Barriers analysis for reverse logistics in Thailand’s palm oil industry using fuzzy multi-criteria decision-making method for prioritizing the solutions

Abstract

The palm oil industry faces significant pressure because of the environmental problems from the plantation process and the palm oil milling process in their supply chain, which is due to the constraints of natural resources and the growing ecological awareness among customers. Reverse logistics (RL) has been considered as a systematic approach for the palm oil industry to improve the environmental impact. However, there are more barriers which cannot be detected in the RL. Therefore, this study proposes robust and flexible strategies for overcoming such barriers by focusing on identifying and ranking the solutions on RL adoption in Thailand’s palm oil industry. We use a fuzzy multi-criteria decision-making (MCDM) methods based on fuzzy analytical network process (ANP) and Vise Kriterijumska Optimizacija I Kompromisno Resenje (VIKOR) to prioritize related solutions and strategies. Firstly, fuzzy ANP is applied to identify the weighting of each barrier using pairwise comparison, while the final ranking of the solution on RL adoption is obtained through VIKOR. This model will help business analysis and supply chain managers formulate both short-term and long-term, flexible decision strategies for successfully managing and implementing RL adoption in the supply chain scenarios. In addition, sensitivity analysis is used to illustrate the method, and comparisons with other existing methods is also presented in this study.

1 Introduction

There is an increasing demand for palm oil worldwide. Thailand is the world’s third-largest palm oil producing country in the world. The increased demand in Thailand for palm oil for food, cosmetics and biodiesel consumption has resulted in a rapid expansion of several environmental impacts. In recent years, the palm oil plantation in Thailand has become a very serious industry since the Ministry of Energy has launched a policy of Alternative Energy Development Palm (AEDP) 2015–2036. This has raised concerns on the increasing environmental impacts of the processing of palm oil and the lack of environmental regulations (Choong and Mckay 2014).

The palm oil industry refers to the whole complexity of the industrial activities that are directly linked to the production of palm oil. The operation in the palm oil industry consists of many processes, for example, production waste, solid waste, as well as waste management (Saswattecha et al. 2015). These environmental impacts should be considered with a reduction in sources to be used by the industry to facilitate environmental protection. On behalf of a producer in the palm oil industry, we are capable of putting efforts on reducing the environmental impacts by proactively encouraging RL practice to create an environmental sustenance in the supply chain management (SCM).

Reverse logistics or so called “Closed-loop Supply Chain (CLSC)” include activities associated with the collection, and subsequent recovery or disposal of used products (Ilgin and Gupta 2010). More specifically, it concerns with issues such as encouragement of product reclaiming, recycling, remanufacturing, reuse, and disposal (Govindan et al. 2012). Many companies have applied the RL concept to their policies and strategies regarding sustainable development, focusing on reducing waste and creating value added from the return of end-of-life products. Moreover, RL is considered a mutual responsibility of both producers and consumers to minimize the generation of wastes by promoting reuse, remanufacturing, recycling as well as safe disposal of discarded items meanwhile contributing to sustainability and circular economy (Bouzon et al. 2016). Despite the positive impact of RL implementation, there are certain types of barriers that impede the adoption of RL. Each type of barrier requires specific treatments and priorities in accordance with its characteristics, resources, capabilities, and strategies. Particularly, flexible and feasible solutions should be proposed and ranked based on a priority basis (Prakash and Barua 2015).

Previous studies pointed inadequate attention to barriers and drivers of RL implementation in various industries and many countries (Abdulrahman et al. 2014; Rogers and Tibben-Lembke 2011; Zhu and Geng 2013). Accordingly, there has been a few research conducted to explore critical barriers and solutions in Thailand’s palm oil industry which is a rapidly growing sector regarding production. More emphasis on the significance of RL has driven a large number of organizations in Thailand’s palm oil industry to practice it. However, there are merely a few companies that have successfully adapted RL practices while many have failed due to various barriers to their RL implementation and there are cannot be detected in the RL process. Thus, to solve these problems, companies must understand and consider the ranking of the barriers that affect RL practices implementation as well as focusing on the ranking of the solutions to minimize the effects of these challenges.

This motivated us to look into the issues relates to RL implementation, specifically to identify and evaluate possible solutions for overcoming RL challenges in Thailand’s palm oil industry. Initially, we identify related barriers, then propose the solutions relevant to the issues and finally, we rank the solutions into orders on a priority basis. Thus, we are intended to propose a method which can robust and flexible strategies for overcoming such barriers. The multi-criteria decision-making approach is applied for ranking the solutions of RL adoption whereas the involvement of people’s opinions is expressed through linguistic variables, hence this study applies evaluation using precise and specific numerical values. We apply fuzzy set theory combined with ANP which allows uncertainty and fuzziness in decision-making. The ANP is a multiple-criteria decision-making technique usually applied to the selection and evaluation of criteria that contains both quantitative and qualitative analysis (Onut et al. 2011). So that the model can be extracted by considering the impact of each criterion on every other criterion using comparison matrices.

Moreover, the fuzzy set provides an appropriate method for modeling the fuzzy situation involving complicated relationships, including feedback and dependency, which allows the decision-making process to be conducted in a clear and systematic way (Rabbani et al. 2014). When dealing with human cognition, the lack of certainty is an unavoidable problem and the fuzzy set is very useful in such situations by applying mathematical techniques for resolving the ambiguity problem with the different kinds of evaluation information without information losing, which can be yield more reasonable result. Furthermore, VIKOR method is employed to decide the priority and ranking of solutions for RL practices. Such information can be applied by involving organizations in Thailand’s palm oil industry to develop practical strategies regarding green supply chain management (GSCM). Beside, this model will help business analysis and supply chain managers formulate both short-term and long-term, flexible decision strategies for successfully managing and implementing RL adoption in the supply chain scenarios.

The remainder of this work is organized as follows: Sect. 2 is a brief on related literature about the concepts adopted for this research. Section 3 defines the barriers and solutions of RL adoption. Section 4 introduces the fuzzy ANP and VIKOR approaches. The ranking of the solutions for RL adoption is illustrated in Sect. 5. Section 6 explains the results and discussions, including sensitivity analysis and comparisons methods, and finally, Sect. 7 provides the conclusion.

2 Related literature

2.1 Barriers to reverse logistics in the palm oil industry

Strict environmental regulations and dramatic decreases in raw material resources have driven the needs of RL adoption in various industries. As a result, RL plays important roles in the reverse flow of CLSC that focuses on product take-back and value recovery by wholly or partially reusing raw materials extracted from end-of-life products. RL includes the process of reducing and disposing of wastes generated from production, packaging, product usage altogether with the process of reverse distribution such as product take-back (Rogers and Tibben-Lembke 1999). Although some companies perceive RL as a potential tool for enhancing competitiveness and have implemented RL into their supply chains, many still focus only on forward supply chain which is the main revenue generation unit.

Previous researches have discussed the multiple barriers of RL, so as Bei and Sun (2005) proposed RL implementation model and identified critical RL barriers in Chinese manufacturing regarding multiple aspects such as management, finance, policy making as well as infrastructure. Likewise, Prakash and Barua (2016) studied the barriers of reverse logistics adoption in Indian electronics industry, demonstrated the identification and ranking of solutions for the problems and at the same time developed the evaluation tool for third party reverse logistics partner selections. Bouzon et al. (2015) investigated RL-implemented supply chain dealing with its barriers by focusing on the recovery of end-of-life (EOL) products that mostly contain raw materials from the mining and minerals industry, leading to identification and analysis of the interactions among the barriers of RL developments in Brazil as well as evaluation of the barriers to reverse logistics under a multiple stakeholders’ perspective analysis. In the same way, Zailani et al. (2017) empirically analyzed the barriers of product return management in Malaysia automotive industry, likewise Xia et al. (2015) analyzed and identified possible internal barriers to RL in China remanufacturers of auto parts.

In Thailand, palm oil manufacturers have encountered some obstacles that discourage the implementation of reverse logistics practices. A large number of organizations have faced a lot of barriers, in fact inadequate focus on environmental issues, insufficient government support, lack of knowledge in reverse logistics practices, or market price sensitivity. Moreover, some manufacturers still have a poor understanding of the international standards such as new management procedures, additional environmental laws and regulations, etc., all of which negatively impact the adoption of RL practices in Thailand’s palm oil industry. The barriers to implementation of RL practices and possible solutions are discussed in Sect. 3.

2.2 Fuzzy ANP and VIKOR

This study proposes the analysis of RL barriers based on fuzzy multi-criteria decision-making (FMCDM). In classical MCDM, the ratings and the weights of the criteria are visible, but under many conditions, crisp data are inadequate to model the actual situation. To overcome this limitation, fuzzy MCDM is proposed as it allows fuzzy assessment and multiple expert judgments. Fuzzy set theory is employed responding to the presence of imprecise and vague information in the process of barriers evaluation. A significant amount of literature on FMCDM has been published, and some have stated approaches to the various characteristics of decision-making problems with vagueness (Galankashi et al. 2015; Govindan et al. 2013; Tseng et al. 2014; Uygun and Dede 2016). Selecting from the set of available MCDM approaches, a two-stage solution methodology including the fuzzy ANP and VIKOR is proposed for this study. Fuzzy ANP method has recently been applied to many research fields, either alone or in combination with other methods (Chatterjee and Kar 2017; Khan et al. 2018).

Besides, some studies combined both fuzzy ANP and VIKOR methods. Yalcin et al. (2012) applied fuzzy multi-criteria decision-making methods in which fuzzy AHP, TOPSIS and VIKOR were utilized for the evaluation of the financial performance of manufacturing industries in Turkey. Similarly, a site selection decision framework that utilized fuzzy ANP-VIKOR for large commercial rooftop PV systems based on the sustainability perspective was presented by Wu et al. (2018). Awasthi et al. (2018) integrated a fuzzy AHP-VIKOR approach based framework for selection of sustainable global suppliers and also, Liu et al. (2018) integrated ANP-VIKOR methods for choosing sustainable suppliers with interval type-2 fuzzy sets. Furthermore, Prakash and Barua (2015) proposed a fuzzy AHP-TOPSIS method for prioritizing the solutions to overcome the barriers to reverse logistics adoption and also proposed some other alternative approaches such as ANP, elimination and choice translating reality (ELECTRE), preference ranking organization method for enrichment evaluations (PROMETHEE) and the rough set theory.

Based on the review of the previous literature, various types of multiple-criteria decision-making methods, as well as hybrid methods of MCDM, has been widely used to assist decision-makers regarding ranking criteria of relevant alternatives in different situations. However, only a few research studies have applied hybrid methods of fuzzy ANP and VIKOR to RL adoption problems in the palm oil industry. Therefore, the context of the palm oil industry in Thailand is still unexplored. To fill this gap, we aim to study the hybrid methods of fuzzy ANP and VIKOR in the aspect of RL practice and implementation problems in Thailand’s palm oil industry as presented in Sect. 5.

3 Reverse logistics barriers and solutions in Thailand palm oil industry

3.1 Reverse logistics barriers practices

In developed countries, RL is more commonly implemented as a result of the enforcement of the laws and regulations requiring producers to conduct product take-back programs to recover the value of end-of-life products. The adoption of RL in Thailand’s palm oil industry is challenging due to the lack of government support, insufficient knowledge in reverse logistics practices, as well as, the markets being price sensitive. In Thailand, product return management is often viewed as an extra cost of a business and is usually conducted in an unorganized manner. Successful RL implementation requires technical and financial support from the government along with coordination and cooperation from supply chain partners. Critical analysis of the RL barriers can provide information for decision-makers so that they can utilize their plans effectively. Several supporting studies have investigated the significant barriers to RL practices adoption to provide guidelines to successful implementation of RL practices. In this study, the barriers found in the perspective of Thailand palm oil industry are classified into five criteria along with their sub-criteria. The identification of barriers is performed through the literature review and through discussion with the decision-making groups and experts who have proficiency in this field. The barriers identified in this study are outlined in Table 1.

Table 1 Barriers of RL implementation along with criteria and sub criteria

3.2 Solutions to reverse logistics practices

A variety of situation-based solutions to overcome the barriers of RL practices adoption have been proposed by academicians, researchers, consultants, experts and palm oil industry associates (see Table 2).

Table 2 Solutions of RL practices

4 Research methodology

In this study, a three-phase methodology is applied to prioritize the solutions that can be utilized to overcome the barriers that might face the adoption of RL practices. A fuzzy ANP is used to obtain the weights of the criteria of the barriers and VIKOR is used to prioritize the solutions of RL practices adoption. Although fuzzy ANP can be used for decision-making by itself, it will become more efficient when integrated with other decision support tools in multi-criteria decision-making process. The fuzzy framework is very beneficial for handling problems with impreciseness and uncertainty. The proposed three-phase methodology for ranking the solutions of RL practices is shown in Fig. 1.

Fig. 1

Proposed three-phase methodology for ranking solutions of RL practices

4.1 Phase 1: identification of RL barriers and solutions

In this phase, two decision-making groups have been formed. The first group includes experts, researchers, industry people (senior managers and logistics personnel), industry associates and customers. Initial analysis refers to the first decision-making group identifying and evaluating RL adoption barriers and analysis from the literature review. The latter half of the analysis focusing on solution identification and evaluation refers to another decision-making group, which includes RL and supply chain experts, industry people, and industry associates.

4.2 Phase 2: fuzzy ANP

Fuzzy set theory offers the capacity to solve problems that involve the uncertainty resulting from human cognition and thought processes (Chen and Chang 2011; Chen et al. 2012; Wang and Chen 2008, 2009; Chen 1998). It was developed by Azizzadeh et al. (1965) and has provided a means of classifying objects through a membership continuum, in which a membership level is allocated to each object, at a value between zero and one. A fuzzy number can be defined as a fuzzy set $\tilde {A}=\{ (x,{\mu _A}(x)),x \in R\}$, in which x represents real line values, R: − ∞ < x < + ∞ and ${\mu _A}(x)$ is formed by continuous mapping between R and the closed interval [0, 1]. Fuzzy logic has three principal membership functions, namely triangular, monotonic, and trapezoidal (Chatterjee et al. 2015). In fuzzy environments, the application of triangular fuzzy numbers (TFN), shown as (l, m, u), are usually the most effective model for the processing of information as a result of their computational simplicity (Tang 2009). Expert evaluations offered in linguistic formats that are used in MCDM studies can be analyzed using TFNs and linear presentation of each TFN is shown below:

$${\mu _A}(x)=\left\{ {\begin{array}{*{20}{l}} {0,}&{x<l,} \\ {x - l/m - l,}&{l \leq x \leq m,} \\ {u - x/u - m,}&{m \leq x \leq u,} \\ {0,}&{x>u.} \end{array}} \right.$$

(1)

ANP was first introduced by Saaty (1996) to address the issues of real-world feedback and interdependence between the study criteria and the alternatives. It is a generalized form of AHP (Hsu et al. 2012). This technique can be used as an option to the hierarchy found in the original AHP through the implementation of a network that allows ANP to model the interactions among the decision elements to resolve complex non-linear problems. In this way, ANP allows for a systematic analysis of all types of interactions and is, therefore, a practical tool in MCDM situations (Uygun et al. 2015). ANP processes are outlined as follow:

Step 1 Performing pairwise comparisons using triangular fuzzy numbers.

Decision-makers are required to perform a number of pairwise comparisons with the responses using the TFNs shown in Table 3. The use of TFNs permits the relative significance of each paired item to be evaluated at the same level (Vinodh et al. 2011). This research employs triangular fuzzy numbers as indicators of comparison ratios, the format is that of the importance of element i over element j from the perspective of the decision maker k, shown as $\tilde {a}_{{ij}}^{k}$ where $\tilde {a}_{{ij}}^{k}=(1,1,1)$ if $i=j$, and $\tilde {a}_{{ij}}^{k}=(l_{{ij}}^{k},m_{{ij}}^{k},u_{{ij}}^{k})$ if $i \ne j$, for $i,j=1,2, \ldots ,n$, and $k=1,2, \ldots ,K$. The fuzzy judgment matrix for decision maker k, ${\tilde {A}_k}$ is given by:

Table 3 The membership function of the triangular fuzzy numbers for criteria and sub-criteria rating

$${\tilde {A}_k}=\{ \tilde {a}_{{ij}}^{k}\} =\left( {\begin{array}{*{20}{c}} {(1,1,1)}&{(l_{{12}}^{k},m_{{12}}^{k},u_{{12}}^{k})}& \cdots &{(l_{{1n}}^{k},m_{{1n}}^{k},u_{{1n}}^{k})} \\ {(l_{{21}}^{k},m_{{21}}^{k},u_{{21}}^{k})}&{(1,1,1)}& \cdots &{(l_{{2n}}^{k},m_{{2n}}^{k},u_{{2n}}^{k})} \\ \vdots & \vdots &{}& \vdots \\ {(l_{{n1}}^{k},m_{{n1}}^{k},u_{{n1}}^{k})}&{(l_{{n2}}^{k},m_{{n2}}^{k},u_{{n2}}^{k})}& \cdots &{(1,1,1)} \end{array}} \right)$$

(2)

for $k=1,2, \ldots ,K$, where n denotes the number of related elements for the level, and $\tilde {a}_{{ij}}^{k}=(l_{{ij}}^{k},m_{{ij}}^{k},u_{{ij}}^{k})=(1/u_{{ij}}^{k},1/m_{{ij}}^{k},1/l_{{ij}}^{k})$.

Step 2 Computation of consistency and aggregation of the judgment matrices.

The purpose of consistency analysis is to ensure the reliability and accuracy of the judgment results. According to Buckley (1985), the fuzzy matrix ${\tilde {A}_k}$ in Eq. (2) is deemed consistent if $\tilde {a}_{{ij}}^{k} \approx \tilde {a}_{{ij}}^{k} \otimes \tilde {a}_{{ij}}^{k}$, for $i$, $j$ = 1, 2…, n, whereby $\approx$ is fuzzy equal to, and $\otimes$ denotes fuzzy number multiplication defined by,

$$({l_1},{m_1},{u_1}) \otimes ({l_2},{m_2},{u_2})=({l_1}{l_2},{m_1}{m_2},{u_1}{u_2}).$$

If the judgment matrix fails the consistency test, the decision-makers must redo the pairwise comparisons until all of the fuzzy judgment matrices compiled by the experts are proven consistent. This is to certify that the construction of the aggregate fuzzy judgment matrix is acceptable. For all groups of decision-makers, the aggregate fuzzy judgment matrix will be:

$$\tilde {A}={\{ {\tilde {a}_{ij}}\} _{n \times n}};\quad {\tilde {a}_{ij}}=({l_{ij}},{m_{ij}},{u_{ij}}),$$

(3)

where ${l_{ij}}$, ${m_{ij}}$ and ${u_{ij}}$ are calculated using the following (Chang et al. 2009);

$${l_{ij}}=\mathop {\hbox{min} }\limits_{k} (l_{{ij}}^{k}),\quad {m_{ij}}=\mathop {\hbox{min} }\limits_{k} (m_{{ij}}^{k}),\quad {u_{ij}}=\mathop {\hbox{min} }\limits_{k} (u_{{ij}}^{k}).$$

(4)

Step 3 Using Chang’s extent analysis approach for local priority.

Chang’s technique of extent analysis (Chang 1996) is used in this research study to assess the weighting of attributes for decision-making problems involving multiple attributes. The local priority can then be calculated using the aggregate fuzzy judgment matrix from Eq. (3). The following steps are included in Chang’s extent analysis technique:

1. The fuzzy synthetic extent value in terms of the element (i = 1, 2…, n), can be defined as:

$$\begin{aligned} {{\tilde {S}}_i} & \equiv ({l_i},{m_i},{u_i})={\sum\limits_{{j=1}}^{n} {{{\tilde {a}}_{ij}} \otimes \left[ {\sum\limits_{{i=1}}^{n} {\sum\limits_{{j=1}}^{n} {{{\tilde {a}}_{ij}}} } } \right]} ^{ - 1}} \\ & \approx \left( {\frac{{\sum\nolimits_{{j=1}}^{n} {{l_{ij}}} }}{{\sum\nolimits_{{i=1}}^{n} {\sum\nolimits_{{j=1}}^{n} {{u_{ij}}} } }},\frac{{\sum\nolimits_{{j=1}}^{n} {{m_{ij}}} }}{{\sum\nolimits_{{i=1}}^{n} {\sum\nolimits_{{j=1}}^{n} {{m_{ij}}} } }},\frac{{\sum\nolimits_{{j=1}}^{n} {{u_{ij}}} }}{{\sum\nolimits_{{i=1}}^{n} {\sum\nolimits_{{j=1}}^{n} {{l_{ij}}} } }}} \right) \\ \end{aligned}$$

(5)

where ${l_{ij}}$, ${m_{ij}}$ and ${u_{ij}}$ are given by Eq. (4).

2. The degree of possibility of${\tilde {S}_i} \equiv ({l_i},{m_i},{u_i}) \geq {\tilde {S}_j} \equiv ({l_j},{m_j},{u_j})$ for two elements for $i$ and $j$ is defined as Eq. (6)

$$V({\tilde {S}_i} \geq {\tilde {S}_j})=\left\{ {\begin{array}{*{20}{l}} {1,}&{{\text{if}}\;{m_i} \geq {m_j}} \\ {0,}&{{\text{if}}\;{l_j} \geq {u_i}} \\ {\frac{{{u_i} - {l_j}}}{{({u_i} - {m_i})+({m_j} - {l_j})}}}&{{\text{otherwise}}} \end{array}.} \right.$$

(6)

As shown in Fig. 2, the value of $V({\tilde {S}_i} \geq {\tilde {S}_j})$ indicates the greatest intersection point ordinate between any pair of fuzzy membership functions.

Fig. 2

Degree of a possibility of ${\tilde {S}_i} \equiv ({l_i},{m_i},{u_i}) \geq {\tilde {S}_j} \equiv ({l_j},{m_j},{u_j})$

Calculate the degree of possibility for ${\tilde {S}_i}$ to exceed all other n − 1 TFN, so that,

$$\begin{aligned} d^{\prime}({{\tilde {S}}_i}) & \equiv V({{\tilde {S}}_i} \geq {{\tilde {S}}_1}, \ldots ,{{\tilde {S}}_{i - 1}},{{\tilde {S}}_{i+1}}, \ldots ,{{\tilde {S}}_n}) \\ & =\mathop {\hbox{min} }\limits_{j} V({{\tilde {S}}_i} \geq {{\tilde {S}}_j}),\;{\text{for}}\;j=1,2, \ldots ,n,j \ne i. \\ \end{aligned}$$

(7)

3. Obtain the normalized priority vector $W={({w_1},{w_2}, \ldots ,{w_n})^{\text{T}}}$ of the fuzzy judgment matrix $\tilde {A}$ where ${w_i}$ is given by:

$${w_i}=\frac{{d^{\prime}({{\tilde {S}}_i})}}{{\sum\nolimits_{{j=1}}^{n} {d^{\prime}({{\tilde {S}}_j})} }},\quad i,{\text{ }}j={\text{ }}1,{\text{ }}2 \ldots ,n.$$

(8)

Step 4 The development and analysis of the supermatrix and the limit supermatrix.

The various process elements will display interdependence effects that can be represented by the supermatrix, which can describe three different types of relationships: (1) independence from the subsequent criteria and sub-criteria, (2) interdependence between the criteria and sub-criteria levels, and (3) interdependence among the criteria and sub-criteria themselves (Vinodh et al. 2011). After computing the priority weights of all criteria and sub criteria (in Step 3), we compose the unweighted supermatrix. The supermatrix is a multi-block matrix in which each block depicts the relationship between two nodes in the network. Each column in every block is a priority weight vector is entered into supermatrix according to the proper flow of influence among nodes. If there is no influence between two elements, the corresponding weight in the supermatrix will be zero. By normalizing the unweighted supermatrix, the weighted supermatrix is obtained in which the sum of all columns is unity. The weighted supermatrix is powered by 2k + 1, where k is an arbitrarily large number. When there is very small change in the arrays of supermatrix by sequent powers, then a convergence on the importance weights is achieved and this new matrix is the limit supermatrix. The relative weights can be derived from this limit supermatrix (Shafiee 2015).

4.3 Phase 3: ranking solutions for reverse logistics using the VIKOR

VIKOR technique was formulated by Opricovic (1998) and is based on the notion of using compromise to assess the different standards of various competing projects. It is possible to use the MCDM model along with VIKOR to provide a ranking system for the alternatives (Opricovic and Tzeng 2004). The approach works by finding the ‘positive-ideal’ solution (or so called the aspired level) and also the ‘negative-ideal’ solution (or so called the worst level). The technique employed to modify the ANP matrix using VIKOR is explained as follows in the case where the alternatives are given by ${S^1},{S^2}, \ldots ,{S^k}, \ldots {S^m}$, the performance scores for alternative ${S^k}$ in terms of the jth criteria are given by ${f_{kj}}$; ${w_j}$ denotes the influential weight, or relative importance, of the jth criterion, such that $j=1,2, \ldots ,n$ where $n$ indicates the number of criteria. The process of VIKOR development commenced from the initial form of ${L_p}$-metric (Liou et al. 2011).

$$L_{k}^{p}={\left\{ {\sum\limits_{{j=1}}^{n} {\left[ {\frac{{{w_j}\left( {\left| {f_{j}^{*} - {f_{kj}}} \right|} \right)}}{{\left( {\left| {f_{j}^{*} - f_{j}^{ - }} \right|} \right)}}} \right]} } \right\}^{1/p}}$$

(9)

in which $1 \leq p \leq \infty$; $k=1,2, \ldots ,m$ while the influential weight ${w_j}$ is obtained from the ANP. Subsequently, $L_{k}^{{p=1}}$ (as ${S_k}$) and$L_{k}^{{p=\infty }}$ (as ${Q_k}$) are utilized by VIKOR for the creation of the ranking and the gap measure (Huang et al. 2009; Liou et al. 2011; Opricovic and Tzeng 2007; Serafim and Gwohshiung 2002):

$${S_k}=L_{k}^{{p=1}}=\sum\limits_{{j=1}}^{n} {\left[ {\frac{{{w_j}\left( {\left| {f_{j}^{*} - {f_{kj}}} \right|} \right)}}{{\left( {\left| {f_{j}^{*} - f_{j}^{ - }} \right|} \right)}}} \right]}$$

(10)

$${Q_k}=L_{k}^{{p=\infty }}=\mathop {\hbox{max} }\limits_{j} \left\{ {\frac{{\left( {\left| {f_{j}^{*} - {f_{kj}}} \right|} \right)}}{{\left( {\left| {f_{j}^{*} - f_{j}^{ - }} \right|} \right)}}\quad j=1,2, \ldots ,n} \right\}.$$

(11)

In this case, ${\hbox{min} _k}L_{k}^{p}$ denotes the compromise solution indicating the synthesized gap, which needs to be minimized and will be chosen so that the value most closely approaches the aspired level. Furthermore, when p is small, such as a p value of 1, the group utility becomes the focus, but in contrast, if the p value tends towards infinity, greater importance in prior improvement is placed upon the individual maximal regrets or gaps for each criterion or dimension. As a result, ${\hbox{min} _k}{S_k}$ places stress upon the maximum group utility; in contrast, ${\hbox{min} _k}{Q_k}$ emphasizes the selection of the minimum from among all the maximum individual regrets or gaps for the priority improvements shown. There are four steps which form the process of using the VIKOR compromise ranking algorithm which are based on the criteria outlined above.

Step 1 Determine a level which is aspired or tolerable. The best $f_{j}^{*}$ values (or aspired level) are first of all calculated along with the worst $f_{j}^{ - }$ values (or tolerable level) for each of the functions of the criteria$j=1,2, \ldots ,n$. For example, if the jth function represents benefits:$f_{j}^{*}={\hbox{max} _k}\,{f_{kj}}$ and $f_{j}^{ - }={\hbox{max} _k}\,{f_{kj}}$then it is possible for decision-makers to set these values, so that $f_{j}^{*}$ is the aspired level and $f_{j}^{ - }$ is the tolerable level. Moreover, the original rating matrix can be converted to a normalized weight-rating matrix using Eq. (12)

$${r_{kj}}=\frac{{\left( {\left| {f_{j}^{*} - {f_{kj}}} \right|} \right)}}{{\left( {\left| {f_{j}^{*} - f_{j}^{ - }} \right|} \right)}}.$$

(12)

Step 2 Calculating the group utility mean and the maximal regret. The calculations of these values are performed by Eq. (13)

$${S_k}=\sum\limits_{{j=1}}^{n} {{w_j}{r_{kj}}} \;\;\left( {{\text{the}}\;{\text{synthesized}}\;{\text{gap}}\;{\text{for}}\;{\text{all}}\;{\text{criteria}}} \right),$$

(13)

$${R_k}=\mathop {\hbox{max} }\limits_{j} \{ {r_{kj}}\left| {j=1,2, \ldots ,n} \right.\}$$

(14)

which denotes the maximal gap for the k criterion in the case of priority enhancement.

Step 3 Calculate the index value. The index value is calculated by

$${Q_k}=v\frac{{({S_k} - {S^*})}}{{({S^ - } - {S^*})}}+(1 - v)\frac{{({R_k} - {R^*})}}{{({R^ - } - {R^*})}}$$

(15)

in which $k=1,2, \ldots ,m$, ${S^*}=\hbox{min}\,{S_k}\;$ “or” ${S^*}=0$, ${S^ - }=\hbox{max}\, {S_k}\;(or{\kern 1pt} \;{S^ - }=1)$ and ${R^*}={\hbox{min} _i}\,{R_i}\;$ “or” setting ${R^*}=0$ and ${R^ - }={\hbox{max} _i}\,{R_i}$ or setting ${R^ - }=1$, Eq. (15) can be written as ${Q_k}=v{S_k}+(1 - v){R_k}$ in which $v$ represents the weighting for the maximum group utility strategy. In contrast, the weight of individual regret is $(1 - v)$.

Step 4 Rank or improve the alternatives for a compromise solution

The alternatives must be placed in a decreasing order using the values ${S_k}$, ${Q_k}$ and${R_k}$. An acceptable solution can then be offered using the alternatives ${S^1},{S^2}, \ldots ,{S^m}$.

5 An application of the proposed method to case analysis

A three-phase methodology is employed in this study to prioritize and rank the barriers and the solutions related to RL practices in the palm oil industry of Thailand. We conduct a case study to test the developed model. For confidentiality, we will maintain the anonymity of the company, as such, we assign “ABC” as the name of the company. The methodology is explained as follow:

5.1 Phase 1: identification of RL barriers and the solutions for these barriers

Selected decision-makers included 10 experts, who comprise three senior managers (supply planning, logistics department, and commercial department), three operational managers, and four senior managers (marketing, finance, quality, and purchasing). In this research, 5 criteria and 19 sub-criteria are applied (see Table 1). Similarly, five experts (RL and supply chain senior members) suggested 10 solutions to overcome these barriers determined by the literature review and a group discussion (see Table 2).

5.2 Phase 2: calculate the weight of barriers of RL practices using fuzzy ANP

The decision group makes pairwise comparisons of the 5 criteria and 19 sub-criteria to assign weightings to the criteria. For example, it could be determined that the “management barriers (MB)” category is more important than “legal barriers (LB)”, and an expert may rate this judgment as “very strongly more important”. This can then be related to the TFNs as seen in Table 3 and applied through Eq. (2), as shown in Table 4.

Table 4 One of the experts’ pairwise comparison matrix of the main criteria

The consistency test calculations are conducted and the aggregate fuzzy judgment matrix is calculated using Eqs. (3) and (4), respectively. Next, Chang’s extent analysis is applied to determine the priority weights. Based on the aggregate fuzzy judgment calculations previously made, the priority weights of the criteria and sub-criteria are established. Equation (5) is then used to generate the fuzzy synthetic extent values. Equation (6) is employed to determine the degree of possibility, whereas Eq. (7) finds the non-normalized weights of the five criteria. Following this, Eq. (8) is applied to the calculation of the normalized priority weights. The comparison matrix for the main criteria and Chang’s priority weights calculated using MATLAB are listed in Table 5.

Table 5 The main criteria and the final Chang’s priority weights

Furthermore, the same process of weight calculation, the amount of interdependency among sub-criteria, and the amount of sub-criteria impacts on each other was measured and then placed in the correct columns based on their paired comparisons to create an unweighted supermatrix, as can be seen in Table 6. This unweighted supermatrix is then normalized to transform it into the weighted supermatrix, in which the sum of each column is equal to one. Subsequently, to obtain the final weights of the evaluation criteria, the weighted supermatrix is raised to the power of 2k + 1 until the values of each column are stabilized and equal. Finally, the limit supermatrix is compiled as shown in Table 7.

Table 6 (Unweighted) supermatrix for the RL barriers criteria

Table 7 The stable matrix of ANP when the power of 2k + 1

5.3 Phase 3: ranking the solutions for RL practices using VIKOR

The ten solutions are identified hereafter as suppliers ${S^1},{S^2}, \ldots ,{S^{10}}$. The performance of the solutions was assessed by seeking out the opinions of specialists with expertise in Thailand palm oil sector. A rating scale of 0–4 was used for the evaluation, where 0 represents very poor performance and 4 denotes excellence. The mean scores were then taken for each of the ten solutions, and the VIKOR technique was employed to determine the indices for ranking ${S_k}$, ${Q_k}$ and ${R_k}$ by applying Eqs. (13)–(15), and the ${r_{kj}}$ for each of the solutions in the five criteria was calculated using Eq. (12), as seen in Table 8. To illustrate the various considerations of ${Q_k}$, v = 0.7 was set to compare the results with ${S_k}$ which provides decision-makers with an additional option for consideration of group utility and individual regret when making comparisons.

Table 8 The final ranking of the solutions for RL practices

According to Table 8, it is easy to know that the total gap of solution S¹⁰ is 0.307 (when v = 0.7), which is better than the other nine solutions.

6 Result and discussions

The hybrid fuzzy ANP and VIKOR methods resulted in a highly systematic process, which was helpful for the decision-makers in the selection of the best choices regarding the RL practices barriers and solutions and to solve the problems through the processes of prioritizing and ranking based on the comparisons to determine which of the barriers are the most significant.

This approach was applied within the context of the RL practices implementation in the palm oil industry in Thailand to enhance and develop the implementation of RL practices in the palm oil business supply chains. The results also provide an awareness of the benefits of RL practices to the business operations and the reduction of environmental impacts. The order of weight values begins with the most significant barriers to RL practices, which were shown to be MB ≻ LB ≻ TB ≻ MRB ≻ IB, as listed in Table 5. It can be seen that management barriers are the most significant with regard to RL practices implementation. The ranking values of the management barriers are MB4 ≻ MB1 ≻ MB5 ≻ MB3 ≻ MB2 (see Table 7), in which “RL is not integrated with the supply chain business processes” is the barrier with the highest weight value, and “lack of awareness regarding RL” is the barrier with the lowest weight value of all of the management barriers.

The ranking values of the legal barriers are LB2 ≻ LB1 ≻ LB3 ≻ LB4, in which “loopholes in laws and regulations on waste management in Thailand” is the barrier with the highest weight value, and “minimal public focus on environmental issues” is the barrier with the lowest weight value among all of the legal barriers. Similarly, the ranking values of the technological barriers are TB3 ≻ TB1 ≻ TB2, in which “lack of flexibility with regard to changing from the traditional system to new ones” is the barrier with the highest weight value, and “lack of technological infrastructure needed to adopt RL” is the barrier with the lowest weight value of all of the technological barriers.

The ranking values of market-related barriers are MRB1 ≻ MRB3 ≻ MRB2, in which “uncertain returns & demand” is the barrier with the highest weight values, and “uncertain quality & quantity of returns” is the barrier with the lowest weight value of all of the market-related barriers. Finally, the ranking values of the infrastructural barriers are IB3 ≻ IB2 ≻ IB4 ≻ IB1, in which “lack of coordination with logistics partners” is the barrier with the highest weight value, and “lack of infrastructure facilities” is the barrier with the lowest weight value of all of the infrastructural barriers.

According to the $S$ values, the rankings of the solutions for RL practices in descending order are S¹⁰ ≻ S⁵ ≻ S² ≻ S⁴ ≻ S⁶ ≻ S⁹ ≻ S³ ≻ S⁷ ≻ S¹ ≻ S⁸. The rankings of the solutions for RL practices based on the $R$ values are S¹⁰ ≻ S² ≻ S⁴ ≻ S⁵ ≻ S³ ≻ S⁶ ≻ S¹ ≻ S⁷ ≻ S⁸ ≻ S⁹. According to the crisp $Q$ (v = 0.7) index values, the final rankings of the solutions for RL practices are S¹⁰ ≻ S⁵ ≻ S² ≻ S⁴ ≻ S⁶ ≻ S⁹ ≻ S³ ≻ S⁷ ≻ S¹ ≻ S⁸ (see Table 8). The results indicate that the S¹⁰, “Create top management awareness and support”, is the best alternative from the solutions and S⁸, “Establish strategies for outsourcing to third parties for EOL”, is the lowest ranking of the solutions for all of the $S$, $R$ and $Q$ index values. Therefore, the decision-makers and other stakeholders in Thailand’s palm oil industry should focus on the rankings of both the barriers to RL practices and the solutions for RL practices implementation to find most appropriate guidelines for solving certain problems.

6.1 Sensitivity analysis

The sensitivity analysis was conducted to determine the effects on the evaluation process and ranking of the solutions for RL adoption by variations in the priorities. It is suggested that slight changes in relative weight will result in major changes in the final rankings. As human judgment input is utilized to calculate the weights for the listed categories of solutions for RL, it is thereby recommended that the variation in the final ranking is assessed by assigning different weights to check the consistency and the experts’ influence in the decision-making process.

The results of sensitivity analysis show that S¹⁰ has the maximum priority under all of the conditions; when v varies from 0.1 to 1.0, the final rankings were given as S¹⁰ ≻ S⁵ ≻ S² ≻ S⁴ ≻ S⁶ ≻ S⁹ ≻ S³ ≻ S⁷ ≻ S¹ ≻ S⁸ (see in Table 9). It implies that “Create top management awareness and support” (S¹⁰) and “Determine RL as part of a sustainability program” (S⁵) are the selected preferences of the decision-makers (see Fig. 3) indicating that the proposed framework is robust and there was only insignificant influence of the ratings given by the experts.

Table 9 Ranking of the solutions for RL practices in sensitivity runs

Fig. 3

Results of the sensitive analysis

6.2 Comparisons with other existing methods

To verify the feasibility and validity of the proposed method, we provide a comparative analysis on the performance including fuzzy ANP-VIKOR method and fuzzy Decision-making Trial and Evaluation Laboratory (DEMATEL) method.

According to the fuzzy DEMATEL method (Wu and Lee 2007; Lin and Tzeng 2009) the generalized direct-relation fuzzy matrix can be obtained by normalizing the aggregated direct-relation matrix. Then compute the total influence matrix $\mathop T\limits^{\sim }$ can be used to obtain the total influence matrix, where I denote the identity matrix defined below:

$$z=\frac{1}{{{{\hbox{max} }_{1 \leq i \leq n}}\sum\nolimits_{j}^{n} {{{\tilde {a}}_{ij}}} }}\;{\text{where}}\;z>0$$

(16)

$$\mathop T\limits^{\sim } =\mathop {\lim }\limits_{{k \to \infty }} (\tilde {X}+{\tilde {X}_2}+ \cdots +{\tilde {X}_h})=X{(I - X)^{ - 1}}.$$

(17)

Then using the total influence matrix $\tilde {T}$ get to the total row $(\tilde {r})$ and column $(\tilde {s})$, which is shown as follows:

$$\tilde {r}={[{\tilde {r}_i}]_{n \times 1}}={\left[ {\sum\limits_{{j=1}}^{n} {{{\tilde {t}}_{ij}}} } \right]_{n \times 1}},$$

(18)

$$\mathop s\limits^{\sim } ={[\mathop {{s_j}}\limits^{\sim } ]_{n \times 1}}={\left[ {\sum\limits_{{i=1}}^{n} {\mathop {{t_{ij}}}\limits^{\sim } } } \right]_{{\kern 1pt} 1 \times n}},$$

(19)

where $\tilde {T}=[{\tilde {t}_{ij}}]$, $i,\,j=1,2, \ldots ,n$.

In addition, $\tilde {r}$ shows the sum of the direct and indirect effects of factor i on other dimensions/criteria if it denotes the sum of the ith row in matrix $\mathop T\limits^{\sim }$. Conversely, $\tilde {s}$ represents the sum of the direct and indirect effects that factor j received from other factors if it denotes the sum of the jth column of matrix $\mathop T\limits^{\sim }$. Furthermore, $(\tilde {r}+\tilde {s})$ provides an index of the strength of influences that are given and received when $i=j$ (i.e., sum of row and column aggregates). Specifically, $(\tilde {r} - \tilde {s})$ refers to the degree of the role played by factor i in the given problem. Table 10 shows the results and it is very clear that the ranking order of solutions obtained by fuzzy DEMATEL method is S¹⁰ ≻ S⁵ ≻ S² ≻ S⁴ ≻ S⁶ ≻ S⁹ ≻ S³ ≻ S⁷ ≻ S¹ ≻ S⁸.

Table 10 Values $(\tilde {r}+\tilde {s})$ and $(\tilde {r} - \tilde {s})$

According to the previous comparison analysis, using fuzzy ANP-VIKOR method, we can obtain the same results with other two methods. We found that the merits of the proposed method, it considers the uncertainty of evaluations the solutions, which can simulate actual situation better. In the process of the priority weight of the criteria, the ANP method is extracted by considering the impact of each criterion on every other criterion using comparison matrices. Moreover, it deals with mixed decision information including the linguistic terms can solve the complex problem with the different kinds of evaluation information without information losing, which can be yield more reasonable result. Finally, the VIKOR method can determine a compromise solution using the utility weight of each solution, the attitude of experts can be reflected by adjusting utility weight. Hence, above mention a comparative analysis, our proposed approach can get the same result that the proposed method is more consistent with the reality and accurate.

7 Conclusion

The competitive environment is currently putting pressures on companies and forcing them to adopt RL practices for sustainable business operations. However, due to the presence of barriers, it is difficult to implement RL practices successfully. Thus, there is a need to provide the solution for these barriers. As it is quite challenging to implement all of the solutions simultaneously, it is therefore beneficial to prioritize a solution for proper implementation to surmount these obstacles. This study presents a robust multi-criteria decision-making method for prioritizing the solutions to resolve the problems resulting from the barriers to RL adoption. This has been performed through the identification of the barriers based on the review of the literature, industry experts and industry associates, and then linguistic ratings of the criteria being assigned by the decision-making team.

In this study, explores possible criteria of barriers under 19 barriers and selected optimized solutions using integrated Fuzzy ANP and VIKOR framework. Fuzzy ANP is used to calculate the relative weights of the barriers and can be extracted by considering the impact of each criterion on every other criterion using comparison matrices. Finally, VIKOR is applied to prioritize the solutions. The results indicate that “Create top management awareness and support” is the highest ranked solution in this study on RL adoption. The proposed framework is supported by an empirical case study of Thailand’s palm oil industry with regard to overcoming the barriers to its adoption of RL.

The further comparison with fuzzy DEMATEL methods, we can observe that the proposed method is capable solving MCDM in most effective and efficient way. We found that the merits of the proposed method, it considers the uncertainty of evaluations the solutions, which can simulate actual situation better. In the process of the priority weight of the criteria, the ANP method is extracted by considering the impact of each criterion on every other criterion using comparison matrices. Moreover, it deals with mixed decision information including the linguistic terms can solve the complex problem with the different kinds of evaluation information without information losing, which can be yield more reasonable result.

Moreover, the rankings of the solutions can function as guidelines that provide support to decision-makers and top management to determine policies and strategies for solving the problems caused by the barriers hindering RL practices implementation. Such information can be applied by involving organizations in Thailand’s palm oil industry to develop practical strategies regarding green supply chain management (GSCM). Beside, this model will help business analysis and supply chain managers formulate both short-term and long-term, flexible decision strategies for successfully managing and implementing RL adoption in the supply chain scenarios. In the future, this work can be extended by employing several approaches, such as fuzzy DEMATEL-ANP, ELECTRE, PROMETHEE and prospect theory.