How to Measure Sustainability in the Supply Chain Design: An Integrated Proposal from an Extensive and Systematic Literature Review

Abstract

The increase in the world population and resource scarcity has led to the introduction of environmental concepts such as sustainability and sustainable supply chain design (SSCD). However, there is a lack of consensus among researchers on how to measure sustainability in SSCD. Therefore, the authors propose a novel approach to measuring sustainability in the context of SSCD by developing an integrated, tractable, and representative metrics framework. The methodology corresponds to a quantitative approach involving bibliographic examination and statistical techniques. First, the authors conducted a systematic literature review by formulating research questions and a search protocol, searched for relevant articles, and conducted a quality assessment on full-text reviews to obtain metrics for measuring sustainability in SSCD from the literature. Then, they defined aggregation criteria representing their inclusion relationship by merging associated metrics. The authors then used Cluster Analysis (CA), a multivariate statistical technique, for grouping the metrics. Consequently, twelve clusters were distinguished from 541 research articles, grouping 51 metrics from different sustainability dimensions. It shows the strong connection among the sustainability dimensions, i.e., they must be assessed holistically. Then, we proposed reducing the 51 metrics to 5 to evaluate sustainability in the SSCD, allowing us to focus on a reduced number of indicators.

1. Introduction

The world population has doubled in the last fifty years, while vital resources have become increasingly limited [1]. However, several companies contribute more to resource depletion and environmental problems due to their increased raw material and energy consumptions [2]. In light of resource scarcity, certain environmental concepts have been incorporated into the design and management of production systems. One such concept is sustainability, which refers to the capacity of enterprises to meet their immediate financial needs while ensuring that they, as well as others, can meet their future needs without compromise [3]. From a holistic perspective, sustainability denotes a form of development that fulfills present requirements while ensuring that the capacity of future generations to fulfill their own needs remains intact [4].

The multidimensional nature of sustainability has been defined in recent literature as the strategic attainment and integration of an organization’s social, environmental, economic, political, and technological aspects [5,6,7,8,9,10,11] through the systemic coordination of the main inter-institutional business processes [12]. Consequently, both governmental and societal concerns have been raised about environmental protection and corporate social responsibility, leading to constant pressure on companies to reassess their supply chains—not only in terms of economic objectives but also environmental, social, political, and technological concerns [13,14]. This is reflected in the increase in company sustainability reports in the last 20 years [15].

This viewpoint introduces novel factors that must be considered when designing supply chains, a practice now known as sustainable supply chain design (SSCD). SSCD aims to effectively measure and achieve sustainability dimensions, primarily by aligning with the Sustainable Development Goals (SDGs) outlined by the United Nations (UN) [16]. Over the past few years, numerous studies have been conducted across various production sectors, including applications within the healthcare industry [17], big data [18], fuels [19], energy [20], textile [21], and water resources [22,23].

In practice, research on SSCD has utilized a wide range of metrics and methodologies to address each dimension of sustainability [11,24,25]. Several literature reviews have demonstrated how SSCD could effectively incorporate sustainability [24,26,27,28,29,30].

Table 1. Related literature review assessment.

Ref	Years	N Articles	Sustainability Dimensions
			Economic	Social	Environmental
[37]	1997–2010	36	Total cost, net revenue	Profit sharing, employment, and income distribution	LCA-based environmental impacts: energy demand and CO2 emissions, natural capital, or  resources
[28]	January 2008 to October 2020	354		Customer service level, attendance to demand, and reduction of work accidents	CO2 emission, use of energy and/or the number of tailings
[30]		54	Total annualized supply chain cost, annualized profit, total profit, revenue, NPV	Accrued jobs, land use changes, traffic annoyance	GHG emissions, Eco-Indicator 99, non-renewable energy use, water use, pollution, CO2 emissions, Impact 2002+
[29]	1995–2017	188	Overall costs, NPV, raw material availability and energy potential, payback period calculation, prices, energy potential	Incomes, calorie consumption, energy access, people in water stressed areas, child deaths, employment, health and safety	Eco-indicator 99; ReCiPe, GHG emissions, cumulated energy demand, global warming potential, acidification potential, primary energy use, land use efficiency, energy consumption, particle emissions, agriculture land use, climate change.
[38]	1997 to July 2016	146	Total cost, risks on investment, efficiency, NPV, total profits, financial revenue, total transportation cost, logistics cost of raw material collection, transport distance, unit cost, economic potential, conditional value-at-risk, marginal delivery cost	Job opportunity, social impact, number of workers, total service level	GHG emissions, total GHG emission savings, net energy out, environmental impact, global warming potential
[39]	Up to Dec. 2019	112	Resource productivity indicator, total costs	Job creation	Waste and emissions related, CO2 emissions, GHG emissions, Eco-Indicator 99, non-renewable energy use, water use and pollution, Impact 2002+
[40]	2000–2015	over 20,000	Cost of production	Food security, human health	GHG emissions, air quality (non-GHGs emissions), soil resources, land use change, water resources
[41]	2000 to 2014/2015		Net income from sales, productivity in primary feedstock production, number, and capacity of routes for critical distribution systems, capacity use and flexibility, gross value added, energy diversity	Employment created, incidences of occupational injury, illness and fatalities in the production process, uncertainty of tenure and land rights	GHG emissions in production, soil organic carbon maintained, non GHG emissions, water withdrawn, pollutant loadings to waterways and bodies of water related to raw material obtention, area and percentage of lands of high biodiversity converted for production, net energy ratio in individual process steps, the  change in diversity of total primary energy supply
[32]				Employment, occupational accidents, unemployment, hazardous work, vulnerable employment, social security, access to clean water	GHG emissions and the use of basic resources, air pollution, damage to species richness, energy consumption, waste production, CO2 emissions
[33]	2008–2019	132			Pollution, soil degradation, product losses and waste, GHG emission, resources consumption, environmental damage or stress
[35]					Carbon footprint, water footprint, ecological footprint, land footprint, nitrogen footprint, phosphorus footprint, chemical footprint, PM2.5 and PM10 footprints, ozone footprint, material footprint, biodiversity loss
[42]		78	Economic performance, financial performance	Human rights, community development	Low-carbon products, low-carbon logistics, low-carbon production, energy consumption
[24]	2000-2015	190	Production performance metrics	Product safety, work safety	Ecological footprint, emissions, pollution
[36]	2012–2015	979	Total supply chain cost, net revenue, profit		GHG emissions
[43]	2006–2016	85	Profitability, cost, revenues, NPV	Job generation, food security, respects for property land rights, social acceptability, working conditions	GHG emission, waste management, wastewater management, biodiversity conservation and protection, energy efficiency
[44]	1997–2012	71	Overall cost, overall profit, NPV, financial revenue, risk on investment, transport cost	Number of jobs, social footprint	GHG emissions, maximize energy return in the conversion facility, minimize energy used in the supply chain, maximize net energy profit
[27]		10 Reviews + 188 articles	Cost reduction, profit, NPV, expected return, economic output, financial risk, total value of purchasing	Service level, number of accrued jobs, hours of employment, injury rate, satisfaction levels of stakeholders and customers, social risks	GHG emissions, energy consumption and water consumption, waste production, CO2 equivalent, CO2 emission per capita, embodied carbon footprint, air pollution, global warming
[45]	1999 to May 2016	220	Cost, profit, NPV, risk	Job creation, safety, health, number of working hours, discrimination, satisfaction, and  poverty aspects	Global warming, LCA impacts, waste reduction, recycling, biodiversity, renewable energy consumption
[46]	1995–2018	198		Number of jobs created by the supply chain, number of workdays missed by employees due to health problems, ethical supply chains, equitable treatment of stakeholders, education and training, social justice, and diversity.	CO2 emissions, natural resources utilization, and product recovery
[29]	2000–2017	50	Profit, cash flow, delivery lead time, customer satisfaction, trade level, budget variance, total cost, capacity utilization, production effectiveness, product quality	Employment, occupational health and safety, local communities, food to energy competition, jobs created, job opportunities created, social benefits	Eco-Indicador 99, Recipe 2008, Impact 2002+, global warming potential, pollution, CO2 emissions, NO2 emission, CO emission, volatile organic compounds, water usage, green appraisal scores, carbon trading, new technologies, new material for products, water quality, fossil fuel consumption
[47]	1900–2018	40	Total cost, total profit, inventory, routing costs, product waste cost	Storage and distribution of infectious medical waste and hazardous material, customer dissatisfaction	Total carbon emissions from logistics operations, carbon emissions by pricing them, reducing waste generation, collection of waste
[48]			Net cash flow generated	Employment	Net GHG emissions, emissions from carbon stock change due to land use, potential environmental risk, land use intensity, energy use, materials use, fertilizer and pesticide use, chemicals used for raw material obtention, water use, wastewater to be treated
[49]	2015–2018	113	Reliability, responsiveness, flexibility, financial performance, quality, transportation costs and establishment costs of facilities, logistics activity costs, purchasing, carbon emission cost, profit, total cost, NPV	Work condition, human health and safety, societal commitment, customer issues, business practices	Environmental management (environmental certification owned by the company), use of resources (use of raw or recycled material, water, and energy from the surrounding area), pollution (methane (CH4) and nitrous oxides (NOx), carbon dioxide CO2)), dangerousness, natural environment
[34]	1990–2014	87	Cost of facility investment, feedstock purchase and transportation, pollution cost, logistics costs, total annual cost, wastewater treatment costs	Work conditions, social commitment, customer issues, human rights, and business practice	Methods (Eco-Indicator 99, Impact 2002+, CML92, Recipe), Impact category and indicators
[50]	2005–2016	333	Total cost, service quality	Customer service level	CO2 emission
[51]			Food versus fuel debate, efficiency, and energy balance, and increasing biofuel budget programs	Poverty reduction potential, land and crop indirect impacts, and effects on social resources, such as water utility systems	GHG emission, water resources quality, soil degradation and loss of biodiversity
[26]	1987 to March 2019	247	Total cost, profit, NVP	Food quality and safety, food security, social welfare, job generation and equality, supporting small enterprises, public and dietary health, consumer price fairness, food donation, corporate social responsiveness investment, social cost of GHG emissions	Carbon footprint and emissions, biomass energy production, waste disposal and food loss, land use and erosion, energy consumption, water use and contamination, LCA impacts, freshness-keeping effort, green effort, organic agriculture

shows that, among the related literature reviews, those with no details regarding the years considered or number of articles reviewed correspond to narratives reviews; this means that they are based essentially on the researcher’s experience [31]. In addition, Table 1 shows that by each dimension of sustainability, there are several aspects assessed. For example, regarding the environmental dimension, refs. [29,32] integrate the Eco-indicator 99 and ReCiPe 2008, each considering different impact category indicators at the midpoint (as acidification potential) or endpoint levels (as damage to ecosystem quality). On their behalf, ref. [33] assessed the use of essential resources such as land, water, and materials, as well as air pollution represented by footprints of $N{O}_x$ and $S{O}_2$ emissions and fine particulate matter ($P{M}_{2.5}$) emissions. They also considered the damage to species richness as a consequence of pollutants, GHG emissions, and the use of land and water. Meanwhile, ref. [34] assessed the pollution emitted into air and water and considered resource consumption as energy or water. Ref. [35] considered impact categories and indicators of climate change, biochemical oxygen demand, damage to human health, and water footprint, as well as performance measures such as residual waste generated, GHG emissions, energy consumption, and amount of recycled material. With even more detail, ref. [36] described several footprints as follows. Carbon footprint or GHG footprint considers carbon dioxide ($C{O}_2$), methane ($C{H}_4$), and nitrous oxide (N₂O) emissions to the atmosphere. Water footprint measures both the consumption of freshwater as a resource (including both blue and green water) and the use of freshwater to assimilate waste. The latter component refers to a greywater footprint. The ecological footprint measures land appropriation to produce renewable biomass resources and uptake waste via $C{O}_2$ sequestration. The land footprint measures the land required to supply food, materials, energy, and infrastructure, expressed in physical hectares or equivalent land units (global hectares). The nitrogen footprint measures the emissions of reactive N to the atmosphere and water bodies. The phosphorus footprint measures P’s use as a resource and P’s losses to water bodies. The chemical footprint accounts for all chemical substances released into the environment, which may ultimately lead to ecotoxicity and human toxicity impacts. The $P{M}_{2.5}$ and $P{M}_{10}$ footprints measure particulate matter pollution in the atmosphere. These are also included in the chemical footprint. The ozone footprint measures the emission of gases controlled or due to be controlled under the Montreal Protocol in terms of ozone-depleting potential weighted kilograms. The material footprint measures the use of materials from a consumption perspective, allocating all globally extracted and used raw materials to domestic final demand (metal ores, nonmetallic minerals, fossil fuels, and biomass (crops, wood, wild fish catch, etc.)). Finally, biodiversity loss measures the impact as a result of different pressures, such as land and water use or chemical pollution.

Note that frequently up to one metric is assessed by sustainability, which varies depending on the research [37,50]. It implies several possible metric combinations for the SSCD, considering the large number of metrics that can be evaluated for each sustainability dimension [37,50]. Thus, it should be emphasized that there is currently a lack of consensus among researchers regarding the optimal metrics to accurately represent each sustainability dimension and how to depict the overarching concept of sustainability within the framework of supply chain design. This research tendency has implied different approaches and metrics to assess sustainability and, then, the following question emerges: How do you measure sustainability in sustainable supply chain design (SSCD)?

It leads to the need for a comprehensive and integrated framework to depict the sustainability measure in the SSCD, evidence of at least two new significant problems to be addressed [52,53,54,55]. First, adopting multiple metrics to evaluate each sustainability dimension could search for a feasible solution, ideally optimal, by any resolution method approach. Second, a particular solution from a limited set of metrics could have substantive differences in terms of results in comparison with another metric’s selection, seeking an isolated goal and avoiding a comprehensive vision of sustainability and the relationships of its components [50]. In addition, similar metrics could be considered in more than one dimension. For instance, both logistic cost (from the economic dimension) and greenhouse gas (GHGs) emissions by transport (from the environmental dimension) require distance between the supply chain actors as a parameter for their computation. Another example is total carbon emission (from the environmental dimension) and the carbon emission cost (from the economic dimension), where the former, weighted by the carbon cost parameter, provides the latter.

Our Contribution

In this paper, the objective is to propose an integrated, tractable, and representative metrics framework to measure the five sustainability dimensions: Economic, Social, Environmental, Political, and Technological, which allows us to address the problems related to measuring sustainability in the sustainable supply chain design (SSCD). This research is based on a quantitative approach involving mainly bibliographic examinations and multivariate relational and statistical techniques. To our best knowledge, this report describes a novel approach that has not been followed in previous research in a sustainable setting. Formally, our contributions are threefold: First, we conduct an exhaustive literature review to analyze the measuring of each of the five sustainability dimensions. This process follows a systematic literature review process through a practical and methodological analysis, distinguishing temporal trends, countries, the main production sectors, methodologies, decision-making levels, and metrics considered to measure each sustainability dimension from 541 published papers available in the Web of Science (WoS) database, until the year 2020. Second, we work on the above-obtained results and develop an integrated metrics framework based on aggregation criteria and Cluster Analysis (CA) methods. It allows for the representation and identification of the relations among different parameters and metrics to be computed/optimized in each of the five sustainability dimensions in SSCD from the literature. In addition, it provides a systemic scheme to incorporate other new metrics from future research. In practice, we propose 12 clusters and a reduced group of metrics to measure sustainability as a basis for novel decision-aid models for production systems and logistics design. It will support and facilitate sustainability management in supply chain design for decision makers in the industry. Third, we discuss our findings and their theoretical and managerial implications, leaving open questions to be addressed in future work about sustainability in SSCD and providing insights from our results to guide answers from research and practice perspectives.

The paper is structured as follows, Section 2 introduces the proposed methodology by integrating a literature review and statistical analysis. Section 3 presents relevant results regarding trends in supply chain scientific literature and sustainability measure identification. Then, Section 4 describes the implications of those results on measuring sustainability in supply chain design. Finally, Section 5 presents an overview of the main results and their implications as well as future research questions.

2. Materials and Methods

2.1. Literature Review

To conduct our exhaustive and systematic literature review, we adopted the search methodology for a systematic literature review presented by [56] because it generalizes the stages and steps for a successful literature review. This methodology includes three major stages: (i) planning the review, (ii) conducting the review, and (iii) reporting the review. The initial phase involves recognizing the need for a review, determining research queries, and constructing a review protocol. The subsequent stage entails identifying and selecting primary studies and extracting, analyzing, and synthesizing pertinent data. Lastly, the third stage involves the dissemination of the resultant findings.

In particular, the research questions for the initial phase are defined as follows:

• What methodologies have been used to measure sustainability in the SSCD?
• At what decision-making level has sustainability been measured in the SSCD?
• How has sustainability been measured in the SSCD?

Then, the review protocol considers as keywords the concepts related to these questions, which are formulated as the search string: ((“Green Supply Chain” OR “Sustainable Supply Chain”) AND (Design OR Conception)) OR ((“Supply Chain Design”) AND (Sustainable OR Sustainability)). Note that the search string does not include decision making or metric-related keywords in order to not restrict the search.

In the second stage, we establish a search strategy corresponding to search articles available in the Web of Science (WoS) database, which is widely regarded as the foremost scientific citation search and analytical information platform [57]. This search strategy focuses on articles published up to December 2020, utilizing keywords that are searched for within the database’s Title, Abstract, and Keywords sections. Note that no initial date was selected to identify the first related literature. The inclusion criteria involve evaluating whether the research articles identified are relevant to the research queries. Furthermore, the screening procedure involves the initial review of the titles and abstracts to identify articles that satisfy the inclusion criteria, as Figure 1 shows. Then, in the third stage, we performed a refined quality assessment on a full-text review to select the articles for data extraction. After the literature search and selection, the literature assessment focused on the research questions defined for the data analysis, particularly to obtain the metric used for measuring the sustainability in SSCD from the literature.

2.2. Aggregation Criteria, Cluster Analysis (CA), and Reduction Rules

Specifically, we determined parameters and metrics from the literature assessment and defined aggregation criteria to represent the inclusion relationship between them. It is formally defined as follows: “An element (A) aggregates another element (B) if and only if the element (B) correspond to the previous calculations required to obtain the value of the element (A)”. For example, profit maximization (A) integrates the total supply chain costs ($B_1$) and the revenues ($B_2$) by adding them. This reduction follows similar initiatives in other research communities, such as the scheduling setting, where the reduction allows the representation and identification of the relations among different parameters and objective functions of the scheduling problems (see the “scheduling zoo” initiative in [58] for details). To our best knowledge, this report describes a novel approach that has not been followed in previous research in a sustainable setting. In this case, we formally identify and define sets of parameters, auxiliary metrics, and final metrics from the measuring analysis of sustainability in SSCD provided by the literature review, stating the relationships between them based on the defined aggregation criteria. We remark that the final metrics are stated from the metrics recognized from the literature review, merging other associated metrics. The parameters and auxiliary metrics are identified from the considered final metrics.

This procedure allows for assessing the relationship among the different metrics, as Figure 1 presents, to analyze the interrelationship among the sustainability dimensions. To analyze it, we consider the multivariate statistical technique, Cluster Analysis (CA), which groups elements to achieve the maximum homogeneity within each group and the highest difference between groups based on the relationships among the metrics [59]. CA can be performed in Gephi open-source software for graph and network analysis [60]. The obtained results allow the construction of a set of directed acyclic graphs, where a directed arrow represents the aggregation criteria to a single metric from the aggregated metric. In this representation, we remark that many metrics can aggregate a metric, and the node size of each metric is directly defined by its number of aggregated metrics. In practice, we obtain an interconnected network among all parameters and metrics used to measure sustainability in the SSCD. In this network, the sustainability dimensions integrated into each cluster and the relationships among the clusters would lead to understanding the interrelationship among the sustainability pillars.

Furthermore, to introduce the reduction rules, consider pollution generation and the pollution cost. In this case, one metric is contained in the other because a pollution cost factor is multiplied by the pollution production. Then, the pollution cost can be understood as a more complex metric or integrated at a higher level. Therefore, the objective is to identify the metrics at the higher level of integration. It would lead us to understand which metrics are a particular case of another metric. Finally, the more complex metrics or objective functions could be selected to measure sustainability in SSCD from five dimensions: Economic, Social, Environmental, Political, and Technological, since they all integrate other metrics.

3. Results

3.1. Literature Assessment

Following the review protocol, we found and scrutinized 1147 articles, of which only 541 research articles met the refined quality standards required for data extraction. During the initial screening, 422 articles were excluded, of which 63 were review articles, 82 did not involve supply chain design, 152 evaluated sustainability drivers, and 125 performed sustainability effects evaluations. The latter two categories involved ex post assessments, which were not within the scope of this research focusing on ex ante assessments. Additionally, 184 articles were excluded from the full-text review, of which four were review articles, 32 did not perform supply chain design, 85 evaluated sustainability drivers, and 63 assessed sustainability effects.

This section details the data extracted from the 541 research articles to solve the research questions presented in the previous section.

3.1.1. Trends in Related Literature

What methodologies have been used to measure sustainability in the SSCD?

The analysis of research articles based on methodology reveals that the majority, 62.85%, employ optimization models (O), followed by evaluation studies (Ev) with 17.01%, and simulation (S) with 10.91%. These details are depicted in Figure 2. The combined use of optimization and simulation (O-S) amounts to 3.14%, while only three articles employ optimization, simulation, and evaluation (O-S-Ev) jointly [61,62,63].

Most of the research articles classified as optimization developed mixed-integer linear programming models [64,65,66,67]. However, mixed-integer nonlinear programming models were also presented [68,69,70]. Research articles integrating several decision-making levels mainly develop two-stage models to incorporate uncertainty [70,71,72,73,74]. Even though stochastic mixed-integer linear fractional programming models to tackle multiple uncertainties regarding feedstock supply and product demand were developed [75]. Furthermore, research articles that assessed several sustainability dimensions frequently integrated multiple objective functions [65,73,74,76,77,78]. These multi-objective models have been solved with the Epsilon-constraint method [79,80]; particle swarm [67]; weighted sum methods, such as weighted Tchebycheff and augmented weighted Tchebycheff [66,77]; genetic algorithms [67,70], such as non-dominated sorting genetic algorithm [81], non-dominated sorting genetic algorithm-II [82], and tabu search [83], among others. In addition, game-theoretic approaches seeking optimal supply chain configurations were found [84,85]. Furthermore, DEMATEL methodology [86] and intuitionistic fuzzy-TOPSIS [87] have been applied to evaluate the suppliers’ characteristics for its selection. Besides, other evaluation research articles address the environmental impacts of the supply chain through the Life Cycle Assessment [88]. The methodologies applied in the simulation research articles include Multi-agent-based simulation [89], Discrete Event Simulation [90], and System Dynamics [91]. Furthermore, the research articles, including optimization and simulation, applied Monte Carlo to address uncertainty effects on supply and demand [92,93]. Even when optimization is the most used methodology to integrate sustainability in the supply chain design, future research should include uncertainty studies through evaluations or simulations.

At what decision-making level has sustainability been measured in the SSCD?

In the literature, three levels of supply chain decision making are distinguished according to the time horizon, the uncertainty, and the activities involved [94]. The strategic level at the base of the decision-making structure covers decisions such as facility location, storage capacity, production capacity, and supplier selection, among others [95]. These are long-term decisions taken with high levels of uncertainty, and they are the basis of tactical and operational decisions, designing the principal supply chain structure [96]. The tactical level covers aspects such as production and distribution planning, production allocation, transport capacities, inventories, and the management of safety stocks [97]. Finally, at the top is the operational level, integrating short-term or daily decisions, such as job execution, vehicle loading, unloading, and order delivery [98]. These decisions involve lower uncertainty degrees than the other decision-making levels. Consequently, we classify the research articles selected by the following criteria. A document accounting for the strategic decision-making level must address a long planning cycle of several years. Furthermore, a research article considering the tactical decision-making level deals with a shorter planning cycle (6 months to a year). Meanwhile, the research articles on the operational decision-making level involve weekly or daily planning tasks.

Figure 3 shows the number of publications assessing the different decision-making levels, either individually or integrated.Although it exhibits that the authors have focused mainly on the strategic aspects, in percentage terms, 41.96% of the research articles studied consider only decisions at the strategic level mainly related to supplier selection [99,100,101,102] and facility location [78,103], 5.18% involve decisions from the tactical level related to inventory strategies [104,105,106] and 18.11% only assess decisions from the operational level, devoted to scheduling [107], pricing [108,109,110,111], and transportation decisions [109,112,113], among others. The above reflects the essential importance of the strategic decision-making level in the supply chain. Furthermore, only 27 research articles consider the three decision-making levels, such as in [114,115,116,117,118,119,120,121,122,123,124], mainly developing models with more than one stage. This reduced the number of research articles due to the requirements for complex models and significant computational calculations, compared with the integration of one decision-making level, in the search for an optimal supply chain, considering optimization is the main approach used in the SSCD, as Figure 2 shows.

Related to the supply chain decision-making levels considered in the SSCD, we observed that at least 34% of the research articles integrate more than one level. Moreover, they provide interesting proofs in an integrated SC design, considering different planning horizons, indicating the need for uncertainty inclusion in the SSCD.

How has sustainability been measured in the SSCD?

Figure 4 shows the distribution of research articles according to the dimension of sustainability covered. Furthermore, 28% of the research articles integrate economic and environmental aspects; 17% focus on economic, social, and environmental dimensions (a set of dimensions called triple bottom line (TBL)); 11% corresponds to research articles devoted only to environmental aspects; the economic dimension is studied in isolation by 9%; and 5% of the research articles focus only on social aspects. It shows that environmental and economic aspects lead the sustainability studied in SSCD.

Only seven research articles integrate the extended definition of sustainability (i.e., environmental, economic, social, political, and technological), published between 2010 and 2020. Dev and Shankar [115] extend the knowledge of the limits of green supply chain management (GSCM) elaborated by [125] by finding a hierarchy of interactions between the sustainable boundary enablers with interpretive structural modeling methodology. The boundaries include environmental, economic, cultural, legal, political, technological, and temporal aspects.

Then, in the context of energy transition policy, ref. [126] investigate whether reframing the bioenergy supply chain design can allow sustainable regional development targets.The enablers studied include environmental factors, such as reduced agricultural fertilizer use, economic aspects such as biogas filling station installation, social aspects such as satisfying biogas demand, political aspects such as converting public sector vehicles to biogas, and technological aspects, such as stabilizing manure processing. Finally, ref. [127] focused on supporting managerial decision making based on a Delphi with domain experts and literature synthesis in the same trend. The supply chain activities ranked include the reduction of pollution to air, water, and land, minimizing energy and material consumption, reducing noise levels, utilizing renewable and alternative forms of inputs, and the discussion, investigation, and selection of alternative methods/options.

Ref. [128] developed a model for SSCD based on the ANP (analytic network process) methodology. It presents a case study applied to the electrical goods industry in Germany assessing environmental criteria (ISO 14001), windows for delivery, occupational health and safety (ISO 45001), corporate social responsibility (CSR), demand volume customers influence on distribution and manufacturing orders, and the duration of the product lifecycle.

Furthermore, ref. [82] focuses on Phase III biorefineries (mix feedstock and multiple products) in the Colombian context and develops a multiobjective optimization model solved with an adapted non-dominated sorting genetic algorithm II (NSGA II). It assesses the property concentration of cultivable lands, the net present value and transportation costs, the potential workstations, the governmental subsidies for the industry, and compare production technologies.

Meanwhile, ref. [129] investigates the impact of information sharing on the decisions and profits of the manufacturer and the retailer. The developed game theory models aim for the equilibrium of both the manufacturer and the retailer profits, including aspects such as the environmental impact of a product, promotional campaigns to capture the consumers’ attention, expected consumer surplus, subsidy policies to encourage consumers to purchase, and new technology to manufacture green product introduction by the manufacturer. Finally, ref. [130] developed research for hydrogen fuel cell vehicles applied to the Occitania Region in France, seeking an optimal hydrogen supply chain with the sequential application of an optimization strategy and a multi-criteria decision-making tool. The optimization model presents a social cost-benefit analysis, including $C{O}_2$ and pollution emissions, platinum depletion, externality costs and net present value, noise, a subsidy policy scenario assessment, and the evaluation of different production technologies.

The Supplementary Materials shows the research articles’ classification in detail according to the methodological analysis performed in this work.

3.1.2. Metrics Describing Sustainability in SCCD

Considering most of the research articles related to the SSCD are approached by optimization, the metrics describing sustainability could be represented as objective functions. Thus, Figure 5 presents a detailed description of the 51 objective functions to be optimized from the 541 research articles studied. Note that a number is given for each objective function (metric) in the second column, this number facilitates the relationship between the definition and the acronym presented in the Appendix B.

The main objective functions and optimization criteria considered to assess the economic aspect are minimizing total costs, maximizing profits, and minimizing transportation costs. Likewise, the main objectives sought in the social dimension are the maximization of job opportunities and social welfare. Regarding the environmental dimension, the main aim is to minimize $C{O}_2$ emissions, environmental impact, GHG emissions, and water use. Finally, for the political and technological aspects, it is sought to increase the high-quality green products in the market, assure food security, maximize the desired effects of the regulations, and minimize the related cost of innovative production technologies.

It is worth noting that economic functions constitute the majority (16 objective functions), followed by environmental functions (15 objective functions). Additionally, seven objective functions can be categorized into more than one dimension of sustainability, denoted by an asterisk in Figure 5. For instance, reducing taxes paid corresponds to the economic dimension, while it is also related to tax collection in the political dimension. Besides, maximizing high-quality green goods and/or services could be classified into social or political sections. Finally, the cost and net present value related to technologies could be classified in the economic section.

It should be noted that the objective functions described in this study apply to a general SSCD. Hence, some objective functions may be more suitable for a particular SSCD than others. Furthermore, the analysis identified 51 objective functions, leading to a many-objective optimization problem. Solving such a problem results in a set of nondominated solutions known as a Pareto-optimal set (POS) or Pareto front [131]. However, solvers for such problems are sensitive to the number of objectives considered, as computational costs increase with more objectives, making solution visualization and analysis more complex [45,132]. Therefore, considering the large number of sustainability metrics and the need for an integrated approach to SSCD, it is crucial to develop efficient many-objective models and dimensionality reduction techniques that effectively address different aspects of sustainable development [51].

Other topics such as the distribution of research articles focused on SSCD by year, the number of related research article applications in the SSCD by country, and the main production sectors in SSCD development are analyzed from the literature review. These allow us to evidence the SSCD as a relevant topic worldwide with the constant growth of related research articles. Furthermore, the leading countries are Iran and China, who focus on goods production, such as automotive and manufacturing products. However, Latin America, the Caribbean, and Africa were left behind. In the same vein, much remains to be done related to using residues in producing new products, fuels, and energy. See details in Appendix A.

3.2. The Aggregation Criteria and Cluster Analysis (CA)

From the above literature review, 51 objective functions (metrics) are recognized to be considered in the measure of sustainability in the SSCD by the decision maker.

To start the aggregation criteria, we initially merge the objective functions n° 16 and n° 47 associated with the net present value (NPV) (see Figure 5). Thus, we formally consider 50 final objective functions (metrics) and identify 58 auxiliary functions and sets of 48 parameters, stating the relationships among them based on the defined aggregation criteria.

The obtained results are described in detail in Appendix B and allow us to construct a set of directed acyclic graphs. The aggregation criteria are represented by a directed arrow to a single objective function from the aggregated function, as shown in Figure 6. Note that many objective functions can aggregate an objective function, and the node size of each one is directly defined by its number of aggregated functions. For instance, the total greenhouse gas emissions in the supply chain (TGHGESC) involve waste, wastewater, transport, production, and infrastructure GHG emissions. The total production cost (TPC) involves the raw material acquisition, water, energy, and fuel costs. Furthermore, social welfare (TSW) integrates the NPV, ROI, consumer surplus, social impacts, capacity use, environmental impacts, health impacts, and weighted customer satisfaction.

By considering the relationships among the final objective function, auxiliary function, and parameters based on the defined aggregation criteria, we analyze and reduce the number of functions to measure sustainability by considering the multivariate statistical technique called the cluster analysis (CA) method [59]. It provides a graph and network analysis using Gephi open-source software [60]. Figure 6 shows the results obtained, where the relationship between the parameters, auxiliary functions, and final functions allows us to identify 12 clusters, which are colored to improve the visualization of each one. In addition, we analyzed the cluster in terms of sustainability dimensions involved by its parameters and functions, as

Table 2. Cluster analysis according to the addressed sustainability dimensions. Economic (EC), social (SO), environmental (EN), political (PO), technological (TE).

Area	Dimension	Description
(1)	EC - EN	This cluster is related to waste generation by considering the waste emission amounts and their environmental and economic impacts.
(2)	EN	This cluster is related to wastewater generation by considering the wastewater emission amounts and their environmental and economic impacts. In addition to the total avoided emissions related to waste recovery.
(3)	EC–EN	This cluster is the economic and environmental impacts related to fuel, raw materials, and water consumption in production.
(4)	EC–EN	It involves the environmental impacts related to fuel consumption in transport, as well as the logistics cost.
(5)	EN–SO	This cluster includes consumer satisfaction and energy balance linked through the production assessment. It involves customer satisfaction by considering the maximum customers coverage, the customer service, quality of products, and delivery time to customers.
(6)	EC–EN–SO	It involves social welfare linked to other clusters, in addition to the total capacity use, the consumer surplus, and the health impacts related to the supply chain emissions.
(7)	EN–SO–PO	This cluster involves the social impact by considering hazardous materials used, occupational accidents, infrastructure redundancy, social impact according to geographic characteristics by considering the customers and suppliers’ geographical selection, and food security.
(8)	EC–TE–PO	It includes the governmental expenditures related to subsidies and taxes, the investment related to infrastructure implementation as well as the financial metrics: net present value and return on investment.
(9)	EN	This cluster involves the total greenhouse gas emissions in the supply chain and the environmental impact related to all the emissions in the supply chain.
(10)	EN–SO	This cluster assesses the infrastructure implementation by considering the emission amounts generated and their environmental impacts, in addition to the number of infrastructures implemented and the job creation impact related to Gini Index, poverty levels, gross domestic product, among others.
(11)	EN	It includes the emissions related to the raw material acquisition as well as its sustainable classification. Besides, the energy consumption links this cluster with clusters 4 and 5 to reach the energy balance calculation..
(12)	EC–EN	It involves the total emissions cost and the total emissions produced in all the supply chain stages.

shows.

Concerning the cluster assessment, we note that each cluster involves a different stage in the supply chain. For instance, cluster 8 includes the infrastructure and technologies implementation for production operations, while Cluster 10 evaluates the impact of this implementation. Then, cluster 11 considers the provisioning stage, while cluster 4 evaluates the transport in the entire supply chain. Cluster 3 includes the consumables necessary for production, such as water, fuel, and raw materials, while Cluster 9 measures the emissions generated in production. Clusters 1 and 2 measure emissions of waste and wastewater generated in the production stage. Clusters 5 and 6 refer to the distribution of products by measuring customer satisfaction and surplus. Finally, cluster 12 measures the costs of all emissions generated in the supply chain, while cluster 8 assesses the financial aspects of the supply chain. Then, this distinction of metrics by cluster allows us to distinguish what material flow and information (parameters) are required to assess sustainability in each section or stage of the supply chain.

Regarding the sustainability dimensions involved in each cluster, we remark that the environmental and economic dimensions are in eleven (92%) and six (50%) clusters, respectively. It shows the importance of environmental and economic dimensions in the SSCD and their relation with the other dimensions to be evaluated. The social and political dimensions are in four (33%) and two (17%) clusters, respectively. In contrast, the technological dimension is in only one cluster (8%), evidencing the incipient assessment and relations of these aspects in SSCD. Furthermore, considering the interactions among the sustainability dimensions, five (42%) clusters simultaneously assess the economic and environmental metrics. Meanwhile, four (33%) clusters integrate environmental and social metrics, while three (25%) only assess environmental metrics. It defends the hypothesis of the possibility of finding similarities between the metrics, grouping them despite belonging to different dimensions of sustainability. In addition, it shows the strong interrelationship among all the sustainability pillars, which reinforces the need for a holistic assessment of sustainability.

Additionally, at larger nodes, which represent a larger number of function and parameter aggregations, we can highlight metrics such as TEIC. It aggregates the environmental impacts by categories, including those associated with wastewater, waste, transportation, and production. Similarly, TEmCost considers the costs associated with the emissions generated throughout the entire supply chain, also accounted for as emissions in the TTESC metric. TSI represents the social impact and includes the impact on food safety, infrastructure redundancy, accidents associated with production technologies, the impact of implementing facilities according to the selected geographical location, and the impact associated with the fixed and variable work generated. Finally, TSW represents social welfare and incorporates several sustainability dimensions by evaluating the importance of net present value, return on investment, environmental impact, impact on human health, and social impact, among others. By including various metrics of several sustainability dimensions, this is observed as an alternative to the inclusion of sustainability to all its extensions through weighting the metrics it incorporates.

3.3. The Aggregation Criteria and Reduction Rules

Then, to understand which metrics are a particular case of another, we have developed Figure A2 in Appendix C. It separates the metrics by level, increasing in level as metrics are added. Then, we have a set of five metrics representative of sustainability as follows: (1) total social welfare (TSW), (2) total products obtained with incipient technologies (TPIT), (3) total raw materials acquired from sustainable suppliers (TRMSS), (4) total sustainable raw material used (TSRM), and (5) total governmental expenditures (TGE). Note that TSW integrates: net present value (NPV), return over investment (ROI), total social impact (TSI), total environmental impact (TEI), total human impact (THI), total consumer satisfaction (TCSat), total consumer surplus (TCSur), and total implemented capacity use (TICU). This reduced number of metrics to consider when integrating sustainability in the SSCD is a manageable number for both multi-objective optimization and decision-maker assessments. Furthermore, these five metrics at the higher level of integration consider the five sustainability dimensions: Economic, Social, Environmental, Political, and Technological. Finally, note that TSW could be the only metric assessing sustainability by considering weights to integrate TPIT, TRMSS, TSRM, and TGE.

4. Discussion

The proposed integrated metrics framework provides decision makers in the industry with a systematic approach to defining and integrating sustainability metrics in sustainable supply chain design (SSCD). This framework will allow decision makers to identify and prioritize sustainability metrics and facilitate decision making in SSCD.

The metrics assessment based on aggregation criteria and cluster analysis (CA) method offers an integrated view of the relationship between the metrics and the sustainability pillars. It reveals the holistic nature of sustainability and indicates that the sustainability dimensions should not be analyzed separately but as a whole. This task is complex to perform if we consider the different sustainability measurement guides or even the UN sustainable development goals, which consider a large number of indicators to be evaluated. In the SSCD context, the large number of metrics found in the related literature show this complexity. Therefore, the reduced group of metrics proposed to measure sustainability will simplify the process of measuring sustainability in SSCD and reduce the burden of considering an unmanageable number of metrics. It will support and facilitate sustainability management in supply chain design for decision-makers in the industry.

In addition, this proposal made tractable the SSCD problem from an optimization point of view since it enables researchers and practitioners to design optimal sustainable supply chains through the typical multi-objective solution methods to evaluate five objective functions.

The proposed framework lays the basis for novel decision-aid models for production systems and logistics design. Because this research was focused on the strategic decision-making level, further research could assess the ex post assessments following the proposed methodology to identify and integrate the sustainability metrics.

5. Conclusions

This paper proposes an integrated, tractable, and representative metrics framework to measure the five sustainability dimensions in the sustainable supply chain design. This research has been based on an exhaustive and systematic literature review, multivariate relational statistical techniques, and reduction rules. To our best knowledge, this report describes a novel approach that has not been followed in previous research in a sustainable setting.

In the review process, 541 research articles were analyzed in depth, where most of the literature assesses strategical decisions by applying optimization as the principal methodological approach. Other topics observed from the literature review allowed us to expect a clear linear research trend for evaluating sustainability aspects in the SSCD, identifying that the principal research countries seeking SSCD are Iran, China, and the United States of America, which are focused mainly on the automotive sector and consumer goods production. Furthermore, the sustainability dimensions most studied are economical and environmental. Fifty-one metrics to measure sustainability in the SSCD are described based on the literature review. Among these, 16 correspond to the economic aspects, 15 to environmental, 12 to social, and 4 to political and technological dimensions. They can be understood as objective functions to be optimized, considering optimization is the most applied methodology. From the sustainability metrics recognized in the literature, we identify parameters and auxiliary functions by applying the aggregation criteria. Then, the cluster analysis obtained 12 clusters showing the strong interrelationship among the sustainability dimensions. Finally, following the reduction rules, a reduced number of 5 objective functions to measure sustainability in the SSCD is proposed, evidencing the measure of social welfare as a potential metric to integrate all sustainability dimensions.

Consistently, interesting practical and policy implications emerge from the research. Firstly, it reveals the exponential growth of SSCD-related research since formulating the Sustainable Development Goals in 2015. As a result, it has led to an unmanageable number of metrics to consider when integrating sustainability into supply chain design. Secondly, the research proposes a limited set of metrics that make optimization tractable through different methodologies to solve the SSCD multi-objective problem. Thirdly, the proposed limited set of metrics facilitates decision making for stakeholders by reducing the number of indicators to observe to make a decision. This research has important implications for supporting the integration of sustainability in productive sectors by providing a managerial-level understanding and allowing the development of optimized supply chain structures for sustainability.

The proposed methodology provides a systemic framework to incorporate additional metrics or objective functions. Hence, considering this research work conducted a literature review up to December 2020, it is advisable to conduct periodic updates every five years.

For future research, some associated research questions are proposed to be addressed, which could facilitate the sustainability measure and analysis in the design problem of a sustainable supply chain:

• How do we integrate the different objective functions in an index/value of sustainability in SSCD?
• How do we define a validation process for it?

The first question invites us to study and analyze these research results from multi-objective and many-objective optimization perspectives to obtain an index/value of sustainability in SSCD, considering the unique features of each productive sector. It requires analyzing and evaluating the five metrics found with a higher level of integration since they could be integrated into a unique metric by weighting them according to their relevance. Moreover, the relevance of each metric could vary depending on the production sector (energy, waste, water, and others) and the organizational setting. This leads to the second question, which is about defining a validation process based on historical management reports and expert knowledge from relevant actors such as government authorities, industry, and the community.