Green marketing is a business strategy to meet consumer needs in an environmentally friendly way, aiming to reduce negative impacts while improving product quality. Its implementation is important for organizations to remain competitive, meet consumer expectations for sustainable products, and support environmental sustainability.

Consumer behavior refers to consumers’ attitudes, actions, and interests related to purchasing, using, evaluating, and disposing of goods or services, influenced by emotional, cognitive, and environmental factors (Kholik, 2020). This study not only focuses on purchase decisions but also on emotional and mental responses that help businesses understand consumer attitudes and market interests.

Anggraeny (2023) identifies indicators influencing the intention to use eco- friendly shopping bags, namely awareness of environmental benefits, commitment to environmental preservation, positive attitudes toward sustainable products, as well as social norms and external influences.

According to Chen & Chang (2012), green trust is consumers’ trust in green products through the fulfillment of obligations, product guarantees, services, and employee care. Such trust encourages consumers to purchase products that do not harm the environment (Yahya, 2022).

According to Amoako et al. (2020), green knowledge refers to understanding environmental impacts, the benefits of eco-friendly products, sustainable practices, and the ability to choose green products, all of which play a vital role in shaping consumer attitudes and behaviors.

Based on Fadilla et al. (2021), environmental concern includes awareness of pollution, the desire to reduce the impacts of consumption, attention to sustainability, and willingness to change lifestyles for environmental preservation. This reflects the level of individual concern and commitment to environmentally friendly practices.

According to Fadilla et al. (2021), green attitude is measured by the belief that using tumblers reduces plastic waste, a positive attitude toward eco-friendly products, the belief that tumblers are a responsible action, and the preference for choosing them over single-use bottles. These reflect awareness of environmental benefits and promote pro-environmental behavior.

This study employed the Partial Least Squares Structural Equation Modeling (PLS-SEM) method with the assistance of SmartPLS version 4.1 software to test the model of relationships among latent constructs measured through indicators developed in the questionnaire. PLS-SEM was chosen because it is suitable for studies with complex models, relatively limited sample sizes, and an exploratory approach aimed at developing and verifying theories (Hair et al., 2019; Ghozali & Latan, 2020; Hair et al., 2021). The data analysis process was carried out in two main stages, namely:

The first stage was the evaluation of the outer model, or measurement model, which aims to assess the validity and reliability of the constructs. This analysis involved testing the following:

Table 4.9 Convergent Validity & Construct Reliability Test Results

Source: SmartPLS 4.1 Output, 2025

The results of the outer model analysis show that all constructs meet the requirements of convergent validity and reliability. The majority of outer loadings are above 0.70, while Cronbach’s Alpha (α), Composite Reliability (CR), and AVE values exceed the minimum thresholds. Green Knowledge (X1) with GK1 and GK4 (0.60–0.69) remained acceptable, with α = 0.722, CR = 0.827, and AVE = 0.551. Environmental Concern (X2) was stronger (α = 0.806, CR = 0.873, AVE = 0.634), while Green Attitude (X3) was also high (α = 0.838, CR = 0.893, AVE = 0.680). Green Trust (Z) demonstrated the best results (α = 0.894, CR = 0.926, AVE = 0.759), and Intention to Use (Y) scored the highest (α = 0.916, CR = 0.941, AVE = 0.801). Therefore, all constructs are feasible to proceed to the inner model analysis.

Figure 4.1 illustrates the path relationships among the five research variables: Green Knowledge (GK), Environmental Concern (EC), Green Attitude (GA), Green Trust (GT), and Intention to Use (IU). The R² values indicate that 59% of the variance in GT can be explained by GK, EC, and GA, while 73% of the variance in IU is influenced by all four variables. In terms of contribution, GA has the strongest effect on GT (0.446), followed by EC (0.356), while GK contributes the least (0.089).

With respect to the intention to use environmentally friendly products, GA again emerges as the most dominant factor (0.409), followed by GT (0.381). In contrast, the direct effects of GK (0.138) and EC (0.061) are relatively small, suggesting that their roles are more prominent in shaping attitudes and trust rather than directly influencing intention. Overall, these findings underscore that positive attitudes and trust toward green products are the key drivers of pro- environmental product use behavior, while knowledge and concern act as supporting factors.

Discriminant validity ensures that each construct is truly distinct from others.

It is typically assessed using the following methods:

Table 4.10 Discriminant Validity Test Results – Fornell–Larcker Criterion

Source: SmartPLS 4.1 Output, 2025

The discriminant validity analysis using the Fornell–Larcker Criterion confirms that all constructs meet the requirement, where the square root of AVE is greater than the inter-construct correlations. For instance, √AVE for EC (0.796) is higher than its correlations with GA (0.589), GK (0.450), GT (0.659), and IU (0.616). Similar results apply for GA (0.824), GK (0.742), GT (0.871), and IU (0.895), all of which exceed their correlations with other constructs.

This finding affirms that each construct represents a unique concept without conceptual overlap, thereby satisfying discriminant validity. Consequently, the measurement model is valid and can be advanced to the inner model analysis.

Discriminant validity via cross loadings ensures that each indicator is more strongly associated with its intended construct than with other constructs. The results confirm that all indicators satisfy this criterion. For example, indicator EC3 loads 0.890 on Environmental Concern (EC), much higher than on GA (0.560), GK (0.446), GT (0.648), and IU (0.572). Similarly, GA3 records a loading of 0.940 on GA, higher than on other constructs, while other indicators such as GK2 (0.861), GT2 (0.908), and IU2 (0.936) also follow this pattern.

These results suggest that all indicators in the model demonstrate strong discriminant ability, accurately representing their intended constructs without overlap. This reinforces the reliability of the measurement instrument and supports both conceptual and structural validity of the measurement model, enabling the study to proceed to the inner model analysis.

Table 4.12 Discriminant Validity Test – HTMT Ratio

Source: SmartPLS 4.1 Output, 2025

The discriminant validity test using the Heterotrait-Monotrait Ratio of Correlations (HTMT), which is more sensitive than the Fornell–Larcker criterion in detecting potential construct overlaps (Henseler et al., 2015), indicates that discriminant validity is established when all HTMT values are < 0.90. The evaluation shows that all construct pairs meet this criterion. For instance, EC with GA (0.682), GK (0.564), GT (0.759), and IU (0.707), confirming that EC is distinct from the other constructs.

The highest values are GA–IU (0.877) and GT–IU (0.852), which, while approaching the threshold of 0.90, are still acceptable given their theoretically strong relationship. Thus, all constructs are empirically unique without overlap, and the measurement model is valid and reliable, allowing the study to proceed confidently to the inner model analysis.

The second stage is the assessment of the inner model, or structural model, which aims to evaluate the relationships among latent variables in the research model. The structural model analysis seeks to test the causal relationships between constructs and to assess the extent to which the model explains the dependent variables. Evaluation was conducted through several key statistical indicators, including multicollinearity, path significance, coefficient of determination (R²), effect size (f²), predictive relevance (Q²), and overall model fit.

Multicollinearity was assessed using the Variance Inflation Factor (VIF) for each path in the model. This test is essential to ensure that no multicollinearity exists among independent variables that could bias the estimation of path coefficients. Based on Hair et al. (2017), acceptable VIF values are below 5, while a more conservative threshold is VIF < 3.3.

Table 4.13 Collinearity (Multicollinearity) Test Results – VIF

Source: SmartPLS 4.1 Output, 2025

The results show that all paths in the model have VIF values within acceptable limits, with the highest being GT → IU (2.438) and GA → IU (2.214). Although these are the highest values, they remain well below the thresholds of 5 and the more conservative 3.3. Other paths, such as GK → GT (1.416) and GK → IU (1.435), are even lower, indicating no multicollinearity issues.

This confirms that each independent construct contributes uniquely without substantial overlap in explaining the dependent variables. Thus, the model is free from multicollinearity problems, parameter estimates are reliable, and the interpretation of construct relationships can be conducted accurately, enabling further structural analysis.

The relationships among constructs in the structural model were tested using the bootstrapping technique to obtain stable parameter estimates without normality assumptions. Three key indicators were considered: path coefficient, t- statistic, and p-value. A path is considered significant if t ≥ 1.96 and p < 0.05 at α = 5%. This test serves as the empirical basis for evaluating hypothesis support and for drawing theoretical and practical conclusions.

Table 4.14 Path Significance Test Results (Path Coefficients)

Source: SmartPLS 4.1 Output, 2025

Based on the results of Table 4.12, the path analysis using the Partial Least Square (PLS) method yielded the following findings:

Not significant, with an original sample value of 0.089, a t-statistic of 1.053, and a p-value of 0.293 (> 0.05). Thus, H1 is rejected. The rejection of H1 indicates that although individuals may have good knowledge regarding environmental issues, such knowledge does not automatically build trust in the benefits of using tumblers. This may be due to external factors such as limited direct experience, access to products, or lack of tangible evidence on the effectiveness of tumblers in reducing plastic waste, so that knowledge alone is insufficient to foster trust.

Significant, with a p-value < 0.001. Therefore, H2 is accepted, indicating that individuals’ concern for the environment directly shapes trust in environmentally friendly products.

Significant, with a p-value < 0.001. Thus, H3 is accepted, showing that a positive attitude toward environmentally friendly products contributes to building consumer trust in tumblers.

Significant, with an original sample of 0.381 and a p-value of 0.000. H4 is accepted, reinforcing the role of Green Trust in shaping the intention to use tumblers.

Significant, with a p-value of 0.025. Thus, H5 is accepted, showing that individuals’ environmental knowledge can directly influence their intention to use tumblers, even without mediation by Green Trust.

Not significant, with a p-value of 0.383. Therefore, H6 is rejected, suggesting that environmental concern, although high, is not sufficient to directly drive the intention to use tumblers. This finding highlights that concern must be channeled through trust (Green Trust) in order to affect behavioral intention, as evidenced in the mediation test of H9.

Significant, with a p-value of 0.000. Hence, H7 is accepted, showing that a positive attitude toward environmentally friendly products can directly encourage the intention to use tumblers.

Regarding the mediation test through the Green Trust (GT) construct, the results are as follows:

Not significant, with a p-value of 0.339. This means there is no mediation effect, and H8 is rejected. Thus, Green Trust does not mediate the relationship between knowledge and the intention to use tumblers. This is consistent with H1, which was also not significant, indicating that environmental knowledge tends to influence intention directly (H5) rather than through trust.

Significant, with a p-value of 0.009. Therefore, H9 is accepted, confirming that environmental concern can enhance the intention to use tumblers through the development of trust.

Significant, with a p-value of 0.003. Hence, H10 is accepted, indicating that a positive attitude not only directly influences intention but is also reinforced through the mediation of Green Trust.

This was analyzed to determine the proportion of variance in the dependent construct explained by the independent constructs. A high R² value indicates good predictive power of the model. In general, the interpretation of R² refers to the following criteria: R² ≥ 0.75 (substantial), 0.50–0.75 (moderate), 0.25–0.50 (weak), and < 0.25 (very weak). In this study, the R² values obtained for the main constructs indicate fairly good explanatory power, especially for Intention to Use, which is the primary dependent variable.

Table 4.15 Results of the Coefficient of Determination (R²) Test

Source: SmartPLS 4.1 Processing Output, 2025

Based on the interpretation criteria of the R-square (R²), this research model demonstrates fairly good predictive ability. The R² value for Green Trust (0.590) falls into the moderate category, meaning that 59% of GT variability can be explained by Green Knowledge, Environmental Concern, and Green Attitude. Meanwhile, the R² value for Intention to Use is 0.730, which also lies in the moderate category but close to strong, indicating that 73% of the variation in the intention to use environmentally friendly tumblers can be explained by the four constructs.

These results indicate that the structural model has high explanatory power for the intention to use environmentally friendly products, making it empirically reliable. Thus, the model is not only statistically appropriate but also strengthens the theoretical validity of the relationships among constructs in this study.

This was calculated to evaluate the individual contribution of each construct to the dependent construct. Effect size provides additional information beyond the R² value, especially in assessing the specific impact of each path. The f² values are interpreted as follows: f² ≥ 0.35 (large), 0.15–0.34 (medium), 0.02–0.14 (small), and < 0.02 (not significant). The f² values in this model show that some constructs have substantial effects, while others contribute on a moderate to small scale.

Table 4.16 Effect Size Test Results (f²)

Source: SmartPLS 4.1 Output, 2025

In this study, Green Attitude (GA) was found to have a medium effect on both Green Trust (GT) and Intention to Use (IU), each with an f² value of 0.280, making it the most influential factor in the model. Environmental Concern (EC) also had a medium effect on GT (f² = 0.193), but its direct impact on IU was very small (f² = 0.007) and insignificant.

Meanwhile, Green Knowledge (GK) only had a very small effect on GT (f² = 0.014), below the minimum threshold of 0.02, indicating weak contribution. Thus, GA occupies the most dominant position in explaining both GT and IU compared to the other constructs.

Obtained through the blindfolding procedure, which assesses how well the model can predict systematically omitted observed data. A Q² value > 0 indicates that the model has adequate predictive relevance. The higher the Q² value, the better the model’s ability to replicate actual data values. In this study, all constructs showed positive Q² values, reinforcing the conclusion that the model has good predictive relevance.

The Q²predict value is used in PLS-SEM analysis to measure the model’s predictive ability for endogenous constructs. According to Hair et al. (2019), Q² > 0 indicates predictive relevance, with the general interpretation as follows: Q² > 0.35 (strong), 0.15 < Q² ≤ 0.35 (moderate), and Q² ≤ 0.15 (weak).

Table 4.17 Predictive Relevance Test Results (Q²)

Source: SmartPLS 4.1 Output, 2025

Based on the results, the Green Trust (GT) construct obtained a Q² value of 0.550, while the Intention to Use (IU) construct had a Q² value of 0.643. Both values are well above the 0.35 threshold, indicating that this research model has very strong predictive ability for the two main constructs.

These findings not only strengthen the structural validity of the model but also confirm that it can effectively perform out-of-sample prediction. This means that the model is relevant for projecting consumer behavior in the context of using environmentally friendly products such as tumblers, influenced by variables such as green knowledge, environmental concern, green attitude, and green trust. Thus, the Q²predict values provide strong support for the reliability of the model in explaining and predicting sustainable behavioral intention in the future.

This was conducted by reviewing the R² values of the dependent constructs and assessing the overall path structure in explaining the studied phenomenon, namely the intention to use environmentally friendly products such as tumblers. The substantial R² values of the key constructs indicate that the model has a strong theoretical structure and solid empirical support, leading to the conclusion that the proposed structural model is suitable for hypothesis testing and further interpretation.

Table 4.18 Model Fit Test Results

Source: SmartPLS 4.1 Output, 2025

The evaluation results of model fit with PLS-SEM show an SRMR value of 0.084 for both the saturated and estimated models, which is below the 0.10 threshold and therefore acceptable (Hair et al., 2021). The low values of d_ULS (1.466) and d_G (0.854) also support the adequacy of the model fit.

Although the Chi-square value is relatively high (431.530), this indicator is not the main focus in PLS-SEM. Meanwhile, the NFI value of 0.732 is still considered acceptable for exploratory research (Henseler et al., 2016). Thus, the structural model in this study meets the model fit criteria and can be reliably used for further hypothesis testing.

The results of the first hypothesis test (H1) show that green knowledge has no significant effect on green trust (original sample = 0.089; t-statistic = 1.053; p- value = 0.293), hence H1 is rejected. This means that although respondents possess knowledge about environmental issues and eco-friendly products, it does not automatically increase their trust in products such as tumblers. This finding differs from Chen & Chang (2012), who stated that consumer understanding forms the foundation of green trust. In this study, knowledge alone appears insufficient due to skepticism toward “eco-friendly” labels or doubts about companies’ commitments.

Descriptive data support this, particularly on indicator GK4 (“I know how to choose eco-friendly products, such as looking for the ‘eco-friendly’ label when shopping”), which recorded the lowest mean of 4.04. This indicates that some respondents remain doubtful or lack understanding in recognizing truly trustworthy eco-friendly products.

The second hypothesis test (H2) shows that environmental concern has a positive and significant effect on green trust, with an original sample value of 0.356, t-statistic of 3.493, and p-value of 0.000. This means H2 is accepted: the higher respondents’ environmental concern, the greater their trust in eco-friendly products. This finding is consistent with Mamun et al. (2020) and Panopoulos et al. (2023), who emphasized that individuals with higher environmental concern are more likely to trust sustainable products because these align with their ecological values.

Support for this result is evident in the descriptive data for environmental concern, particularly indicator EC1 (“I feel compelled to protect the environment whenever I see a polluted river or litter in public places”), which obtained the highest mean of 4.63. This reflects respondents’ genuine empathy toward environmental issues and explains why their concern strengthens trust in products perceived as ecological solutions.

The results of the third hypothesis test (H3) indicate that green attitude has a positive and significant effect on green trust (original sample = 0.446; t-statistic = 4.234; p-value = 0.000), thus the hypothesis is accepted. A positive attitude toward eco-friendly products is proven to foster trust in the quality and sustainability commitment of these products. This finding is in line with Chen et al. (2022), who argued that pro-environmental attitudes—shaped by perceived benefits and moral responsibility—play a critical role in forming green trust.

Empirical support is evident in indicator GA1 (“Using a tumbler can help reduce plastic waste in the environment”), with a mean of 4.61. This demonstrates that the majority of respondents hold positive views on sustainability, thereby strengthening trust in green products.

The fourth hypothesis (H4) is accepted, as green knowledge was found to have a positive and significant effect on intention to use a tumbler (original sample = 0.138; t-statistic = 2.243; p-value = 0.025). This means that the greater one’s knowledge of environmental issues and eco-friendly products, the stronger their intention to use a tumbler. This finding is consistent with Liu et al. (2020) and Amoako et al. (2020), who confirmed that environmental knowledge fosters green consumption behavior because awareness of the dangers of single-use plastics drives consumers to switch to sustainable products.

Empirical evidence can be seen in indicators GK2 (mean = 4.41) and GK3 (mean = 4.43), suggesting that knowledge of eco-friendly products and sustainable practices generates awareness that ultimately increases the intention to use a tumbler.

The fifth hypothesis (H5) is accepted, as green trust has a positive and significant effect on intention to use a tumbler (original sample = 0.381; t-statistic = 4.029; p-value = 0.000). This means that the higher consumers’ trust in eco- friendly products, the stronger their intention to use them. This result is consistent with Gilal et al. (2020), who emphasized that trust in green products is a key factor in consumer adoption decisions.

Empirical support is reflected in indicator GT2 (mean = 4.46), showing that respondents believe tumbler producers have a genuine commitment to the environment. This trust encourages consumers’ confidence in adopting tumblers as a sustainable lifestyle choice.

The sixth hypothesis (H6) is rejected because environmental concern does not significantly affect intention to use a tumbler (original sample = 0.061; t- statistic = 0.872; p-value = 0.383). This implies that although respondents express high concern for the environment, it does not automatically translate into an intention to use a tumbler. This result differs from Nguyen et al. (2019), who identified environmental concern as a predictor of pro-environmental behavior. Other factors, such as trust, accessibility, and perceived costs, may play a more determining role in intention.

Descriptive results reinforce this phenomenon, with indicators EC2 (mean = 4.28) and EC4 (mean = 4.49) reflecting high environmental concern. However, this concern alone is not strong enough to directly influence intention without a mediating factor, such as green trust, to strengthen consumer confidence in the product.

The seventh hypothesis (H7) is accepted, as green attitude has a positive and significant effect on intention to use a tumbler (original sample = 0.409; t- statistic = 4.615; p-value = 0.000). This is consistent with the Theory of Planned Behavior (Ajzen, 1991), which asserts that attitude is a primary determinant of intention. Positive attitudes such as pride, responsibility, and belief in the benefits of green products are proven to encourage sustainable tumbler use.

Descriptive data further reinforce this finding, as shown by GA3 (mean = 4.52) and GA1 (mean = 4.61), which reflect respondents’ pride and confidence in the positive value of using eco-friendly products. Thus, positive attitudes serve as a strong driver of green consumption intention.

The eighth hypothesis (H8) is rejected because green trust does not significantly mediate the effect of green knowledge on intention to use a tumbler (O = 0.034; t = 0.955; p = 0.339). This means that while environmental knowledge can directly influence intention, the indirect path through trust is not strong. Trust in eco-friendly products is not automatically formed solely from individuals’ knowledge.

According to Chen and Chang (2012), green trust requires critical understanding as well as positive product experiences. In this study, indicator GK4 (“I know how to choose eco-friendly products, such as looking for the ‘eco- friendly’ label when shopping”) had the lowest mean (4.04), reflecting respondents’ limited ability to identify green attributes. This explains why knowledge alone is insufficient to build trust, resulting in the absence of a mediation effect.

The ninth hypothesis (H9) is accepted, as green trust significantly mediates the effect of environmental concern on intention to use a tumbler (O = 0.136; t = 2.611; p = 0.009). This means that environmental concern can increase intention indirectly through trust in green products. The higher the environmental concern, the greater the confidence in the credibility of eco-friendly products, which in turn strengthens the intention to use them.

This finding is consistent with Panopoulos et al. (2023), who emphasized that the effect of environmental concern on pro-environmental behavior is often mediated by trust. Descriptive data support this, as seen in indicator EC1 (“I feel compelled to protect the environment whenever I see a polluted river or litter in public places”), which achieved the highest mean of 4.63. This suggests that strong environmental concern fosters trust, which ultimately enhances the intention to use eco-friendly tumblers.

The tenth hypothesis (H10) is accepted, as green trust significantly mediates the effect of green attitude on intention to use a tumbler (O = 0.170; t = 2.989; p = 0.003). This means that a positive attitude toward green products builds trust, which subsequently strengthens intention. Within the framework of the Theory of Planned Behavior, attitude serves as an important antecedent that shapes trust before actual behavior occurs.

This result is supported by indicator GA1 (“I believe that using eco-friendly products is the right and valuable choice”) with a mean of 4.61, and GA3 (“I feel proud when using eco-friendly products in public”) with a mean of 4.52, both reflecting respondents’ positive attitudes. Consistent with Chen et al. (2022), a positive attitude facilitates consumer trust in green products, with trust acting as a psychological mechanism linking ecological values to behavioral intention.

The results of this study provide the basis for several practical policy recommendations that can be implemented by various stakeholders, including the government, businesses, educational institutions, and environmental communities.

The study found that green knowledge influences the intention to use a tumbler but does not significantly foster green trust. Therefore, more practical educational strategies are required, such as hands-on classes on identifying eco- friendly products, recognizing eco-labels, and distinguishing genuine green products from greenwashing.

Additionally, campaigns like “One Household, One Tumbler” involving schools and neighborhood associations can enhance environmental literacy while promoting household-level tumbler usage. Collaboration with social media and eco-friendly influencers can expand educational outreach, while retailers and tumbler producers can provide incentives (e.g., discounts or rewards) to motivate consistent sustainable behavior. These strategies are expected not only to strengthen green trust but also to reinforce usage intention, foster long-term behavioral change, and encourage society to adopt a more eco-friendly lifestyle.

The study indicates that environmental concern alone is insufficient to drive the intention to use a tumbler without trust. Hence, policy interventions are needed, such as providing refill stations in public spaces accessible only to tumbler users, to foster habits and facilitate eco-friendly practices. Governments can also offer incentives, such as price discounts or reward points, for individuals using personal tumblers when purchasing beverages.

Pro-environmental attitudes significantly influence both trust and intention to use a tumbler. To encourage this, companies or organizations could implement “green identity campaigns”, such as a “Tumbler Challenge” on social media or in schools, where participants showcase their daily tumbler use. Rewards could be non-material, such as certificates of recognition or acknowledgment in local green communities.

Trust in eco-friendly products serves as the key link between attitudes and intention to use. Producers should provide authenticity proof of their eco-friendly claims, such as official labels, production process videos, or customer testimonials. Governments can verify green products and monitor false claims. Practical initiatives could include monthly exhibitions of local green products involving producers, consumers, and certification bodies, to build trust through direct interaction.