors-
analyze the data in terms of decision-making techniques. ened by choosing a technique that is incompatible with the
The successful execution of a construction project is pro- case, for instance, an extremely larger pairwise comparison
foundly impacted by accomplishing the right decision dur- with the qualitative approach produces doubtful results when
ing the selection process. In some cases, heavy reliance on the model criteria are extensive likewise in AHP, ANP, and
human-based selection was adopted even after the inclusion PROMETHEE. Apart from the aforementioned limitations
of MCDM methods such as in (Hasnain et al. 2017; Semaan in MCDM, the addition of higher probabilistic and larger

13
Automated two‑stage continuous decision support model using exploratory factor…

hybrid methods adding further complexity for instance in the et al. 2017; Tomczak and Jaśkowski 2018) which is in con-
case of (Vahdani et al. 2013; Zavadskas et al. 2016; Hasnain trast to public sector regulations.
et al. 2017; Taylan et al. 2017; Hashemi et al. 2018).
The proposed model developed a novel hybrid system
based on the triplet combination of EFA-MACBETH- 11 Conclusion
SMART. The hybrid system is proposed based on the limi-
tations of MACBETH and SMART as a single method. The present work aims to develop a novel automated two-
MACBETH and SMART techniques have fundamental stage continuous decision support model for contractors’
advantages over other MCDM techniques; nonetheless, assessment and selection. The idea of the discontinuous
the ranking problem of MACBETH and weightages cal- progression of the technical phase in the final selection
culations in SMART have remained unresolved. The said stage in past models unrest the authors. Extensive model
inherent issues in both techniques are resolved via a triplet criteria were designed to assess the contractors based on
combination which turns as EFA-MACBETH-SMART for the concept of Critical Criteria, Value-Added Criteria, and
the first time ever. Besides, the concept of two-stage models Desirable Criteria. The model criteria were initially ranked
in contractor selection has been attempted by (San Cristóbal via Exploratory Factor Analysis (EFA) and later weightages
2012; Liu et al. 2017; Cheaitou et al. 2019; Marcarelli and were determined using MACBETH technique in M-MAC-
Nappi 2019; Zhao et al. 2019); however, these models oper- BETH software. The exhaustive assessment of model crite-
ate under the principle of discrete approach, and continuous ria was performed using SMART technique which produced
assessment has not been addressed. The present model prin- the performance grading assessment levels. Each contrac-
cipally operates under the continuous assessment system, tor in the competition was assessed based on technical and
which recognizes the prominence of technical assessment financial bids. The model also preserves the concept of effi-
in the final stage and also retains the concept of value for cient use of public resources in addition to supporting the
money. This model proposes R ­ T/F ratios where the client can technically highest bidders. The idea of extensive contrac-
choose a suitable ratio to evaluate contractors. tor selection using an extensive criteria assessment system
Various models in the past have developed that are not was kept under priority consideration. Furthermore, the
developed with the aim of easiness and simple procedures concept of a simple and straightforward model that sounds
in their model so that those can easily be adopted in a real practical on the ground was resolved using an automated
scenario. In contrast to this, overlapping and burdening of assessment system which is another major contribution of
models with a complicated and large number of MCDM in this study. Hybrid novel combinations of decision-making
a single case seems a common problem in past models. This techniques were implemented based on their true applica-
is confirmed from the (Holt 2010) who reviewed the con- bility. The inherent problem of ordinal data in MACBETH
tractor selection models of two decades and concluded that was resolved using EFA; moreover, the primary issue of
the developed models are additionally complex, henceforth, computing weightages in SMART was resolved using MAC-
are not suitable for the public sector. The term easiness in BETH. This unique combination of EFA-MACBETH and
the application is directly linked with the above two terms, MACBETH-SMART turns to EFA-MACBETH-SMART
i.e. model criteria and mathematical approach, alongside triplet-combination that resolves the inherent issues of deci-
with selection mechanism. In past models, the term hybrid sion-making methods. The findings obtained from the analy-
methods have been presented with complex calculations. A sis of extensive model criteria suggests that 73 criteria (out
large number of extremely complex MCDM methods in a of 76) are the most influential for assessing the contractor.
single case is challenging to apply in a real case scenario, The Critical Criteria category obtained the highest weight-
especially in the public sector where people believe in sim- age of 50%, whereas, Value-Added Criteria and Desirable
ple and straightforward processes such as in (Cheng and Criteria category attained 30%, and 20% weightage respec-
Kang 2012; Zavadskas et al. 2016; Borujeni and Gitinavard tively. Study finds that the final selection of contractor can
2017). Moreover, the selection of contractors based on cer- be performed on various technical bid/financial bid ratio
tain vague criteria such as time, risk, health, and safety, etc. ­(RTF) such as 80/20; 75/25; 70/30; 65/35; and 60/40. A
with bid price creates further complexity in the models, for hypothetical case of contractors’ assessment system tested
instance in the models by (Plebankiewicz 2012; Yang et al. on a few bid price assumptions such as ∅ = variations from
2012; Rao and Rathish 2018; Ye et al. 2018). A few models 5 to 10% (5%, 7.5%, 10% respectively by each contractors),
included bid price during the technical stage, such as (Watt ∇ = variations from 5to 15% (5%, 10%, 15% respectively
et al. 2010; Yang et al. 2016; Semaan and Salem, 2017) and by each contractors, and ϑ = variations from 5 to 20% (5%,
other models evaluated contractors without bid price i.e., 10%, 20% respectively by each contractors). The study con-
(Nieto-Morote and Ruz-Vila 2012; Yang et al. 2016; Taylan cludes that in 80/20 ­RTF ratio, L4 contractor is always a
winner, whereas, in other cases of ­RTF ratios, L3 contractor

13
 A. R. Khoso et al.

could also win the competition based on a few conditions Condition 2 (ii): [quantifying the level of attractiveness]
described under Sect. 9. [ ]
The research supports that each contractor must be ∀ i, j, k, l ∈ S ∶ (i, j)Λ(k, l) ∈ Q ∶ u(i) − u(j) ∕[u(k) − u(l)]
assessed individually based on their corresponding perfor- (8)
mance assessment level. The research further supports the Further,
idea that a contractor must be assessed continuously until the
final selection stage, and technically high contractor which i <Qi j (9)
lies in higher performance levels must win the contractor if
quotes a financial bid under the umbrella of a feasible bid. k <Qi l (10)
Furthermore, the higher benefits must be given to a high-
Eq. A4 and A5 describe the relation between elements
level performer, and correspondingly more compensation
such as i and j, and k and l respectively on the scale of Q
is offered in quoting the bids other than the lowest price.
such that j is Q times greater than i, and l is Q times greater
The study thus concludes through a hypothetical case that
than k. At the scale Q, if i is strongly attractive than j and
a contractor with the highest technical bid score must win
similarly, k is extremely attractive than l; equation A4 and
the award even if not being the lowest but proposed a fea-
equation A5 turns to equation A6 and A7 respectively.
sible bid. Furthermore, research finds that the lowest bid
contractor can also win a contract providing high technical u(i) − u(j) = 5∩ (11)
assessment score.
u(k) − u(l) = 6∩ (12)
∀, ∩ must meet the necessary condition say u(i), u(j), u(k),
Appendix A:Preliminaries in MACBETH u(l) ∈ [0,100].
Applying the Condition 1 and Condition 2 and solving
Let S is a set of finite elements and ∀ i, j, k,l (∈ S) is a subset the equation A6 and A7, the following additive value model
of another number Q [∀ Q ∈ {0, 1, 2, 3, 4, 5, 6}]. To rank would generate as mentioned in equation A8 and A9.
the criteria, the set S must satisfy Condition 1 of the linear
programming from classical MACBETH. n
∑
(13)
( )( )
Condition 1: [For ranking the criteria]. U(S) = wm um
m=1
Say, i, j, k, l represent the four different judgments on a
seven-point semantic scale of differences such that i is more n
attractive than j, and k is more attractive than l, then the first
∑
wm = 1 > 0 (14)
condition can be followed as; m

∀ i, j, k, l ∈ S ∶ [i is more attractive than j ⇔ i

> j Λ k is more attractive than l ⇔ k > l] (6)
Appendix B:Preliminaries in SMART​
e attractiveness can be found through semantic scale.
Condition 1 in classical MACBETH satisfies through direct SMART likewise MACBETH operates on the elemen-
rating or swing weight method where the fundamental inten- tary principle of additive value model. The utility values
tion is to rank the criteria in decreasing order. This repre- in SMART can be calculated by multiplying the criteria
sents the ordinal information (ranking the criteria) from the weightage with their expected utility values. Hence the ear-
DMs. However, the process of MACBETH is based on the liest step is to develop objective weightages. The weightage
assumption of converting the ordinal information into car- ­(wα) can be calculated by the normalization process using
dinal information (based on differences of attractiveness). Eq. B1. The normalization process produces the final criteria
This conversion can be satisfied by following Condition 2 weightage, later on, the criteria value (performance values)
of linear programming of MACBETH. ­(Vak) can be computed.
Condition 2 (i): [Relation as measure of attractiveness
between two elements]. 𝛽
∑
(15)
( ) ( )
From Condition 1 we have the information about the Vak = w𝛼 ∗ Vk
order of criteria and say ∀ (i, j) ⇔ (k, l) ∈ Q (here Q denotes 𝛼=i

the measure of difference of attractiveness), then; The utility value of each criterion can be calculated using
Eq. B2, the value is normalized on a scale of 0–1.
∀ i, j, k, l ∈ S ∶ [iQj ⇔ u(i) > u(j)Λ kQl ⇔ u(k) > u(l)]
(7)

13
Automated two‑stage continuous decision support model using exploratory factor…

Δ𝛼𝛽 − Δmin André FJ, Riesgo L (2007) A non-interactive elicitation method for
Vk = (16) non-linear multiattribute utility functions: theory and application
Δmax − Δmin to agricultural economics. Eur J Oper Res 181:793–807. https://​
where; wα is the relative weightage of each criteria/sub- doi.​org/​10.​1016/j.​ejor.​2006.​06.​020
Awwad R, Ammoury M (2019) Owner’s perspective on evolution of
criteria (from 1 to 100). Vk is the utility value of each crite- bid prices under various price-driven bid selection methods. J
ria/sub-criteria [0 to 1 scale; 1 = highest, 0 = lowest]. ∆min Comput Civ Eng 33:1–12. https://​doi.​org/​10.​1061/​(ASCE)​CP.​
is the minimum scale value. ∆max is the highest scale value. 1943-​5487.​00008​03
Bagal HA, Soltanabad YN, Dadjuo M, Wakil K, Ghadimi N (2018)
Risk-assessment of photovoltaic-wind-battery-grid based large
industrial consumer using information gap decision theory. Sol
Supplementary Information The online version contains supplemen- Energy 169:343–352. https://​doi.​org/​10.​1016/j.​solen​er.​2018.​05.​
tary material available at https://d​ oi.o​ rg/1​ 0.1​ 007/s​ 12652-0​ 21-0​ 3186-w. 003
Ballesteros-Pérez P, González-Cruz MC, Cañavate-Grimal A (2013)
Acknowledgement We are thankful to our experts who acted as deci- On competitive bidding: scoring and position probability graphs.
sion-makers and directed this work. Int J Project Manag 31:434–448. https://​doi.​org/​10.​1016/j.​ijpro​
man.​2012.​09.​012
Funding This research did not receive any specific grant from funding Ballesteros-Pérez P, Skitmore M, Pellicer E, González-Cruz MC
agencies in the public, commercial, or not-for-profit sectors. (2015) Scoring rules and abnormally low bids criteria in con-
struction tenders: a taxonomic review. Constr Manag Econ
Data availability The required data will be provided upon request. 33:259–278. https://​doi.​org/​10.​1080/​01446​193.​2015.​10599​51
Barfod MB, Salling KB (2015) A new composite decision support
framework for strategic and sustainable transport appraisals.
Code availability Not applicable. Trans Res Part A Policy Pract 72:1–15. https://​d oi.​o rg/​1 0.​
1016/j.​tra.​2014.​12.​001
Declarations Bendaña R, del Caño A, Pilar de la Cruz M (2008) Contractor selec-
tion: fuzzy-control approach. Can J Civ Eng 35:473–486.
Conflict of interest The authors declare that there is no conflict and https://​doi.​org/​10.​1139/​L07-​127
competing interest with any individual or any organization while ex- Benson NF, Kranzler JH, Floyd RG (2016) Examining the integ-
ploring and writing this research. rity of measurement of cognitive abilities in the prediction of
achievement: comparisons and contrasts across variables from
higher-order and bifactor models. Can J Civ Eng. https://​doi.​
org/​10.​1016/j.​jsp.​2016.​06.​001
Birjandi AK, Akhyani F, Sheikh R, Sana SS (2019) Evaluation and
References selecting the contractor in bidding with incomplete informa-
tion using MCGDM method. Soft Comput 23:10569–10585.
https://​doi.​org/​10.​1007/​s00500-​019-​04050-y
Abdelmegid MA, González VA, Poshdar M, O’Sullivan M, Walker Borujeni MP, Gitinavard H (2017) Evaluating the sustainable mining
CG, Ying F (2020) Barriers to adopting simulation modelling in contractor selection problems: an imprecise last aggregation
construction industry. Autom Constr 111:1–13. https://​doi.​org/​ preference selection index method. J Sust Mining 16:207–218.
10.​1016/j.​autcon.​2019.​103046 https://​doi.​org/​10.​1016/j.​jsm.​2017.​12.​006
Abdelrahman M, Zayed T, Elyamany A (2008) Best-value model Brook M (2017) Estimating and tendering for construction work. J
based on project specific characteristics. J Construct Eng Manag Operat Res 158:308
134:179–188. https://​doi.​org/​10.​1061/​(ASCE)​0733-​9364(2008)​ Brugha CM (2004) Phased multicriteria preference finding. Eur J
134:​3(179) Oper Res 158:308–316. https://​doi.​org/​10.​1016/j.​e jor.​2003.​
Abedinia O, Zareinejad M, Doranehgard MH, Fathi G, Ghadimi N 06.​006
(2019) Optimal offering and bidding strategies of renewable Brunjes BM (2020) Competition and federal contractor performance.
energy based large consumer using a novel hybrid robust-sto- J Pub Admin Res Theory 30:202–219. https://​doi.​org/​10.​1093/​
chastic approach. J Clean Prod 215:878–889. https://​doi.​org/​10.​ jopart/​muz027
1016/j.​jclep​ro.​2019.​01.​085 Bryman A, Cramer D (1997) Quantitative Data Analysis with SPSS
Abudayyeh O, Zidan SJ, Yehia S, Randolph D (2007) Hybrid prequali- for Windows. Routledge
fication-based, innovative contracting model using AHP. J Manag Cheaitou A, Larbi R, Al Housani B (2019) Decision making framework
Eng 23:88–96. https://d​ oi.o​ rg/1​ 0.1​ 061/(​ ASCE)0​ 742-5​ 97X(2007)​ for tender evaluation and contractor selection in public organiza-
23:​2(88) tions with risk considerations. Soc Plann Sci 68:1–12. https://d​ oi.​
Afshar MR, Alipouri Y, Sebt MH, Chan WT (2017) A type-2 fuzzy set org/​10.​1016/j.​seps.​2018.​02.​007
model for contractor prequalification. Autom Constr 84:356–366. Cheng MY, Kang ST (2012) Integrated fuzzy preference relations with
https://​doi.​org/​10.​1016/j.​autcon.​2017.​10.​003 decision utilities for construction contractor selection. J Chin
Albano GL, Bianchi M, Spagnolo G (2006) Bid avarage methods in Inst Eng 35:1051–1063. https://d​ oi.o​ rg/1​ 0.1​ 080/0​ 25338​ 39.2​ 012.​
Procurement. Rivista Di Politica Economica 96:41–62. Retrieved 708510
from https://​mpra.​ub.​uni-​muenc​hen.​de/​id/​eprint/​8997 Cheng EWL, Li H (2004) Contractor selection using the analytic net-
Alptekin O, Alptekin N (2017) Analysis of criteria influencing con- work process. Constr Manag Econ 22:1021–1032. https://​doi.​
tractor selection using TOPSIS method. Mater Sci Eng IOP Pub. org/​10.​1080/​01446​19042​00020​2852
https://​doi.​org/​10.​1088/​1757-​899X/​245/6/​062003 Chou SY, Chang YH (2008) A decision support system for supplier
Anagnostopoulos KP, Vavatsikos AP (2006) An AHP model for con- selection based on a strategy-aligned fuzzy SMART approach.
struction contractor prequalification. Oper Res Int J 6:333–346. Expert Syst Appl 34:2241–2253. https://​doi.​org/​10.​1016/j.​eswa.​
https://​doi.​org/​10.​1007/​BF029​41261 2007.​03.​001

13
 A. R. Khoso et al.

Costa BE, Chagas MP (2004) A career choice problem: An example Hashemi H, Mousavi SM, Zavadskas EK, Chalekaee A, Turskis Z
of how to use MACBETH to build a quantitative value model (2018) A new group decision model based on Grey-Intuitionistic
based on qualitative value judgments. Eur J Operational Res Fuzzy-ELECTRE and VIKOR for contractor assessment prob-
153:323–331. https://​doi.​org/​10.​1016/​S0377-​2217(03)​00155-3 lem. Sustainability 10:1–19. https://d​ oi.o​ rg/1​ 0.3​ 390/s​ u1005​ 1635
Costa BE, Vansnick JC (1994) MACBETH - An interactive path Hasnain M, Thaheem MJ, Ullah F (2017) Best value contractor selec-
towards the construction of cardinal value functions. Internat tion in road construction projects: ANP-based decision support
Trans Operat Res 1:489–500. https://​doi.​org/​10.​1016/​0969-​ system. Internat J Civil Eng 16:695–714. https://d​ oi.o​ rg/1​ 0.1​ 007/​
6016(94)​90010-8 s40999-​017-​0199-2
Costa BE, De Corte J-M, Vansnick J-C (2003) MACBETH (Overview Holt G (2010) Contractor selection innovation: Examination of two
of MACBETH multicriteria decision analysis approach). Internat decades’ published research. Constr Innov 10:304–328. https://​
J Inform Technol Dec Making 11:359–387 doi.​org/​10.​1108/​14714​17101​10600​97
Costa BE, Oliveira CS, Vieira V (2008) Prioritization of bridges and Hurson C, Mastorakis K, Siskos Y (2012) Application of a synergy
tunnels in earthquake risk mitigation using multicriteria decision of MACBETH and MAUT multicriteria methods to portfolio
analysis: application to Lisbon. Omega 36:442–450. https://​doi.​ selection in Athens stock exchange. Internat J Mult Dec Making
org/​10.​1016/j.​omega.​2006.​05.​008 2:113–127. https://​doi.​org/​10.​1504/​IJMCDM.​2012.​046939
Cox R, Sanchez J, Revie CW (2013) Multi-criteria decision analysis Ioannou PG, Awwad RE (2010) Below-average bidding method. J Con-
tools for prioritising emerging or re-emerging infectious diseases str Eng Manag 136:936–946. https://​doi.​org/​10.​1061/​(ASCE)​
associated with climate change in Canada. PLoS ONE. https://​ CO.​1943-​7862.​00002​02
doi.​org/​10.​1371/​journ​al.​pone.​00683​38 Jeremy F, Barkus E, Yavorsky C (2006) Understanding and modeling
Dabrowski M (2014) The simple multi attribute rating technique health behavior: the multi-stage model of health behavior change.
(SMART). DTU Trans Comp Series 6:1–6 J Constr Eng. https://​doi.​org/​10.​1177/​13591​05306​058845
Darvish M, Yasaei M, Saeedi A (2009) Application of the graph theory Jiang H, Zhang Y (2016) An investigation of service quality, customer
and matrix methods to contractor ranking. Int J Project Manage satisfaction and loyalty in China’s airline market. J Air Trans
27:610–619. https://​doi.​org/​10.​1016/j.​ijpro​man.​2008.​10.​004 Manag. https://​doi.​org/​10.​1016/j.​jairt​raman.​2016.​07.​008
Distefano C, Zhu M, Mîndrilã D (2009) Understanding and using factor Jie BL, Huo T, Meng JJG (2016) Identification of key contractor char-
scores: considerations for the applied researcher. Pract Assess acteristic factors that affect project success under different project
Res Evalu. https://​doi.​org/​10.​7275/​da8t-​4g52 delivery systems: empirical analysis based on a group of data
Duarte BPM, Reis A (2006) Developing a projects evaluation system from China. J Manag Eng. https://​doi.​org/​10.​1061/​(ASCE)​ME.​
based on multiple attribute value theory. Comput Operations Res 1943-​5479.​00003​88
33:1488–1504. https://​doi.​org/​10.​1016/j.​cor.​2004.​11.​003 Joerin F, Cool G, Rodriguez MJ, Gignac M, Bouchard C (2010) Using
Ebrahimi A, Alimohammadlou M, Mohammadi S (2016) Identification multi-criteria decision analysis to assess the vulnerability of
and prioritization of effective factors in assessment and ranking drinking water utilities. Environ Monit Assess 166:313–330.
of contractors using fuzzy multi-criteria techniques. Dec Sci Lett https://​doi.​org/​10.​1007/​s10661-​009-​1004-8
5:95–108. https://​doi.​org/​10.​5267/j.​dsl.​2015.8.​001 Khodaei H, Hajiali M, Darvishan A, Sepehr M, Ghadimi N (2018)
Ebrahimnejad S, Mousavi SM, Tavakkoli-Moghaddam R, Hashemi Fuzzy-based heat and power hub models for cost-emission opera-
H, Vahdani B (2012) A novel two-phase group decision mak- tion of an industrial consumer using compromise programming.
ing approach for construction project selection in a fuzzy envi- Appl Therm Eng 137:395–405. https://​doi.​org/​10.​1016/j.​applt​
ronment. Appl Math Model 36:4197–4217. https://​doi.​org/​10.​ herma​leng.​2018.​04.​008
1016/j.​apm.​2011.​11.​050 Khoso AR, Md Yusof A (2020) Extended review on contractor selec-
Edwards W, Barron FH (1994) Smarts and smarter: Improved simple tion in construction projects. Can J Civ Eng 47:771–789. https://​
methods for multiattribute utility measurement. Organ Behav doi.​org/​10.​1139/​cjce-​2019-​0258
Hum Decis Process 60:306–325. https://​doi.​org/​10.​1006/​obhd.​ King G, Keohane RO, Verba S (1994) Designing social inquiry scien-
1994.​1087 tific inference in qualitative research. Princeton University Press
El-abbasy MS, Zayed T, Asce M, Ahmed M, Alzraiee H, Abouhamad Kline P (1994) An easy guide to factor analysis. Routledge
M (2013) Contractor selection model for highway projects using Kog F, Yaman H (2014) A meta classification and analysis of contrac-
integrated simulation and analytic network process. J Constr Eng tor selection and prequalification. Proc Eng 84:302–310. https://​
Manag 139:755–767. https://​doi.​org/​10.​1061/​(ASCE)​CO.​1943-​ doi.​org/​10.​1016/j.​proeng.​2014.​10.​555
7862.​00006​47 Konidari P, Mavrakis D (2007) A multi-criteria evaluation method
Ertay T, Kahraman C, Kaya İ (2013) Evaluation of renewable energy for climate change mitigation policy instruments. Energy Policy
alternatives using macbeth and fuzzy ahp multicriteria methods: 35:6235–6257. https://​doi.​org/​10.​1016/j.​enpol.​2007.​07.​007
the case of Turkey. Technol Econ Dev Econ 19:38–62. https://​ Kumar D, Alappat BJ (2005) Evaluating leachate contamination
doi.​org/​10.​3846/​20294​913.​2012.​762950 potential of landfill sites using leachate pollution index. Clean
Gao W, Darvishan A, Toghani M, Mohammadi M, Abedinia O, Technol Environ Policy 7:190–197. https://​doi.​org/​10.​1007/​
Ghadimi N (2019) Different states of multi-block based forecast s10098-​004-​0269-4
engine for price and load prediction. Int J Electr Power Energy Kundakcı N (2019) An integrated method using MACBETH and EDAS
Syst 104:423–435. https://​doi.​org/​10.​1016/j.​ijepes.​2018.​07.​014 methods for evaluating steam boiler alternatives. J Multi-Criteria
Ghadimi N, Akbarimajd A, Shayeghi H, Abedinia O (2018) Two stage Dec Analysis 26:27–34. https://​doi.​org/​10.​1002/​mcda.​1656
forecast engine with feature selection technique and improved Kundakcı N, Işık AT (2016) Integration of MACBETH and COPRAS
meta-heuristic algorithm for electricity load forecasting. Energy methods to select air compressor for a textile company. Dec Sci
161:130–142. https://​doi.​org/​10.​1016/j.​energy.​2018.​07.​088 Lett 5:381–394. https://​doi.​org/​10.​5267/j.​dsl.​2016.2.​003
Gómez-Limón JA, Martínez Y (2006) Multi-criteria modelling of irriga- Kwak S-J, Yoo S-H, Kim T-Y (2001) A constructive approach to air-
tion water market at basin level: a Spanish case study. Eur J Oper quality valuation in Korea. Ecol Econ 38:327–344. https://​doi.​
Res 173:313–336. https://​doi.​org/​10.​1016/j.​ejor.​2004.​12.​009 org/​10.​1016/​S0921-​8009(01)​00190-2
Gurgun AP, Koc K (2020) Contractor prequalification for green build- Lai KK, Liu SL, Wang SY (2004) A method used for evaluating bids in
ings—evidence from Turkey. Eng Constr Archit Manag. https://​ the chinese construction industry. Int J Project Manage 22:193–
doi.​org/​10.​1108/​ECAM-​10-​2019-​0543 201. https://​doi.​org/​10.​1016/​S0263-​7863(03)​00009-7

13
Automated two‑stage continuous decision support model using exploratory factor…

Lam KC, Yu CY (2011) A multiple kernel learning-based decision evaluation of construction contractors. Autom Constr 59:149–
support model for contractor pre-qualification. Autom Constr 157. https://​doi.​org/​10.​1016/j.​autcon.​2015.​02.​009
20:531–536. https://​doi.​org/​10.​1016/j.​autcon.​2010.​11.​019 Safa M, Yee M-H, Rayside D, Haas CT (2016) Optimizing contractor
Liu B, Huo T, Liao P, Yuan J, Sun J, Hu X (2017) A special Partial selection for construction packages in capital projects. J Comput
Least Squares (PLS) path decision modeling for bid evaluation of Civ Eng 30:1–12. https://​doi.​org/​10.​1061/​(asce)​cp.​1943-​5487.​
large construction projects. KSCE J Civ Eng 21:579–592. https://​ 00005​55
doi.​org/​10.​1007/​s12205-​016-​0702-3 San Cristóbal JR (2012) Contractor selection using multicriteria deci-
Madeira AG, Cardoso MM, Belderrain MCN, Correia AR, Schwanz sion-making methods. J Const Eng Manag 138:751–758. https://​
SH (2012) Multicriteria and multivariate analysis for port per- doi.​org/​10.​1061/​(ASCE)​CO.​1943-​7862.​00004​88
formance evaluation. Int J Prod Econ 140:450–456. https://​doi.​ Semaan N, Salem M (2017) A deterministic contractor selection deci-
org/​10.​1016/j.​ijpe.​2012.​06.​028 sion support system for competitive bidding. Eng Constr Archit
Marcarelli G, Nappi A (2019) Multicriteria approach to select the most Manag 24:61–77. https://​doi.​org/​10.​1108/​ECAM-​06-​2015-​0094
economically advantageous tender: the application of AHP in Taylan O, Kabli MR, Porcel C, Herrera-Viedma E (2017) Contractor
Italian public procurement. J Pub Procurement 19:201–223. selection for construction projects using consensus tools and big
https://​doi.​org/​10.​1108/​JOPP-​05-​2018-​0020 data. Int J Fuzzy Syst 20:1267–1281. https://​doi.​org/​10.​1007/​
Marler RT, Arora JS (2010) The weighted sum method for multi- s40815-​017-​0312-3
objective optimization: new insights. Struct Multidiscip Optim Teixeira De Almeida A (2007) Multicriteria decision model for out-
41:853–862. https://​doi.​org/​10.​1007/​s00158-​009-​0460-7 sourcing contracts selection based on utility function and ELEC-
Marzouk M (2008) A superiority and inferiority ranking model for TRE method. Comput Operat Res 34:3569–3574. https://d​ oi.o​ rg/​
contractor selection. Constr Innov 8:250–268. https://​doi.​org/​10.​ 10.​1016/j.​cor.​2006.​01.​003
1108/​14714​17081​09126​44 Tomczak M, Jaśkowski P (2018) Application of Type-2 interval
Marzouk M (2010) An application of electre III to contractor selection. fuzzy sets to contractor qualification process. KSCE J Civ Eng
Const Res Congress 2010:1316–1324 22:2702–2713. https://​doi.​org/​10.​1007/​s12205-​017-​0431-2
Mateus RJG, Costa BE, Matos PV (2017) Supporting multicriteria Topcu YI (2004) A decision model proposal for construction contractor
group decisions with MACBETH tools: selection of sustainable selection in Turkey. Build Environ 39:469–481
brownfield redevelopment actions. Group Dec Negot 26:495– Vahdani B, Mousavi SM, Hashemi H, Mousakhani M, Tavakkoli-
521. https://​doi.​org/​10.​1007/​s10726-​016-​9501-y Moghaddam R (2013) A new compromise solution method for
Minchin RE Jr, Smith GR (2005) Quality-based contractor rating model fuzzy group decision-making problems with an application to the
for qualification and bidding purposes. J Manage Eng. https://d​ oi.​ contractor selection. Eng Appl Artif Intell 26:779–788. https://​
org/​10.​1061/​(ASCE)​0742-​597X(2005)​21:​1(38) doi.​org/​10.​1016/j.​engap​pai.​2012.​11.​005
Monat JP (2009) The benefits of global scaling in multi-criteria deci- Waara F, Bröchner J (2006) Price and nonprice criteria for contractor
sion analysis. Judgm Decis Mak 4:492–508 selection. J Constr Eng Manag 132:797–804
Opricovic S, Tzeng GH (2004) Compromise solution by MCDM Wang W, Yu W, Yang I, Lin C, Lee M, Cheng Y-Y (2013) Applying
methods: a comparative analysis of VIKOR and TOPSIS. Eur J the AHP to support the best-value contractor selection – les-
Oper Res 156:445–455. https://d​ oi.o​ rg/1​ 0.1​ 016/S
​ 0377-2​ 217(03)​ sons learned from two case studies in Taiwan. J Civ Eng Manag
00020-1 19:24–36. https://​doi.​org/​10.​3846/​13923​730.​2012.​734851
Padhi SS, Mohapatra Pratap KJ (2010) Centralized bid evaluation for Wang J, Yu B, Tam VWY, Li J, Xu X (2019) Critical factors affecting
awarding of construction projects - a case of India government. willingness of design units towards construction waste mini-
Int J Project Manage 28:275–284. https://​doi.​org/​10.​1016/j.​ijpro​ mization: an empirical study in Shenzhen, China. J Clean Prod
man.​2009.​06.​001 221:526–535. https://​doi.​org/​10.​1016/j.​jclep​ro.​2019.​02.​253
Phogat S, Gupta AK (2019) Evaluating the elements of just in time Watt DJ, Kayis B, Willey K (2010) The relative importance of tender
(JIT) for implementation in maintenance by exploratory and evaluation and contractor selection criteria. Int J Project Manage
confirmatory factor analysis. Internat J Quality Reliabil Manag 28:51–60. https://​doi.​org/​10.​1016/j.​ijpro​man.​2009.​04.​003
36:7–24. https://​doi.​org/​10.​1108/​IJQRM-​12-​2017-​0279 Winterfeldt VD, Edwards R (1986) Decision analysis and behavioral
Plebankiewicz E (2012) A fuzzy sets based contractor prequalification research. Cambridge University Press
procedure. Autom Constr 22:433–443. https://​doi.​org/​10.​1016/j.​ Yang IT, Wang WC, Yang TI (2012) Automatic repair of inconsistent
autcon.​2011.​11.​003 pairwise weighting matrices in analytic hierarchy process. Autom
Rao MVK, Rathish VS (2018) Optimal contractor selection in con- Constr 22:290–297. https://d​ oi.o​ rg/1​ 0.1​ 016/j.a​ utcon.2​ 011.0​ 9.0​ 04
struction industry: the fuzzy way. J Instit Eng Series A 99:67–78. Yang J-B, Wang H-H, Wang W-C, Ma S-M (2016) Using data envelop-
https://​doi.​org/​10.​1007/​s40030-​018-​0271-1 ment analysis to support best-value contractor selection. J Civ
Rashvand P, Zaimi M, Majid A, Pinto JK (2015) Contractor manage- Eng Manag 22:199–209. https://d​ oi.o​ rg/1​ 0.3​ 846/1​ 39237​ 30.2​ 014.​
ment performance evaluation model at prequalification stage. 897984
Expert Syst Appl 42:5087–5101. https://​doi.​org/​10.​1016/j.​eswa.​ Ye K, Zeng D, Wong J (2018) Competition rule of the multi-criteria
2015.​02.​043 approach: what contractors in china really want? J Civ Eng
Rayno B, Parnell GS, Burk RC, Woodruff BW (1998) A methodology Manag 24:155–166. https://​doi.​org/​10.​3846/​jcem.​2018.​459
to assess the utility of future space systems. J Multi-Criteria Dec Zavadskas EK, Turskis Z, Eviciene JA (2016) Selecting a contractor
Analy 6:344–354 by using a novel method for multiple attribute analysis: weighted
Rocha de Gouveia M (2002) The price factor in EC public tenders. Pub aggregated sum product assessment with grey values (WASPAS-
Contract Law J 31:679–93. Retrieved from http://​www.​jstor.​org/​ G). Stud Inform Control 24:141–150
stable/​25754​500 Zhao L, Liu W, Wu Y (2019) Bid evaluation decision for major project
Saeedi M, Moradi M, Hosseini M, Emamifar A, Ghadimi N (2019) based on analytic hierarchy process and data envelopment analy-
Robust optimization based optimal chiller loading under cooling sis cross-efficiency model. J Ambient Intell Humaniz Comput.
demand uncertainty. Appl Therm Eng 148:1081–1091. https://​ https://​doi.​org/​10.​1007/​s12652-​019-​01564-z
doi.​org/​10.​1016/j.​applt​herma​leng.​2018.​11.​122
Safa M, Shahi A, Haas CT, Fiander-McCann D, Safa M, Hipel K, Publisher’s Note Springer Nature remains neutral with regard to
MacGillivray S (2015) Competitive intelligence (CI) for jurisdictional claims in published maps and institutional affiliations.

13

View publication stats