Smart green supply chain management: a configurational approach to enhance green performance through digital transformation

2.1 Configuring the smart supply chain dimensions through digital transformation

The study of digital transformation in SCM has been addressed from several streams that used different conceptual labels for this phenomenon. For instance, some authors have investigated the concept of Industry 4.0 in SCM (Manavalan and Jayakrishna, 2019) and the resulting Supply Chain 4.0 (Frederico et al., 2020), the development of Internet of Things (IoT)-enabled supply chains (Ben-Daya et al., 2019; Birkel and Hartmann, 2019; Paolucci et al., 2021) or digital/digitized supply chains (Büyüközkan and Göçer, 2018). We follow the extensive literature review of Meindl et al. (2021) that shows the interconnection of these different concepts under a major label called smart supply chain. Smart supply chain congregates all these perspectives that have the common aim to leverage the operations of the supply chain through the adoption of digital technologies, especially those related to base digital technologies (i.e. generic digital technologies that are used in different applications and domains). These include IoT, cloud computing, big data and artificial intelligence (AI) (Frank et al., 2019; Meindl et al., 2021), and, more recently, also blockchain (Cole et al., 2019). In summary, smart supply chain is a technological construct that results from the digital transformation of SCM (Benitez et al., 2022).

The vertical configuration of the smart supply chain has been studied through two different levels, strategic and operational (Benitez et al., 2022). The strategic level of a smart supply chain comprises the digital transformation perspective that companies want to follow for their SCM system, considering the necessary alignment of supply chain goals with real-time data flow (Benitez et al., 2022; Sturgeon, 2021). Based on this, the digital transformation strategy considers the extent of digital transformation’s impact on the company’s view of its SCM (Nasiri et al., 2020). For instance, General Electric (GE) created a large supply chain spanning an IoT platform focused on short-term results and using technologies whenever possible. Unfortunately, the company did not achieve the desired results. Its share price plummeted because GE Digital lacked a clear and objective strategy to link technologies with outcomes and performance indicators (Davenport and Westerman, 2018). Therefore, this combined approach between digital technologies’ potential and a targeted implementation in the supply chain enables the smart supply chain to promote data flow in SCM activities (Pasi et al., 2020; Shao et al., 2021).

The operational level of the smart supply chain can be divided into two technological layers: base digital technologies and front-end technologies (Frank et al., 2019; Meindl et al., 2021). Base technologies (i.e. generic cross-cutting technologies useful in several domains and applications) for smart supply chain allow sharing of real-time data to make faster decisions and transactions. These base technologies configure technologies, such as IoT, cloud computing, big data analytics, AI and blockchain, which are useful for multipurpose in the SCM activities (Frank et al., 2019; Nasiri et al., 2020). When these base technologies are used as stand-alone technologies in the smart supply chain, they are concerned with the data flow in the supply chain for better integration (Aryal et al., 2018). Thus, the interrelationships between buyers and suppliers, as well as aspects such as supply chain design, are changed through the use of base technologies, which provides more transparency to this integration (Frank et al., 2019). In this regard, supply chain transparency is vital to enabling supply-chain spanning potentials for data analysis and optimization (Müller et al., 2020).

Moreover, these base technologies can be run in the back-end to enable the operation of front-end technologies. Front-end technologies are defined as technologies that execute operational tasks like material handling or quality control of inputs in supply chain operations. Some of them are collaborative robotics, computer simulation, augmented reality or three-dimensional (3D) printing (Dalenogare et al., 2018; Hohn and Durach, 2021). These technologies optimize supply chain processes, such as smart machines for packaging or sensor technologies, ensure logistics specifications and avoid damaged goods (Birkel and Müller, 2020; Frank et al., 2019).

2.2 Configuring the GSCM through external–internal connections

In contrast to general SCM studies, which tend to focus predominantly on market and financial results, the GSCM field also considers social and environmental metrics to understand green performance, as in the studies from Appendix 1. Overall, the GSCM goal is to integrate environmental concerns into the entire supply chain, configuring the traditional operational processes like manufacturing and purchasing and relationships with partners into sustainable practices (Hsu et al., 2016; Laari et al., 2016). When companies look at green practices in SCM, they can align their operations management process to achieve green performance goals and develop GSCM systems (Abdel-Baset et al., 2019; Wu and Pagell, 2011).

As the interplay of internal and external processes is highly relevant to the digital transformation of supply chains (Frank et al., 2019), there is a need for studies that use theories that show how to interrelate green supply chain internal and external activities to achieve enhanced performance. Our five-year literature review on digital transformation and GSCM is evidenced in Appendix 1. Most of the studies focused on the use of specific technologies such as blockchain, IoT or additive manufacturing to support green operations or on the factors and barriers to implementing digital transformation in green operations. However, extant literature lacks evidence on the join implementation of both strategic and technological levels of digital transformation in several dimensions of GSCM. The literature in Appendix 1 acknowledges that strategic and technological dimensions set digital transformation and that GSCM requires internal and external green operations activities. These activities can be supported by digital transformation, but evidence for this is sparse and fragmented in several studies. Furthermore, a holistic and integrated view on Smart GSCM systems is still missing. Therefore, we adopt a configurational theory to explain the configuration and alignment of external activities with suppliers and customers with internal activities, e.g. manufacturing processes, to achieve higher performance (Hult et al., 2006). We follow a horizontal external–internal configurational approach rather than individually studying the supply chain elements. Based on this, we propose that internal and external green supply chain activities should be interrelated to support firms in achieving green performance. Such an integrative view from external and internal activities is relevant for digital transformation. This is because digital technologies have one of their main aims to integrate data flow from different organizational levels (Frank et al., 2019).

According to Zhu et al. (2013), some companies have already understood how to interrelate green relationships (external configuration) and green operations (internal configuration) into a GSCM system to achieve higher green performance. These companies started green internal practices before expanding their activities to external relationships at GSCM because they can have a higher control and management of their green practices. However, implementing or developing solely internal green operations requires large efforts to achieve green performance. Companies depend on external activities such as material, components and subcomponents and environmental compliance with customers in the product development process (Seuring and Müller, 2008). Companies are further dependent on interconnecting internal and external activities to gain green performance (Birkel and Müller, 2020). For instance, reducing resource and energy consumption should be aligned among purchasing, manufacturing and packaging processes to achieve better results, but this will also depend on the quality of inputs received from supplies.

Despite studies in GSCM exploring how external and internal activities are important for performance separately (Appendix 1), there is still a lack of studies investigating how these external and internal levels can be interrelated for the improvement of green supply chain results. Therefore, we integrated the configurational approach to comprehend how the interrelated activities between internal operational activities (green packaging, green manufacturing and green purchasing) (Hsu et al., 2016), and external relationship activities (green supplier relationships and green customer relationships) (Laari et al., 2016) could improve green performance. We aim to understand how this configurational approach of GSCM can be supported by the digital transformation perspective of smart SCM to configure a holistic Smart GSCM that enhances firms’ green performance.

3.1 Smart supply chain and green performance

The literature of digital transformation in GSCM described in Appendix 1 has already acknowledged the use of concepts such as Industry 4.0 (Belhadi et al., 2021), digital technologies (Bechtsis et al., 2021) or digital transformation (Zekhnini et al., 2021) connected to green performance. Several studies identified in the literature of Appendix 1 use a broad approach by connecting the digital transformation issue with sustainable operations rather than only environmental issues (green performance). On the other hand, they tend to focus on specific parts of the digital transformation, either on the strategic side or on the technologies adopted. We use an integrative perspective of what we call smart supply chain to investigate its impact on green performance. In this view, smart supply chain comprises the configuration of three different hierarchical layers: digital transformation strategy, base digital technologies and front-end technologies, and each of them could have contributions to increasing green performance.

The digital transformation strategy can create new business models that provide a new green perspective for the companies to use the generated data toward a strategic position as a green performer in the market (Birkel and Müller, 2020; Hohn and Durach, 2021). Furthermore, besides the digital strategy, base technologies are crucial for green performance because cutting-edge technologies like IoT, cloud, big data and AI facilitate companies’ data flow and management. Several studies in the literature (Appendix 1) have concentrated their focus on this digital dimension, suggesting that tools like blockchain or cloud systems can improve transactions and monitoring of the green practices, which results in better performance (Cole et al., 2019; Gong et al., 2022). Furthermore, as suggested by Frank et al. (2019), data-driven technologies enable the implementation of front-end technologies. These front-end technologies (e.g. simulation, 3D printing, augmented reality) are responsible for smart supply chain implementation, enabling digital platforms with customers and suppliers to maximize green performance, e.g. virtual testing instead of physical prototypes. When companies use front-end technologies strategically, they can also create new industrial applications, improving their competitiveness and promoting their sustainability success (Frank et al., 2019; Zhao et al., 2010). Hence, we propose the following general hypothesis to represent these three dimensions of the smart supply chain:

3.2 Smart supply chain as an antecedent of GSCM

Even though smart supply chain plays an important role in green performance, it also depends on the green supply chain configuration. Companies should structure their external and internal practices to align with their environmental goals (Sarkis et al., 2010; Wu and Pagell, 2011). Hence, according to the configurational approach, the best results in green performance cannot be reached in a supply chain only by using digital strategy and technologies. It also needs to be aligned between external activities (green relationships) and internal activities (green operations). Without this alignment, companies are improving their processes and practices without a clear perspective. Consequently, they deliver short-term transitory results rather than implementing changes that will improve long-term and lasting sustainable results (Wu and Pagell, 2011). Hence, companies need to rethink how they can organize green relationships and green operations in conjunction with smart supply chain to promote green results (Birkel and Müller, 2020). Companies are developing new digital strategies focusing on improving their data flow, adopting base technologies like cloud services to facilitate the information sharing between their process and adopting front-end technologies to improve processes themselves (Frank et al., 2019; Wu and Pagell, 2011). Thus, companies can reach a higher level of competitive advantage toward sustainability when they join efforts to implement smart supply chain and GSCM together.

As smart supply chain consolidates the use of base technologies to improve relationships, companies are using base technologies to collect and analyze data that could create new opportunities for data management and integration towards sustainability (Frank et al., 2019; (Wilhelm and Villena, 2021). In addition, front-end technologies can enhance green relationships with suppliers because companies can use these technologies to better select green supplies. On the other hand, green operations could enhance companies’ performance when driving digital transformation in their supply chains. Providing information using digital platforms with suppliers can help companies develop green digital purchasing, offering new reliable, accurate information about the suppliers and explaining how they implement green practices into their supply chain (Birkel and Müller, 2020; Narayanamurthy and Tortorella, 2021). Furthermore, to deliver the products to eco-friendly customers, companies seek to create new green packages, using as few materials as possible, which can be supported by sensors connected via the IoT, to enhance such packaging solutions (Birkel and Müller, 2020; Kumar et al., 2021; Narayanamurthy and Tortorella, 2021).

Considering the contributions explained by smart supply chain to green relationships and green operations, we propose the following hypotheses:

As we follow a horizontal external-internal configurational view of GSCM, we hypothesize the relationship between green relationships and green operations, being both parts of the GSCM system. In the configurational view of SCM, external activities support internal operations (Flynn et al., 2010), which can be extended to GSCM. For instance, green suppliers can encourage the use of renewable resources in the supply chain. At the same time, eco-friendly customers can demand new green or remanufactured products to mitigate the carbon footprint (Abbey et al., 2015). These requirements from the stakeholders can stimulate new green practices (Zhu et al., 2013). One of such practices is the demand for ISO 14001 certification from suppliers when companies look for supplies (Laari et al., 2016; Hsu et al., 2016). Although green purchasing will prioritize renewable resources for the companies’ processes, companies should also rethink how they manufacture their products regarding environmental impacts. Green manufacturing can be implemented by pressure from green suppliers and customers (Zhu et al., 2013), enabling them to create converging practices to integrate and align sustainable manufacturing with GSCM. Furthermore, supply chain actors can work together to improve their collective recycling systems, improve green product development and enhance recycling technology (Rahmani et al., 2021). Based on the GSCM configuration, green packaging is another important practice influenced by suppliers and customers. For instance, biodegradable packaging has been a requirement from society and governments in the past few years, but its cost is still not competitive for some companies. Therefore, restaurants like iFood and Uber Eats offer green packaging that consumers are willing to pay for, promoting new green practices in packaging operations (Zhu et al., 2013).

Considering these aspects, we propose the following hypothesis:

While green relationships and green operations can have stand-alone contributions to green performance (Hsu et al., 2016; Laari et al., 2016), the configurational view also considers relationships and alignment between the internal and external structure of the organizational system (Flynn et al., 2010), as represented in Figure 1, which results in the contribution of GSCM on green performance, as hypothesized:

4.1 Sampling

We conducted a cross-industry survey with experts in SCM. The target respondents were top executives and managers or directors with strategical decision-making roles in Brazil, many representing foreign multinational companies. The initial sample was composed of approximately 1,500 executives from all Brazilian regions who participate in the Brazilian Council of Purchasing and Supply Executives. We did not select any specific profile of respondents to reduce selection bias and increase the randomization of the sample. However, because companies more engaged in the questionnaire topics are more likely to participate (self-selection bias), we adopted the strategy of sending the invitation through a specific customized channel of the association (which elevates the likelihood of different types of companies to respond). We also conducted post hoc tests for endogeneity and self-selection bias (Section 4.5). Our questionnaire was sent three times to our target respondents via e-mail from the beginning of February to the end of March 2021. We obtained 615 answers with 473 final useful responses (31.5% response rate). Our final sample is composed of the majority (78%) of large companies (more than 500 employees), followed by 17% of medium-sized companies (100–500 employees) and 5% of small companies (less than 100 employees), according to the classification of the Brazilian Institute of Geography and Statistics (IBGE, 2015). The overall respondent profile was essentially of directors (40%), managers (36%), coordinators or supervisors (13%), president/vice/CEO (6%) and owners (5%). Table 1 details all sample compositions in our study.

Regarding revenue, more than 65% of our sample is composed of manufacturing companies that earn more than US$104m per year. Therefore, our sample is characterized by significant players in manufacturing industries. Approximately 25% of manufacturing companies earn between US$19m and US$104m, and only around 7% achieve less than US$19m. Moreover, considering the time of foundation, most companies (around 96%) have more than 10 years of experience in the market, while 4% have less than 10 years of experience, characterizing a relative mature business life, which is important when considering companies with consolidated supply chains as expected in our study. Hence, especially in an emerging economy like Brazil, we argue that enterprises with large revenue and experience in the market are more likely to invest in green processes to achieve new performance levels (Frank et al., 2016; Cousins et al., 2019; Nara et al., 2021). Thus, we opted to choose companies that are already consolidated in the market as those represented by the business association investigated in our sample. In addition, we also analyzed the regions of the sample. Almost 90% (75% from the Southeast and 14% of the South) of our sample are composed of manufacturing companies that are in the most industrialized area of the country (Southeast and South of Brazil) (Marodin et al., 2017).

4.2 Measures and survey instruments

The questionnaire was developed from well-established constructs in extant literature. The items used to measure each construct and respective references are presented in Appendix 2. We also present which hypotheses each construct used in Appendix 2. For digital transformation, we used three constructs nominated [DT_STRATEGY], [BASE] and [FRONT_END]. The first, known as digital strategy, uses a five-item scale that comprises the strategical aspects of digital transformation, e.g. to create a stronger communication network between different sectors (Nasiri et al., 2020). We also considered the exchange of information, collecting large amounts of data, and improving the interface with customers through digitalization as part of the digital strategy (Nasiri et al., 2020). For the second [BASE] and third [FRONT_END] constructs, we followed Frank et al. (2019), who presented these nomenclatures for digital transformation technologies. For base technologies [BASE], which are cross-cutting or general-purpose technologies that organizations need to adopt along with their systems when they want to implement digital transformation. Base technologies (IoT, cloud computing, big data analytics, AI and blockchain) have a comparably broad focus and field of application. They represent the foundation for digital transformation and provide the intelligence for the systems, enabling the “smart” operations in supply chains with more specific and focused technologies (Frank et al., 2019). To measure [BASE] we used a formative construct based on a five-item scale that considers digital technologies like the IoT, cloud computing, big data analytics, AI and blockchain (Frank et al., 2019; Meindl et al., 2021). We also included in this construct blockchain as it is considered a key digital technology that provides different applications for supply chains (Kouhizadeh et al., 2021). In the case of front-end technologies [FRONT_END], we included collaborative robots, computer simulation, augmented reality and 3D printing. We based this decision on Frank et al. (2019) and Meindl et al. (2021) studies, who explain that these frond-end technologies support the development of the smart supply chain. We adopted a formative scale to represent these two constructs (BASE and FRONT_END) because each of these technologies provides different but complementary applications that constitute the digital operations of the companies (Coltman et al., 2008) while they all contribute to providing an advance digital operation. Thus, these two constructs were built through a composite measure using the sum of the items.

For GSCM, we used five constructs divided into green relationships [SUPPLIER; CUSTOMER] and green operations [PACKAGING; MANUFACTURING; PURCHASING]. The green supplier relationship includes cooperation with suppliers for eco-product design, green input logistics development, mutual understanding of environmental issues responsibilities and suppliers’ preference for an environmental management system (Laari et al., 2016). Green customer relationship considers cooperation with customers for eco-product design and mutual understanding of responsibilities in environmental issues. The construct also has items related to customers’ demands, like information on the company’s environmental compliance, ensuring sustainable practices of the company's suppliers and the request to implement an environmental management system like ISO 14000 (Laari et al., 2016). Green packaging includes the company’s reusable packaging, the use of fewer materials, encouragement of the use of reusable packaging and promotion of packaging recycling and reuse programs (Hsu et al., 2016). For green manufacturing, we considered assessing the environmental impact of developing and improving products, product development with recyclable raw material, low use of resources, low impact on the environment and high life span. In the case of green purchasing, we measured this construct with four items, including purchases based on environmental specifications established by product design and purchases conducted with ISO 14001 certified partners (Hsu et al., 2016). We also included minimized environmental impact in purchasing procedures and a purchasing process that follows product labeling standards to minimize environmental impact as items for this construct (Hsu et al., 2016; Green et al., 2012; Zhu et al., 2008).

Regarding the dependent variable, we used one main construct named green performance [GREEN_PERF], which includes performance metrics related to environmental impact in the last three years. We selected this construct as a final measure because our goal was to analyze the influence of digital transformation on the green operations and relationships in GSCM. For analyzing how green metrics are influenced in a GSC configured with interrelated relationships, we used a scale that considers an increase in material recycling, reducing of emissions, reducing of use/waste of resources and decreasing the use of hazardous and environmentally harmful materials (Cousins et al., 2019; Zhu et al., 2008). We measured all the constructs’ items using a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).

Usually, the control variables used in surveys are company size and the industrial sector. However, considering our context, we aim to understand if there is a department responsible for sustainable operations to understand GSCM. Hence, we included the existence of an area/department responsible for sustainable operations in the supply chain inside the firm [GO-area: 1 = yes; 0 = no]. In addition, we also controlled the firm’s size, measured by the number of employees into two main dummy categories [firm_size: 1 = large; 0 = small or medium], following the Brazilian Institute of Geography and Statistics classification (IBGE, 2015).

4.3 Variable operationalization, reliability and validity of measures

To examine unidimensionality, we performed confirmatory factor analysis (CFA). Our model showed the goodness of fit as the reference values for comparative fit index (CFI), Root mean square error of approximation (RMSEA), average variance extracted (AVE), composite reliability (CR) and Cronbach’s alpha fell in the acceptable values (Hair et al., 2018), as shown in Appendix 2. In addition, overall constructs showed high factor loadings, the exception of one item for [MANUFACTURING], which we opted to keep as the construct is already validated in the literature (Hsu et al., 2016) and a general test including all constructs of Green Operations showed the goodness of fit (RMSEA = 0.072; CFI = 0.925). We also measured Green Relationships (RMSEA = 0.081; CFI = 0.971) and Smart Supply Chain (RMSEA = 0.081; CFI 0.947) constructs and obtained goodness of fit in our measures. Also, the final and complete model including all constructs reported goodness of fit (RMSEA = 0.054; CFI = 0.913; Δχ² = 1676.17).

We also checked discriminant validity using a series of two-factor model estimations (Bagozzi et al., 1991). As recommended by Cable and DeRue (2002), we performed pairwise comparisons amid CFA models for each construct, looking for their respective goodness of fit. In the first step, the correlation between the two constructs was restricted to a unit. In the second step, the model restriction was freed, and we calculated the goodness of fit for the original constructs. In this test, all the results showed discriminant validity (Δχ² > 3.84, p < 0.01), evidencing our constructs are measured with theoretically different concepts (Bagozzi et al., 1991).

Finally, we assessed the normality of our data by examining the skewness and kurtosis values. The results suggest that our data is normally distributed since all values were between the thresholds of ±2.58 (α = 0.01) (Hair et al., 2018). We also analyzed the means, standard deviations and correlations for our constructs and control variables incorporated in the model. Appendix 3 summarizes all descriptive statistics, as well as normality and correlations.

4.4 Response bias and common method variance

Considering the potential bias in our survey design, we used a series of procedures and statistical remedies to attenuate this potential issue (Podsakoff et al., 2012). Moreover, because our survey is composed of single respondents, common method variance (CMV) may be a concern (Podsakoff et al., 2012). As an initial procedure, we pretested the questionnaire with 20 executives to verify the clarity of our survey instrumentation. The chosen executives were all related to GSCM and knowledge of digital transformation technologies and strategies. Then, the order of questionnaire blocks (e.g. green operations, green relationships, digital transformation and performance metrics) was randomized to prevent potential associations between the variables. To test the magnitude of CMV, we performed several statistical tests using RStudio. First, we performed Harman's single factor test in which a single factor loads on all measured items from our model. If the total variance extracted by one factor exceeds 50%, common method bias is present in the study (Podsakoff et al., 2003). The single factor explained 32.61% of the total variance, indicating CMV is not a concern in our study. However, because this approach is too simple and other authors (Williams et al., 2010; Simmering et al., 2015) recommend other approaches to measure CMV, especially for single respondents, we also performed the marker variable technique. The marker variable technique considers adding a variable to the survey, which is expected to be theoretically unrelated to the substantive variables measured in the model (Lindell and Whitney, 2001). We used “logistics operations during the pandemic” as a marker, which is used to measure how respondents felt the impact of their logistics operations during the pandemic. We initially included this variable in the correlation matrix with all variables in our model. To GO_ AREA and DT_STRATEGY, the marker variable presents p-value < 0.05, even though this happens with two variables, CMV was not a concern in our survey. Moreover, we included this item in all estimations necessary for hypothesis testing, and the results were compared with the outputs without markers. The results remained stable with adding a marker variable, which means that there were no significant changes in the models. Hence, we concluded that response bias should not be a concern in this data set.

4.5 Endogeneity and robustness checks

Because endogeneity and self-selection bias can jeopardize and biased regression results, we also assessed this in the model (Bascle, 2008). There may be an endogeneity effect on the digital transformation process as companies may have been forced to adopt digital strategies because of the pandemic. At the same time, respondents can self-select to respond to the survey because the companies are more engaged in GSCM. To test endogeneity and self-selection bias, we ran a two-stage least squares (2SLS) regression approach in Stata 16. We instrumented all independent constructs in our model during our hierarchical regression procedure stages. We chose four items related to COVID-19 impact (i.e. demand impact, production impact, purchasing and revenues) in the industrial activities to instrument our independent constructs. COVID-19 has accelerated the adoption of digital technologies and has impacted on the supply chain operations (Zimmerling and Chen, 2021). Therefore, we considered this effect a potential instrument for choosing digital transformation development and green operations and relationships management within a supply chain configuration. According to our initial tests, the explanatory variables showed that our instruments are strong (i.e. the p-values < 0.001, and the minimum F-value is 14.41, p = 0.000). Therefore, we verified whether the explanatory variable should be treated as endogenous and need to be instrumented as proposed in the 2SLS regression model. We performed Stata's estat endogenous procedure using Durbin and Wu–Hausman statistics to evaluate the consistency of our estimators. The tests showed that the hypothesis that the explanatory variable is exogenous could not be rejected during our regression estimation (i.e. all p-values > 0.05).

We instrumented all smart supply chain constructs in the general model [GREEN_PERF] with the other constructs, which comprise all associations and relationships in our study. The results for the instrumented constructs in the general model were: Durbin = 1.741, p-value = 0.627 and Wu–Hausman = 0.565, p-value = 0.638. For the validity of instrumental variables, we analyzed the Sargan χ² and Basmann χ² tests through Stata’s estat overid code which checks if the instruments are uncorrelated with the error term and correctly excluded from the estimated equation (Baum et al., 2003). Our results for the general model were: Sargan χ² test = 0.161, p-value = 0.687 and Basmann χ² test = 0.157, p-value = 0.691.

To ensure the general consistency of our model, we also performed a series of robustness checks. We explored how the results of our regressions analysis might vary using four distinct robustness approaches:

1. removal of control variables;

2. inclusion of a new construct;

3. individual analysis from predictors; and

4. inclusion of a competitive model.

In the first approach, we removed all control variables (GO_area and firm_size) to check if our predictors were not control variables’ artifacts. We found stable results because there were not significant changes in the coefficients of all models in our regression. For the second approach, we included a construct named green practices (RMSEA = 0.078; CFI = 0.986; AVE = 0.42; Cronbach = 0.72; CR = 0.98), including: use of environmentally friendly materials in packaging (0.43), purchasing of materials that do not harm the environment (0.75), assessment of suppliers' environmental compliance (0.70) and use of environmental training effectively in daily activities (0.68). We expected to find significant effects from the relationship between digital transformation and GSCM constructs since green practices are present in GSCM routines (Green et al., 2012; Zhu et al., 2008). The approach showed a significant direct effect of this new construct in all models, confirming the link between green practices, digital transformation and GSCM. For the third approach, we analyzed the individual effect of each construct in our models, and overall, we found consistency with our main findings presented in Table 2 in the Results section. We only have slight changes in [BASE] construct, which showed a positive association in our models when single evaluated. However, this reinforces our H2a and H2b because we found positive associations when [BASE] was single measured despite in the overall assessment with all variables, it did not show any positive and significant associations. Finally, we tested a competitive model for the fourth approach considering smart supply chain as a moderator between GSCM and green performance. However, we did not find statistical support for this competitive model.

4.6 Data analysis

We performed a set of hierarchical ordinary least squares regression models to test our hypotheses. We standardized our independent variables using a mean-centering Z-score to test all relationships. In the first stage, we examined all direct effects of the digital transformation phenomenon [DT_STRATEGY; BASE; FRONT_END] on green relationships [SUPPLIER; CUSTOMER]. In the second stage, we included [SUPPLIER] and [CUSTOMER] as independent variables and regressed them in green operations [PACKAGING; MANUFACTURING; PURCHASING]. In the last stage, we included green operations variables as independent and assessed them in green performance [GREEN_PERF]. We also checked the mediation effects from green relationships [SUPPLIER; CUSTOMER] on digital transformation for green operations [PACKAGING; MANUFACTURING; PURCHASING], and from GSCM on smart supply chain for [GREEN_PERF]. We performed the PROCESS macro from Hayes (2017) to assess the mediation effect. Our final model contains two control variables, eight independent variables and one dependent variable.

We checked the assumptions of normality, linearity and homoscedasticity for our regression analysis for all independent and dependent variables. We analyzed normality through the values of kurtosis and skewness (Appendix 2). Linearity was investigated by plotting the partial regressions for the independent variables, while homoscedasticity was visually examined in plots of standardized residuals against a predicted value. All these requirements were met in our dataset for regression analysis. Finally, we evaluated a possible multicollinearity issue for our independent variables, as Hair et al. (2018) suggested. In multicollinearity, regression estimates are unstable and have high standard errors. Our results indicate a low variance inflation factor (<3.5) for all variables far below the threshold of 10 (Hair et al., 2018).

To assess mediation effects, we calculated the indirect effects of the relationships as suggested by Preacher and Hayes (2008). We adopted the PROCESS analysis developed by Hayes (2017) to check our hypotheses (H2 and H4) associated with mediation. PROCESS analysis allows for a bootstrapping procedure to examine the conditional indirect effects, a more powerful procedure than Sobel’s z-test to test for mediation effects (Zhao et al., 2010). We set up 5,000 bootstrap samples as Preacher and Hayes (2008) suggested.

7.1 Implications for practice

This study contributes to the sustainable development goals proposed by the 2030 Agenda for Sustainable Development of the United Nations[1]. From the 17 goals proposed in the 2030 Agenda, our study is strongly aligned with Goal 12: Ensure sustainable consumption and production patterns. This goal is concerned with adopting practices that can help reduce the carbon and material footprint of companies and, consequently, nations. Our study shows that when companies configure a Smart SCM, they can improve several green performance metrics that contribute to this goal, such as increasing material recycling, reducing emissions and reducing the use of resources, including hazardous and environmentally harmful materials.

For managerial practice, this paper highlights the relationships between the smart supply chain and GSCM, leading to Smart GSCM. Companies that aim to increase green performance need to integrate external and internal GSCM practices, which directly affect green performance. At the same time, companies can apply digital technologies to achieve a higher degree of development of the configurational aspects of GSCM. Furthermore, pursuing a smart supply chain is not sufficient to achieve green performance, but the combination with GSCM is a vital factor in achieving a Smart GSCM. We thus recommend pursuing joint rather than single technological and operational approaches to increase green performance. The configurational view calls attention to how managers should integrate these three perspectives of SCM: digital transformation, external relationships and internal operations. Practitioners can use our model to guide the supply chain transition to smart GSCM through the adoption of base technologies that can later be combined with a digital transformation strategy and front-end technologies to support the configurational green supply chain structure. Using such a configurational Smart GSCM model combined with a correct assessment of digital technology investments can help establish a digital transformation journey in the supply chain (Almeida et al., 2022).

7.2 Limitations and future research

Our study has the limitation that it considers single responses from the companies. Thus, the study could be improved by combining responses and performance data from multitier supplier chain settings, including several operations and supply chain processes interconnected. Therefore, our study is the first step to larger research with more GSCM stakeholders. Further limitations include that the results are based on cross-sectional data. Hence, a long-term investigation presents a promising future research avenue. GSCM and smart supply chain represent emerging concepts that can be expected to develop significantly in the next years. Such an approach could include how smart supply chain is implemented across several supply chain tiers and how this process evolves over time.

Moreover, digital technologies can support supplier integration in the product development process (Ayala et al., 2020) or avoid glitches when such integration happens (Merminod et al., 2021), which can help develop green products aligned with the GSCM system of the firm. Future studies could advance in this direction of research. Finally, although most firms are active in multinational settings and engage in global supply chains, the study focuses on firms from or active in Brazil. Prior research in this market has shown that companies are less open to external collaboration with stakeholders due to industrial uncertainties (Frank et al., 2021). Also, investments in technologies are not well aligned with the expected innovation outputs compared to developed countries (Frank et al., 2016). In another study, Dalenogare et al. (2018) showed that Brazilian companies that invest in digital technologies have not as high priority sustainable goals as European countries. These are some differences that can produce different results when compared to companies from developed countries in the GSCM domain. In such countries, issues like open innovation with suppliers, technology investment to increase innovation outputs and expected benefits to achieving sustainable goals tend to be more consolidated. Thus, the effects of the statistical inferences found in this study should be considered limited to the national context. Future studies should expand such findings to global supply chains where these practices are more consolidated. Moreover, academics could devote effort to analyzing differences in digital and green practices between developed and emerging countries to find the gaps and show the main differences. This would be helpful to understand better the potential tensions, and ways companies in both types of contexts differentiate. Based on this, research policies can be proposed to reduce such gaps in managerial practices that can help achieve global sustainable goals.