Enhancing the cosmetics industry sustainability through a renewed sustainable supplier selection model

ABSTRACT The cosmetics industry requires a long-term sustainable strategy to balance its continuously growing trend worldwide and its resources consumption. In this view, the suppliers’ selection process is gaining more attention affecting products’ overall sustainability. The objective of this contribution is hence to develop and validate the Cosmetics Sustainable Supplier Selection (C-SSS) model allowing the selection of sustainable suppliers for the cosmetic industry, evaluating them in an objective and balanced manner. The model was built relying on both scientific and grey literature, by incorporating the characteristics of existing SSS models usually used separately. The C-SSS enabled to integrate the EMM approach (to reduce the subjectivity), the ANP approach (to evaluate criteria interconnections), and the TOPSIS and ELECTRE models (to create a hybrid compensation model) to support managers in objectively selecting the most sustainable suppliers. The C-SSS model was applied and validated through an industrial use case in a cosmetics Italian company.


Introduction
Nowadays, the sustainability paradigm, which is based on the environmental, social, and economic pillars (i.e. the Triple Bottom Line (TBL)) (Elkington, 2013), is spreading all over the society, becoming a strategic element in governments and companies' plans due to the unprecedent uncontrolled resources usage (United Nations, 2015).In particular, the manufacturing sector is moving towards circular-oriented strategies influencing product design, production, distribution, and consumption (Acerbi, Taisch, 2020;Bjørnbet et al., 2021).Among all, due to its ever-growing size and massive consumption of natural resources, the cosmetics industry is required to update its current operations to face new sustainable requirements and norms (Pereira de Carvalho & Barbieri, 2012;Rocca et al., 2022).For instance, fair-trade labels and sustainability indexes, such as those referred to Corporate Social Responsibility, are introduced to strengthen the relationships with several stakeholders, among which final users, with the goal to clarify the sustainable-oriented characteristics of a certain product (Bom et al., 2019) and facilitate the stakeholders' engagement in the embracement of sustainability pillars (Gong et al., 2019).Indeed, the cooperation with industrial entities, external to the company, and also with final consumers are fundamental to undertake a sustainable and circular-oriented path (Santa-Maria et al., 2021) independently from the industry in which the company operates.The growing attention to sustainability pillars (i.e.TBL) in the cosmetics industry (Ambak et al., 2019;Amberg & Fogarassy, 2019) is hence becoming fundamental.Cosmetic products are increasingly spreading in our daily life activities, and consumers are becoming keener on purchasing green products, as environmental and ethical considerations are increasingly relevant factors in their purchase behaviour (Appolloni et al., 2022).According to Bom et al. (2019), 'cosmetics sustainability' can be defined 'as a complex and multi-faceted issue that cannot be evaluated considering single aspects, but using an integrated assessment about the environmental, social and economic dimensions and about the final product quality and performance'.The increasing importance of the cosmetics industry is visible from an economic perspective too; its value has been estimated at €76.7 billion at retail sales price in 2020 in Europe ('Cosmetics Europe -The Personal Care Association: Cosmetics Industry,' 2021).Also, for the job market, the cosmetics industry represents for Europe a great player, accounting, in 2019, over 2 million jobs, out of which 167,730 were employed for the manufacturing of products ('Socio-Economic Contribution of the European Cosmetics Industry ', 2019).The growing trend of this industry in the global market is evident having a steady increment in the last five years with a pick of the 5,5% in 2018 (Cosmetic industry growth | Statista, 2020).Therefore, considering, on one hand, the increasing growth of the cosmetics industry and, on the other hand, the necessity to start undertaking a path towards more sustainable systems, it is required by cosmetics manufacturing companies to engage in value chains with sustainable-oriented stakeholders (Bom et al., 2020;Fonseca-Santos et al., 2015) and within this context to keep under control and assess the performances of the stakeholders across the entire value chain (Brown & Bajada, 2018).Therefore, starting from the design phase of products, which is responsible for most of the impacts of the subsequent product life cycle stages (Sassanelli et al., 2020), it is fundamental to choose accurately all the product characteristics among which the materials and thus the related suppliers (Acerbi, Taisch, 2020;Acerbi, Taisch, et al., 2020).Actually, in the extant literature emerged that the two key challenges for sustainability initiatives in value chain flexibility are the lack of suppliers' commitment to sustainable products and the lack of sustainable knowledge along the value chain (Dwivedi et al., 2021).Considering that suppliers might influence the sustainability performances of products and of the entire supply chain (Song et al., 2017) their involvement in the sustainability pathway of producers becomes essential.The scientific literature reports some models to properly select sustainable suppliers such as the one by (van Thanh & Lan, 2022) that is focused on the food industry or the one by (Wang et al., 2022) focused on the resiliency of the supply chain to ensure the supply of critical resources.However, there are still some open points not yet covered by them.For instance, it is still lacking a clear definition of criteria to evaluate suppliers specifically for the context of cosmetics companies under the sustainability umbrella; (Atthirawong, 2020) proposed a preliminary study on cosmetics and then, the research by (Tong et al., 2019) can be considered closed to cosmetics being it focused on chemical industry but there is not a comprehensive sustainable-oriented model focused on cosmetics.Moreover, existing Sustainable Supplier Selection (SSS) models usually cover the TBL pillars in an unbalanced manner (e.g.(Hashemi et al., 2015b), which is focused mainly on economic and environmental aspects).In addition, they may not take into account potential subjective judgements (e.g.(Liu et al., 2019)) or interdependences among criteria (e.g.(Amindoust, 2018)) or the possibility to have more than one decisionmaker (e.g.(Hendiani et al., 2020)).There is the need to develop a unique comprehensive model to cover the emerged literature gap and concretely support cosmetics companies in selecting the best sustainable suppliers in accordance with their strategic needs.For this reason, the research objective (RO) of this contribution is to overcome this open gap and develop a novel model allowing the selection of suppliers for cosmetic companies, evaluating them in a balanced and objective manner according to a set of criteria suggesting their sustainability-oriented performances.In doing that, the integration of already existing supplier selection models, which often employ a silos approach, covering the key above mentioned relevant characteristics and weaknesses have been conducted to finally have a complete model tailored on the cosmetics industry.An initial proposal, but neither comprehensive nor focused on cosmetics, was made by (Yu et al., 2019) that included in a single model the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) and the Elimination and Choice Translating Reality (ELECTRE) methods, although it assumed that the criteria are independent among each other.Further details will be given in the literature review and model development sections.
The remainder of this contribution is structured as follows.Section 2 reports the research methodology employed to address the RO.Section 3 elucidates the key criteria to be included for the SSS process tailored on cosmetics companies and it also analyses the already existing SSS models.Section 4 describes the novel model (i.e.Cosmetic Sustainable Supplier Selection model (C-SSS model)) developed after a comparison and selection of the already existing ones identified in the extant literature.Section 5 validates the model based on its application in an industrial case.Section 6 discusses the results highlighting the key practical and theoretical implications.Last, Section 7 concludes the contribution by also highlighting the key limitations opening towards future research opportunities.

Research Methodology
To address the RO, thus developing a comprehensive model for the SSS for cosmetics companies, it has been firstly conducted a review of the extant literature.This review was performed to ensure that we keep into account all the key sustainable-oriented criteria, thus covering the entire TBL, relevant for cosmetics companies and to analyze the already developed SSS models also in terms of structure.Then, the developed model has been applied and validated in a real industrial case.More in detail, the research process followed in this contribution is depicted in Figure 1 and explained below.
(1) A literature review was conducted to analyze especially the SSS models already available in the extant literature to identify the needed criteria and the characteristics of the existing SSS models.This review has been performed mainly on Scopus database, based on the following string: TITLE-ABS-KEY (('sustainabl*' OR 'circular' OR 'green') AND (('supplier*' OR 'vendor*') AND ('selection model*' OR 'choice model*'))) AND (LIMIT-TO (LANGUAGE, 'English')).
This string generated an outcome of 75 contributions, out of which 42 have been considered eligible to be reviewed being them focused on the manufacturing sector, on sustainability-related issues, on supplier selection models and not only on supply chain management.Actually, few contributions were focused specifically on cosmetics and only some of them concerned the chemical, pharmaceutical and food industries (e.g.(van Thanh & Lan, 2022)) which might have some characteristics adherent to the cosmetics one.Therefore, to ensure to grasp all the needed criteria, other documents also coming from grey literature have been reviewed (e.g. the International Organization for Standardization (ISO) standards).Moreover, through a snowball process and suggestions from experts, 40 contributions were added to the systematic literature review results.Out of these 40 contributions, 25 contributions were included in the review to extend the analysis about the cosmetic-related criteria to be assessed during the SSS process.
The other 15 contributions were added to enlarge the analysis over the already existing SSS models to ensure the evaluation of all the possible models' characteristics.This initial stage enabled to achieve the following tasks.
(1) Identification and selection of the criteria to be included in the C-SSS model.This step of the methodology relied on the analysis of contributions and documents available in different databases, thus not only Scopus.To ensure to consider all the criteria required in the SSS process for cosmetics companies (i.e.including norms, regulations, and standards), the analysis of the criteria already used in previous SSS models has been merged with the ISO standards developed for the industry, which were analysed through the Sustainable Development Goals (SDGs) lenses.
(2) Analysis of the eligible models available in the extant literature and selection of those coherent with the RO.A deep analysis of the already existing SSS models available in the extant literature has been performed.This enabled to evaluate their key strengths, to be included in the novel model, and their key weaknesses, to be overcome through the novel model, leading to the integration of different SSS models into a comprehensive one.
(2) Development of the novel C-SSS model.Based on the literature review findings, especially on those contributions tailored on the cosmetics industry (e.g.(Atthirawong, 2020)), a C-SSS model has been developed.As previously anticipated, the model has been conceptualized grounding on the literature findings, both scientific and grey literature, and it has been concretely built to be applied to an industrial use case.The concrete realization of the model was possible thanks to the integration of an excel file and an open-source software, 'SuperDecisions' (SuperDecisions, 2022).These two technical tools are simple to be used and affordable for any type of company while ensuring to cover all the steps required in selecting the proper supplier by a cosmetics manufacturer (all the steps are detailed in the model application section).
(3) Application and validation of the model.As previously anticipated, the model has been applied and validated in a Small and Medium Enterprise operating in the cosmetics industry localized in the north of Italy.
More specifically, the flow chart for the model conceptualization, development, and application is summarized in Figure 2 which shows the steps employed starting from the literature findings.Additional details for the selection of the approaches are reported in sections 3 and 4, summarized in Figure 6 and reported in sub-section 4.1.

Literature review
The increased need for sustainable-related accountability by cosmetics companies to address the requests coming from clients, public opinion, and governments with new

Model Development
The strengths, and the related models and approached, of the analysed models have been integrated in order to overcome the weaknesses observed.Moreover, a comprehensive list of sustainable-related criteria for manufacturing companies have been included in the C-SSS model.

Model Application
The C-SSS model developed has been applied in one industrial case to evaluate the benefits or the potential limitations observed by the company.norms and regulations, led to the need to create resilient and sustainable supply chains (Pereira de Carvalho & Barbieri, 2012).This need requires a careful and thoughtful selection of partners and suppliers (Giannakis et al., 2020).Thus, the SSS process emerged to be a core aspect for the creation of a sustainable cosmetics supply chain (Chai et al., 2013;Fortunati et al., 2020).This process can be defined as 'the process by which firms identify, evaluate, and contract with suppliers [. . .] to reduce purchase risk, maximize overall value to the purchaser, and develop closeness and long-term relationships between buyers and suppliers' (Taherdoost & Brard, 2019).
In general, the criteria for the selection of suppliers have changed over time moving from only quantitative indicators (e.g.price), to a mix of both qualitative and quantitative considerations (Tong et al., 2019).Currently, within these criteria, also the sustainabilityoriented ones need to be considered, giving a new name to the process: 'Sustainable Supplier Selection' (SSS) process defined as 'a comprehensive consideration of economic, environmental and social aspects in supplier selection process' (Yu et al., 2019).
In this regard, the analysis of the required criteria for an SSS process undertaken by a cosmetic company, together with an analysis of the existing SSS models present in the extant literature are elucidated below.

SSS Criteria in Cosmetics
This chapter aims to clarify the key sustainable-oriented criteria to be included in an SSS model for the cosmetics industry looking at both scientific literature (sub-section 3.1.1.)and ISO standards (sub-section 3.1.2).

Criteria from scientific literature
The extant literature, in addition to the standard criteria of price, quality, flexibility and delivery suggested, in most of the contributions (e.g.(Bai & Sarkis, 2010;Hashemi et al., 2015b)) highlighted a set of criteria which can be clustered into four key classes to be considered in the SSS process, especially while dealing with the cosmetic industry: (i) safety, (ii) environmental pollution, (iii) aquatic toxicity and (iv) sourcing and resource consumptions.A brief explanation of the four classes is described below, and Table 1 summarizes the most relevant criteria and sub-criteria identified for the already existing SSS models.
Safety concerns are among the most relevant criteria to be considered when looking at the ingredients and the chemicals used in cosmetics products.To evaluate product safety, manufacturers need to perform a risk assessment starting from product development until the marketing cycle (Engasser et al., 2007).Moreover, there are several standards worldwide to be addressed among which the 1938 Federal Food, Drug and Cosmetic Act and its amendments in the Unites States, and 1976 EU Cosmetics Directive and its revisions in the European Union (Engasser et al., 2007).Another important issue for the cosmetics industry is the animal testing (Sahota, 2014).Although these practices are required to prevent humans from negative consequences on safety issues of cosmetic products, concerns for animals' treatment and wellbeing arose and actions against these practices have been taken (Sreedhar et al., 2020).For instance, in Europe, the first action on the implementation of a ban on sales and import of cosmetics tested on animals was taken in 2013, followed by India in 2014 and New Zealand in 2015 (Sreedhar et al., 2020).This attention let the rising of vegan-friendly and cruelty-free cosmetics brands (InVitro International, 2019).In other countries, this attention is still limited, and no regulations are present although alternatives to animal testing are available (e.g. the Virtual Human Platform (VHP) (Cosmetic Europe, 2020)).
Environmental pollution is another criterion to be included in the analysis, considering that cosmetics products contain a wide range of chemicals and some of them are labeled as 'chemicals of emerging concern' (Dhanirama et al., 2012).In this sense, an important aspect to be considered is the use of microbeads in cosmetics formulations.Microbeads are solid microplastic spheres with small diameter (less than 5 mm) used in cosmetics products to perform for instance the skin exfoliation (Guerranti et al., 2019).Being them composed by plastic material that might be dispersed in the aquatic plants, the substitution with natural ingredients has been considered since they pose a threat to the wellbeing of the ecosystem and to the biodiversity loss (Guerranti et al., 2019).Actually, the use of plastic materials has impact on both the formulation and the packaging.Plastic represents hence one of the most diffused materials for the packaging because of its flexibility and lightness (Sahota, 2014).There are actions to start thinking at different packaging as the proposal by (Cosmetics Europe, 2019;Jaccarini & Refalo, 2017) of biobased packaging and the introduction of design for environment practices.
Aquatic toxicity is another important issue to be considered in this industry since after the use phase, several products enter the aquatic environment, and traditional methods  (Amindoust, 2018;Bai & Sarkis, 2010;Hashemi et al., 2015b;Hendiani et al., 2020;Kaur et al., 2020Kaur et al., , 2020;;Park et al., 2018;Rahmadani & Suparno, 2021;Sahota, 2014;Sreedhar et al., 2020;Tavassoli et al., 2020;Thanh & Lan, 2022;Tong et al., 2019;van Thanh & Lan, 2022;Wang et al., 2022) Environmental Pollution solid waste, and harmful materials released (Bai & Sarkis, 2010;Dhanirama et al., 2012;Fallahpour et al., 2021;Freeman & Chen, 2015;Guerranti et al., 2019;Hashemi et al., 2015b;Hendiani et al., 2020;Kaur et al., 2020Kaur et al., , 2020;;Khalilzadeh & Derikvand, 2018;Kuo et al., 2010;Li et al., 2021;Mohammadi et al., 2017;Rahmadani & Suparno, 2021;Sahota, 2014;Thanh & Lan, 2022;Tong et al., 2019;van Thanh & Lan, 2022;Wang et al., 2022) Aquatic Toxicity • Production of polluting agents • Average volume of wastewater (Bai & Sarkis, 2010;Bom et al., 2019;Hashemi et al., 2015b;Secchi et al., 2016;Vita et al., 2018;Wang et al., 2022) Sourcing and resource consumption (Amindoust, 2018;Bai & Sarkis, 2010;Bom et al., 2020;Fallahpour et al., 2021;Freeman & Chen, 2015;Hashemi et al., 2015b;Hendiani et al., 2020;Jaccarini & Refalo, 2017;Jayant & Paul, 2014;Kaur et al., 2020Kaur et al., , 2020;;Khalilzadeh & Derikvand, 2018;Kuo et al., 2010;Li et al., 2021;Mohammadi et al., 2017;Pulverail & Givaudan, 2013;Rahmadani & Suparno, 2021;Schneiders & Anklin, 2013;Secchi et al., 2016;Tavassoli et al., 2020;Thanh & Lan, 2022;van Thanh & Lan, 2022;Wang et al., 2022Wang et al., , 2022) ) of treatment can be used only to remove a limited amount of this waste (Vita et al., 2018).Among the most diffused ingredients representing a threat for the aquatic environment there are the parabens, UV filters in solar protection, antimicrobial and preservatives agents (Vita et al., 2018).Moreover, the petrochemical derivatives are other ingredients to be kept under control since they generate emissions having a non-renewable nature, and the silicones too should be considered being them bioaccumulate and thus posing threat to aquatic life (Bom et al., 2019).Another key aspect to be kept into account in the cosmetics industry is the quantity of water used in the production process that was reported to be one of the most relevant issues after a Life Cycle Assessment study conducted on a bio-based cosmetic product (Secchi et al., 2016).Additionally, considering the relevant role of the raw materials and ingredients of cosmetics products and of their packaging, also their sourcing and their resource consumptions represent important elements in cosmetics (Bom et al., 2020;Jaccarini & Refalo, 2017;Secchi et al., 2016).The ingredients used come from regions worldwide with different conditions and different resource consumptions.Therefore, it cannot be taken for granted that natural ingredients are more sustainable with respect to artificial or synthetic ones.Indeed, there are several issues concerning environmental and social aspects to be considered especially in the procurement phase when ingredients are selected (Bom et al., 2019).Some certifications have been introduced such as the fairtrade partnerships ensuring to operate in a social and environmental responsible manner (Schneiders & Anklin, 2013).In fact, traceability and transparency along the supply chain are becoming more and more relevant in this sense (Pulverail & Givaudan, 2013), also because the cosmetic value chains and more specifically the networks of suppliers are mainly global.This might bring some concerns such as great levels of uncertainty, hidden costs, and different national regulations (Manuj & Mentzer, 2008).For these reasons, sustainable suppliers' development practices have been put in place to support suppliers in improving their performances and capabilities according to the customer needs creating long-term relationships (Belotti Pedroso et al., 2021;Sancha et al., 2015).

Criteria from ISO linked to SDGs
As reported in the methodology section, the ISO standards specifically designed for the cosmetics industry have been investigated to evaluate the performances of a supplier in terms also of adherence to sustainable-oriented standards already in use in the industrial world.The ISOs were analysed in combination with the SDGs to highlight their sustainability-oriented value as reported in Table 2.

SSS models analysis
Considering the list of criteria to be kept into account in cosmetics, it is important to find a solution allowing to include all these criteria in a comprehensive assessment model by giving them the proper weight.In this regard, the extant literature presents some possibilities in terms of SSS models, even though the majority is not focused on the cosmetics industry.In Table 3 the extensive analysis is described by reporting the type of model adopted, the limitations or strengths highlighted by the authors, the industry in which the model was applied (if available), and the sustainable pillars considered in the criteria used in the model.
These models are all characterised by some strengths and weaknesses.For instance, those focused on cosmetics are not structured neither to capture in a balanced manner the whole panel of sustainability-oriented criteria nor to include more than one decision maker.For this reason, the existing models focused on other industries have been analysed to evaluate how to overcome their limits by combining their useful characteristics, thus integrating them into a unique and comprehensive model.Among all, the main criticalities emerged in the existing models are: (i) subjectivity, (ii) single decisionmaker perspective, (iii) no criteria interconnections and (iv) no cosmetics and sustainable-related criteria included.
The Multi Criteria Decision Models (MCDM) have emerged from the extant literature (Giannakis et al., 2020) as good tools ensuring a balanced evaluation of the potential     trade-offs among criteria.The evaluation is conducted based on experts' judgments and comparisons between criteria or alternatives (Deniz, 2020).Among all, the extant literature presents MCDM such as Analytic Hierarchy Process (AHP), Analytic Network Process (ANP) fuzzy-based approaches, Technique for Order of Preference by Similarity (TOPSIS) (Giannakis et al., 2020;Rashidi & Cullinane, 2019;Zeydan et al., 2011).In addition, to deal with the imprecise data conversion of uncertain and subjective information from the decision-makers into interval of fuzzy numbers, the Fuzzy Set Theory (FST) can be integrated into these models (Awasthi et al., 2018;Azadi et al., 2015;Memari et al., 2019;Rashidi & Cullinane, 2019;Yadavalli et al., 2019).Fuzzy techniques enable to reduce the subjectivity of the evaluation of the decision-makers, highly affecting the MCDM, by translating qualitative linguistic variables into quantitative numeric values (Zadeh, 1965).
Regarding the SSS models, another aspect to be considered is their compensatory nature.It would be possible in fact to compensate a bad performance of a certain criterion with a good performance in another one belonging to a different category as suggested by (Yu et al., 2019).In the sustainability context, a non-compensatory model would better work to ensure not lagging a certain category of criteria (Yu et al., 2019).At the same time, the compensatory model would enable to balance the different criteria and it is well known and diffused into companies.Therefore, hybrid models have been proposed to integrate both the natures, i.e. compensatory and non-compensatory, of the SSS models (Junior et al., 2013;Yu et al., 2019).For instance, it is possible to integrate the TOPSIS model, which has a compensatory nature model currently quite diffused, with the ELECTRE, a non-compensatory nature model, which is coherent with a sustainable oriented evaluation of the suppliers (Yu et al., 2019).Summarizing the main characteristics that necessitate to be considered in the development of a novel SSS model are the following: • The integration of multiple perspectives of decision-makers • The reduction of subjectivity of the decision-makers • The evaluation of the presence of interdependencies and interconnections among criteria • The model compensation nature.

Model conceptualization based on scientific findings
The novel C-SSS model relies on the findings from the extant literature about SSS models.More in detail, three key models were taken as references to build the novel one tailored on cosmetics industry.These three were chosen as main reference since they already validated part of the integration of some of the approaches which emerged in the literature as fundamental to address the weaknesses.Atthirawong (2020) developed a model focused on the green supplier selection process in the cosmetics industry.This model relies on the TOPSIS method which has a compensatory nature and is not properly useful for this context due to the high subjectivity given to the evaluation.Thus, a hybrid model proposed by Yu et al., (2019) has been considered to be inserted in the novel model since it includes both the TOPSIS and ELECTRE approaches backed by the Entropy Measured Method (EMM).This integrated solution allows to reduce the level of subjectivity by asking the decision makers to evaluate the suppliers' capability and not the weights.Moreover, still in accordance with the model proposed by (Yu et al., 2019), a quantitative and objective evaluation is required, and this can be performed thanks to the Fuzzy Time Series (FTS) that hence has been considered too in the developed model.The integrated models and approaches proposed by (Yu et al., 2019) are reported in Figure 3.
However, the model proposed by (Yu et al., 2019) has an important limit since the criteria considered by the model are considered as independent, while their interrelations are needed to be taken into account during the selection process of suppliers.For this reason, it was considered valuable the inclusion in the novel model of an ANP model.The ANP enables to keep into account the independencies present among criteria using this information as input for the TOPSIS model.Anyway, the results should be objective and balanced among them thus, both the EMM and the entropy-weighted ANP evaluation methods were used since they address both objective and subjective judgments as presented also by (Mei et al., 2016).Although this research does not concern the SSS process, it has been taken only as reference as a structured and validated way to link entropy and ANP results.In addition, to strengthen the potential links present between the EMM, the ANP and the TOPSIS model, another model was analysed, the one proposed by (Zhou et al., 2009) even though focused on other industry and topics.Therefore, to address this integration based on already validated models, the research proposed by (Zhou et al., 2009) was included as another grounding model since it considers the TOPSIS approach once the weights have been aggregated.Figure 4 shows the summary of the integrated analysis of (Mei et al., 2016) and (Zhou et al., 2009).
Last, to also include multiple actors in the evaluation phase and to lower the random error possibility made by individuals, it employed the approach used by Shyur & Shih (2006).This model applied the geometric mean to derive group preferences provided by multiple decision makers.Indeed, this model was selected since it already tested the integration between the ANP and the TOPSIS approaches with the geometric mean for the evaluation of a local Taiwanese company's suppliers.
As a result, the model proposed by the authors is an aggregation of the approaches and models applied and already validated by (Mei et al., 2016;Yu et al., 2019) and (Zhou et al., 2009) as depicted in Figure 5.
The final novel C-SSS model structure is reported below in Figure 6 where the specific steps and the related methodologies employed are reported.Immediately below the Figure 6, the steps are described in their details.(Mei et al., 2016) and (Zhou et al., 2009).
Therefore, the extant literature enabled to clarify the key characteristics and criteria to be included in a novel C-SSS model.Based on that, the key steps for the model development are summarized in Table 4.

Model application
To use the C-SSS model, both the suppliers and the company's decision-makers are involved, and two technical tools are employed to make the process smoother as reported in Table 5.Indeed, as previously described in the methodology section, these tools are easy to be applied in industrial cases being cheap and user-friendly.

Marco-phase 1
Starting with macro phase 1 (i.e.criteria selection and evaluation of a network), it is first required to select the decision makers, and make them select the proper criteria to be used for the selection of their companies' suppliers (see analysis in sub-section 3.1.1).The selection of a specific set of criteria (see Step 1 of Table 4) depends on the context of application and on the company's strategy.These criteria can be chosen among those identified from the scientific and grey literature.These criteria were revised and grouped in three main classes corresponding to the TBL pillars, as reported below in Table 6, to ensure the balanced consideration of all the three pillars.
All the criteria are translated into polar questions to ensure an easy adoption of the model by companies.These questions are sent to the suppliers to obtain all the information that need to be evaluated by the decision-makers through the developed model.Indeed, in Step 2 (see Table 4), the ANP model is introduced.This model enables to keep into account both inner and outer dependences, allowing to consider the interactions within and between groups of criteria (Jayant, 2016).The general structure of these models is based on two layers: the control and the network layers.These layers perform respectively the pairwise comparison and evaluate the influences present among criteria in the network (Jayant & Paul, 2014).The ANP implementation requires first to ask the (Mei, Ye and Zeng, 2016) and (Zhou, Li and Yu, 2009) ANP Weights Entropy Weights TOPSIS (Yu, Zhao and Li, 2019) FTS ELECTRE Figure 5. Synthetic structure of the model resulted as aggregation of models proposed by (Mei et al., 2016;Yu et al., 2019) and (Zhou et al., 2009).
decision-makers to build a network making visible the interdependencies among criteria.Secondly, to define the criteria weights by conducting a pairwise comparisons for each relationship between groups and single criteria.Once the ANP model is built, it is possible to create the 'Supermatrix'.This underlines the relationships among criteria by representing the relative importance of an element on the left of the matrix compared to an element at the top of the matrix, with respect to a particular control criterion by giving specific weights (Saaty, 2006).In this matrix, the subjectivity of the weights is then balanced by more objective weights thanks to the use of the EMM (Mei et al., 2016) which is better described in Step 5 (see Table 4).

Macro-phase 2
Entering the macro phase 2, and more specifically the Step 3 (see Table 4), it is required to make sure to use qualitative criteria as the quantitative ones to allow an objective comparison.To address this issue, it has been used the FST (Lee et al., 2009).More specifically, it has been used to convert the qualitative criteria into quantitative values by using the graded mean integration representation method (Jolai et al., 2011), thus by transferring a triangular fuzzy number (TFN) A = a 1 ; a 2 ; a 3 ð Þ into a crisp real number.Having set all the criteria as numeric variables, it is possible to enter the Step 4 (see Table 4) allowing to aggregate the evaluations coming from the several decision-makers.In doing that, geometric mean is used.This is defined as the k th root of the product of k numbers.In the context of suppliers' evaluation, each decision-maker z ¼ 1; . . .; kexpresses his opinion x z ij regarding each supplier i ¼ 1; . . .; m for what concerns criterion j ¼ 1; . . .; n.Equation 1 defines the geometric mean of evaluations made by all decision makers.
ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi The aggregated values can be used as inputs for the Step 5 (see Table 5) requiring the application of the EMM (Mei et al., 2016).According to Shannon (1997), entropy is a measure of the uncertainty of information according to which the more data are dispersed, the bigger the uncertainty is, the greater relevance the criterion has.In this way, each decision on the criteria can be expressed by entropy values, and the relative importance can be determined objectively.More specifically, this model presents three main clusters of criteria (e.g.economic, environment, and social as reported in Table 6) within which there are specific criteria to be assessed.So, considering j ¼ 1; . . .; n as the single criterion and i ¼ 1; . . .; m the different suppliers, the idea is to compute first the normalized decision matrix R ¼ r ij À � based on the nature of the single criterion x ij as follows: Benefit Criterion r ij ¼ x ij ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi Cost Criterion r ij ¼ 1=x ij ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi Figure 7 depicts how it looks like in the excel file Based on the normalized values p ij of r ij À � to be computed as in Equation 4, each entropy measure e j is computed as reported in Equation 5.
The measure e j allows to calculate the importance of the criterion that is greater when the divergence value d j computed in Equation 6 is higher.Finally, the divergence value is used to compute the normalized criteria weights w j based on Equation 7. The denominator consists in the sum of the criteria grouped by sustainable aspects.
Step 6 (see Table 4) aims to integrate the results from the ANP and EMM to ensure the objectivity of the evaluations and in doing that, a linear combination of the two is computed as suggested by (Mei et al., 2016).To allow this integration, it is first necessary to associate the ANP weights with the different sustainable pillars and normalize them to align these results with those of the EMM (which have been already computed for the single pillar of the TBL).The vector containing the ANP weights is composed by the weights of the different criteria j ¼ 1; . . .; n within which n can be considered the sum of all the criteria corresponding to the different sustainable pillars as follows: where a corresponds to the sum of all the economic-related criteria, b to the sum of all the environmental-related criteria and c to those related to the social aspects.Therefore, the three vectors of normalized ANP weights can be obtained by means of Equation 8, Equation 9, Equation 10.
The final weight vector W j j ¼ 1; . . .; n ð Þ is obtained as in Equation 11where subjective weights provided by ANP are reported as w 00 j , and those objective weights computed with the EMM are reported in the equation as w 0 i .This vector multiplied by the evaluation matrix R ¼ r ij À � allows to obtain the weighted normalized evaluation matrix V ¼ v ij À � reported in the Equation 12.
w j ¼ w 0 j w 00 j P n j¼1 w 0 j w 00 j ; j ¼ 1; . . .; n (11) These results are the required inputs to implement the hybrid TOPSIS (Step 7 of Table 4) adopted to evaluate the performances of suppliers with respect to economic, environmental, and social aspects respectively, which require the EMM adoption for every single pillar.The TOPSIS method aims to select the alternative that has (Hwang & Yoon, 1981): • the shortest distance from the Positive Ideal Solution (PIS), solution that maximizes the benefit criteria and minimizes the cost criteria; • the farthest distance from the Negative Ideal Solution (NIS), solution that maximizes the cost criteria and minimizes the benefit criteria.
Based on the weighted normalized decision matrices V ¼ v ij À � computed above, PIS and NIS are calculated as reported in the Equations ( 13) and ( 14), respectively by considering 'I' the cluster of the benefit criteria, and 'J' the cluster with the cost criteria.
To finally compute the relative closeness C i , of each solution i to the best solution with the Equation 17, the distance from PIS and NIS is computed for each solution i, as reported in Equation 15 and Equation 16 respectively: ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi

Macro-phase 3
Entering the third macro-phase means to be able to rank the suppliers and chose the most proper one looking in an aggregated way to the three sustainability pillars.Therefore, the improved ELECTRE method is used to rank all the alternatives (Yu et al., 2019).First, starting from the weights obtained in Step2 related to criteria referred to each sustainability pillar, the cumulated weights for each sustainability aspect are computed as the sum of the single weights (Step 8 of Table 4) with Equations 18, 19 and 20.
These weights are combined with those of the decision matrix that is based on the ranks of suppliers with respect to economic, environmental, and social criteria obtained through the TOPSIS method.Therefore, in Step 9 (see Table 4), the EMM is used and applied to this new decision matrix X to obtain a normalized decision matrix R ¼ r ij À � .Therefore, Equation 8, Equation 9, Equation 10, Equation 11, and Equation 12 are applied to obtain three weights w j one for each sustainable pillar.After Step 8 and Step 9, six weights (i.e. two weights, thus entropy and ANP weights, for each sustainability pillar) are obtained and in Step 10, the linear combination is applied aggregating the six weights obtained with the three weights v j using Equations 11 and 12).
To concretely apply the ELECTRE method, the criteria required to be divided into concordance and discordance subsets C kl and D kl for each pair of supplier alternatives A k and A l (k; l ¼ 1; 2; . . .; m and k ¼ l), where: • The concordance set of criteria C kl is the collection of criteria where A k is better than or equal to A l (Equation 21) • The discordance set D kl contains all criteria for which A k is worse than A l (Equation 22) In this set of criteria, for each pair of suppliers k and l, it is possible to compute all the concordance and discordance indexes c kl and d kl relying on Equations 23 and 24.Based on the discordance index, it computed the revised discordance index d 0 kl (Equation 25) to be included in the computation of the revised weighted aggregated matrix E ¼ e kl ð Þ as reported in Equation 26.
Finally, to evaluate the net advantage value adv k referred to the supplier K (thus the alternative A k ), it is used Equation 27.The higher the adv k the better the alternative is.This value allows to have a single rank of the alternatives in a decreasing order, with respect to all sustainability aspects and related selected criteria.
Wrapping up, thanks to this revised C-SSS model, it is possible to overcome some criticalities emerged in previous works as reported in Table 7 and to enable companies operating in the cosmetics industry to choose the most appropriate suppliers according to their sustainable-oriented strategies and needs.

Cosmetic sustainable supplier selection model application and validation
The proposed C-SSS model was applied in a manufacturing company operating in the cosmetics industry in the north of Italy.The company has a strong commitment towards sustainability, which is also reinforced by the requests coming from their direct clients and final consumers.The application of the model requires the involvement of both the suppliers and the company's decision makers.Therefore, to streamline the information gathering and sharing, two different tools are used: Excel and SuperDecision Software as reported in the methodology section (see section 2, Table 5).
First, the decision-makers must select the criteria, grouped in an excel file, to be considered for the supplier selection.Based on this initial choice, the questionnaire is created and sent to the suppliers.The questionnaire is based on both qualitative and quantitative questions reported on an Excel file.While the quantitative questions can be easily filled with numeric variables, the qualitative ones are detailed questions to which it is possible to answer 'yes' or 'no'.An example is reported below in Figure 8.
The answers are filled in by the decision-makers and their judgments are expressed into a unique scale relying on the FST model as previously described.Figure 9 reports an example of the excel sheet used by the single decision-maker.
The decision-makers are also required to create the network among the criteria.By adopting the ANP model it is possible to keep into account the interdependencies present among the different criteria.This network puts the basis to identify the pairwise comparisons and obtain the weights for each single criterion.The ANP approach is concretely adopted relying on SuperDecisions software.This software requires first to set and populate the clusters of criteria (in this research the criteria are chosen among those reported in Table 6), second it requires to identify and populate the clusters with the available solutions, which in this case correspond to the different suppliers among which to choose, as presented in Figure 10.
Finally, the software enables to make the decision-makers define the inner and outer dependencies present among the criteria as reported in Figure 11 and to ensure the pairwise comparison among the criteria in a scale from −9 to + 9 as reported in Figure 12.
Based on these analyses, it is possible to obtain the weights of the criteria to make them usable automatically by the software.These weights are normalized based on the clusters considered as reported in the example referred to the case under analysis reported in Table 8.
The normalized values are used in the TOPSIS model while the limiting values, those coming from the automatic computation performed by the SuperDecision Software, are used as inputs in the ELECTRE model.These weights, combined with the responses by the suppliers, enable to finally rank the suppliers to choose the best option as reported in the example depicted in Figure 13 (the potential suppliers are only four in this specific case).This combination enables the creation of specific scores for the different suppliers looking on the TBL.Based on that, the model enabled to express a preference on the most suitable supplier and additionally, it enabled to look at the results across the different sustainability-related pillars (i.e.TBL pillars).This second result is valuable when there are two or more suppliers reaching the same final score, but it is required to select only one among the two.In this way, it is possible to decide, for instance, to select the one that obtained the highest score in the environmental-related criteria in case it would be more aligned with the strategic goal of the company in that historical moment.In this specific case, Supplier A was chosen as best option for the company as visible in Figure 13.

Discussion
As emerged from the results obtained after the model adoption (i.e. the C-SSS model), the company was able to integrate the perspectives of several decision-makers, to balance their judgments and finally to rank the different suppliers considering the sustainabilityrelated criteria enhancing their stakeholders' engagement capabilities.Therefore, the C-SSS model acts as facilitator to improve the circular-manufacturing ecosystem giving the possibility to establish stronger partnerships and collaborations among manufacturers and suppliers.Indeed, the selection of specific suppliers stands in the accurate analysis of criteria covering the whole TBL in accordance with the circularity principles, looking for instance at the possibility to recycle resources, reduce resources consumption etc.Moreover, the C-SSS model enabled to overcome the key weaknesses emerged in the already existing models, being it a comprehensive model able to involve different decision-makers while ensuring objectivity in their choices, capable to consider the potential interdependencies present among criteria while ensuring the inclusion of the entire spectrum of sustainability-oriented criteria required for the cosmetics industry.Last, the adoption of the C-SSS model benefitted the company ensuring a deep and thoughtful audit of their suppliers allowing the creation of a stronger and more resilient value chain through their engagement.Indeed, the questionnaire shared with the suppliers increases the suppliers' awareness in terms of the expectations from the company regarding sustainability and circular-oriented certifications or required characteristics.Therefore, the contributions from this research cover both practical and theoretical perspective.
From a practical perspective, the C-SSS model, once used in an industrial case, emphasized its potentialities in addressing a core aspect facilitating the sustainable transition of cosmetics companies, thus the supplier selection process.Through the industrial case, it emerged that the model allows to make collaborate and converge a series of decision-makers, even with contrasting ideas, working for the same company but operating in different departments (e.g.production and quality).Decision-makers from different departments were asked to evaluate and select a set of criteria, under the big umbrella of sustainability for cosmetics companies, required to evaluate the most appropriate supplier for their company.Based on this initial choice, the questionnaire for the suppliers was developed and shared with them to select the best option.Therefore, from the model application, it emerged that the model enables to engage suppliers in the sustainable transition of the producer itself making them aware about the sustainable characteristics required by the company.Actually, thanks to an accurate integration of the needed characteristics of the existing SSS models, it was possible (i) to involve different decision-makers while limiting the subjectivity in their choices, (ii) to consider a wide spectrum of criteria referred to sustainability, together with their potential interdependencies, and finally (iii) to select the most appropriate supplier for the company.This model, hence, based on a mathematical computation, facilitates the company in selecting the best supplier considering among all, the request coming from consumers and governments integrated with the company's strategic objectives.Moreover, it enables to take into account the potential interconnections present among the criteria selected.The set of criteria to be selected by the decision-makers includes those found in the scientific and grey literature in terms of ISO standards mostly linked to the SDGs.In this way, a cosmetics company can conduct an updated vendor rating ensuring the adherence with the current norms and standards for its industry.Moreover, the model modularity and adaptability facilitate its usage also in the future when the set From a theoretical perspective, the C-SSS model allows to overcome the criticalities emerged in the already existing SSS models by integrating their fundamental characteristics (e.g.judgement objectivity, multiple-decision makers, criteria interdependencies, sustainable criteria, etc.) into a unique and comprehensive model.Therefore, it was possible to create an SSS model, the C-SSS model, tailored on cosmetics companies interested in embracing a sustainable-oriented path, which before was not present in the extant literature.The model innovativeness stands not only in the inclusion of all the sustainable-related criteria needed to be considered by a cosmetics company but also in the accurate usage and selection of existing SSS models' characteristics needed for each specific stage of the supplier selection process.The C-SSS model, with respect to the previous models, includes in the selection process a wider set of criteria for cosmetics companies, including standards, covering the whole TBL.It takes into account criteria interdependences and engages several stakeholders (company's departments and suppliers) while keeping high the level of objectivity of the internal stakeholders' choices.More specifically, referring to the existing models, to ensure an agreement among the different decision-makers on the evaluation, it has been introduced the geometric mean on the different evaluations, and to reduce the subjectivity coming from decision makers' evaluations it has been used the EMM approach.Moreover, to keep into account the interconnections that might be present among the criteria considered, the ANP approach has been included in the model too.Last, to ensure the creation of a hybrid compensation model, the TOPSIS and the ELECTRE models have been concurrently adopted.Indeed, the combination of compensating and non-compensating approached within a unique

Conclusions
The RO of this contribution aimed at overcoming the inefficiencies and criticalities emerged in the existing SSS models by developing a novel one tailored on the cosmetics industry.This revised model ensures the coverage of the most recent advancements and requirements for cosmetic companies in regards of sustainability and circularity issues leading to the creation of a circular manufacturing ecosystem.In doing that, it has been conducted firstly a review of the extant scientific and grey literature investigating the key sustainable and circular criteria to be included in the model to ensure the coverage of the TBL pillars when dealing with cosmetic companies.In parallel to this, it has been reviewed the literature about existing SSS models to analyse their key characteristics, their key weaknesses and to evaluate what should have been kept into the revised novel tailored on cosmetics.Based on these analyses, the set of sustainable-related criteria for cosmetic companies has been identified and the SSS models to be integrated have been selected and deeply analysed.Grounding of these findings, the C-SSS model has been developed and applied in an Italian manufacturing company operating in the cosmetic industry.
Regarding the limitations of this research, although the technical tools used (i.e.Excel and SuperDecisions software) are affordable and easy to be accessed by any type of company, ad hoc tools might be considered in the future to be easily integrated with the vendor rating software already used by cosmetics companies.Additionally, a survey could be conducted on a wider set of manufacturing companies operating in the cosmetics industry to verify that the current set of criteria included in the model is exhaustive and complete.In parallel to that, a more extensive literature review focused only on cosmetics sustainable-related criteria might be conducted in future research.Last, the present model should be applied in a wider number of companies in the future to verify its generalizability.

Figure 2 .
Figure 2. Research design for the realization of the C-SSS model.

Figure 7 .
Figure 7. Excel for the score computation.

Figure 9 .
Figure 9. Example of an excel sheet used by a single decision-maker.

Table 1 .
Sustainable-related criteria in SSS models.

Table 2 .
ISO and related SDGs for the cosmetics industry.

Table 4 .
Model development steps.

Table 5 .
Stakeholders involved in the selection process and technical interfaces.

Table 6 .
Criteria for the supplier selection in cosmetics.

Table 8 .
Clustering of criteria evaluation.