Material Flow Analysis for Demand Forecasting and Lifetime-Based Inflow in Indonesia’s Plastic Bag Supply Chain

Abstract

Background: this research presents an integrated approach to enhancing the sustainability of plastic bag supply chains in Indonesia by addressing critical issues related to ineffective post-consumer waste management and low recycling rates. The objective of this study is to develop a combined framework of material flow analysis (MFA) and sustainable supply chain planning to improve demand forecasting and inflow management across the plastic bag lifecycle. Method: the research adopts a quantitative method using the XGBoost algorithm for forecasting and is supported by a polymer-based MFA framework that maps material flows from production to end-of-life stages. Result: the findings indicate that while production processes achieve high efficiency with a yield of 89%, more than 60% of plastic bag waste remains unmanaged after use. Moreover, scenario analysis demonstrates that single interventions are insufficient to achieve circularity targets, whereas integrated strategies (e.g., reducing export volumes, enhancing waste collection, and improving recycling performance) are more effective in increasing recycling rates beyond 35%. Additionally, the study reveals that increasing domestic recycling capacity and minimizing dependency on exports can significantly reduce environmental leakage and strengthen local waste management systems. Conclusions: the study’s novelty lies in demonstrating how machine learning and material flow data can be synergized to inform circular supply chain decisions and regulatory planning.

1. Introduction

Integrating material flow analysis (MFA) with supply chain management (SCM) provides a more effective way to build a sustainable and resilient supply chain, especially when it comes to demand forecasting and managing material inflows over a product’s lifetime. Traditional supply chain models often rely heavily on historical sales data and predictive analytics, which can lead to inefficiencies in resource planning, production scheduling, and waste management [1]. Incorporating MFA allows businesses to clearly track material movements across the supply chain, optimize resource use, reduce waste, and maintain ideal inventory levels [2]. Abu-Bakar and Charnley [3] believed that this approach supports better decision-making by allowing companies to fine-tune their procurement strategies, lower their dependence on raw materials, and integrate circular economy (CE) principles into their operations. Additionally, managing material inflows based on product lifespan, considering factors such as durability, reuse, and recovery, plays a key role in improving efficiency and reducing the environmental impact [4]. However, most SCM frameworks do not fully account for product life cycles when forecasting demand, leading to overproduction and supply chain disruptions [5]. Integrating MFA with SCM enables businesses to develop more precise forecasting models, align material inflows with real consumption patterns, reduce waste, and enhance the sustainability of manufacturing processes [6]. Moreover, emerging technologies like blockchain, artificial intelligence, and the Internet of things (IoT) create new possibilities for real-time tracking and analysis, improving supply chain transparency and resilience [7]. The motivation of this study lies in addressing inefficiencies in supply chains caused by the disconnection between material tracking and demand forecasting, especially in developing economies like Indonesia that face growing environmental and logistical challenges.

MFA has been widely used to evaluate resource efficiency, waste reduction, and CE practices, providing a systematic approach to tracking material usage across various industries [8]. Withanage and Habib [9] have extensively applied MFA in assessing material cycles, identifying inefficiencies in resource consumption, and proposing strategies for minimizing waste. However, most studies focus on macro-level applications, such as national resource flow assessments or sector-wide material accounting, rather than examining their implications at the operational level within individual supply chains [10]. Similarly, existing studies on lifetime-based inflow management emphasize product durability, reuse potential, and end-of-life recovery [11], but often lack real-time tracking mechanisms that align material availability with actual demand [11]. Without a precise approach to monitoring material inflows throughout a product’s lifecycle, industries struggle to optimize their resource planning, leading to inefficiencies in production and inventory control [12]. On the other hand, SCM has predominantly focused on forecasting models, network optimization, and demand planning, relying heavily on historical sales data and predictive analytics to improve decision-making [13]. Furthermore, advancements in digital technologies such as artificial intelligence, blockchain, and the Internet of things (IoT) have introduced new possibilities for enhancing SCM, yet their potential remains underexplored in the context of real-time material flow tracking [14]. Despite these developments, the existing literature rarely integrates MFA and SCM at a granular operational level for demand forecasting based on lifetime-based inflow, especially in developing country contexts. This lack of integration constitutes the main missing link addressed in this study.

Although prior research has significantly advanced the understanding of MFA and SCM as separate disciplines, several critical gaps remain, particularly in their combined application for demand forecasting and lifetime-based inflow management. One major limitation is the insufficient exploration of real-time data integration in MFA studies, which often rely on static datasets rather than dynamic tracking of material movements within supply chains [15]. Aljohani [16] affirmed that this lack of real-time insights limits the ability of businesses to make proactive decisions regarding procurement, production scheduling, and waste reduction. Additionally, while SCM has made strides in predictive analytics, existing models frequently overlook the role of material flow dynamics, resulting in inefficiencies in resource allocation and inaccurate demand projections [17]. The absence of a structured framework that connects material lifecycle data with supply chain decision-making further exacerbates these challenges, preventing businesses from fully optimizing their inflow management strategies [18]. Furthermore, sustainability remains an overlooked aspect in conventional SCM approaches, as many forecasting models prioritize cost and efficiency over long-term environmental impact, limiting opportunities for integrating CE principles such as material reuse and remanufacturing [19]. Therefore, this study is motivated by the need to close this gap by proposing an integrated framework combining MFA and SCM for more accurate, lifetime-based inflow demand forecasting within the supply chain.

The rest of the paper is structured into four sections. Section 2 reviews the relevant literature on MFA and SCM, highlighting existing gaps. Section 3 outlines the research methodology, detailing the data collection and analytical approach. Section 4 presents the findings and discusses their implications for demand forecasting and lifetime-based inflow management. Finally, Section 5 concludes with key insights, limitations, and directions for future research.

2. Theoretical Background

2.1. Material Flow Analysis (MFA) Applications in Sustainability

MFA is a systematic approach used to quantify the flow of materials and resources within a defined system, offering insights into resource consumption, waste generation, and environmental impact [20]. As a key tool in industrial ecology, MFA enables policymakers and businesses to assess material stocks, identify inefficiencies, and develop strategies for sustainable resource management [21]. It does this by tracking inputs, outputs, and accumulation within a system using tools like Sankey diagrams and mass balance equations [22]. MFA supports decision-making in manufacturing, urban planning, and waste management by enabling a data-driven approach to resource efficiency and sustainability [23]. Its ability to reveal consumption patterns and waste generation makes it essential for identifying resource optimization and supporting CE strategies [9]. Additionally, its application enhances supply chain resilience by improving demand forecasting and reducing dependency on virgin materials [23].

MFA has been widely applied in assessing resource efficiency, waste reduction, and CE initiatives. In manufacturing, it helps optimize raw material usage and minimize production waste. In waste management, MFA aids in evaluating material recovery and shaping recycling policies [24]. Its contribution to CE efforts is notable through secondary material flow tracking, which supports closed-loop systems [2]. However, MFA’s application in dynamic, real-time supply chain decision-making is limited. Traditional MFA studies often focus on static system boundaries, which makes it difficult to incorporate changing market conditions, demand fluctuations, and disruptions [25]. Moreover, integration with emerging digital tools such as IoT and AI remains underdeveloped [26], highlighting the need for more adaptive, tech-enabled approaches in sustainable supply chain planning.

2.2. Supply Chain Management Sustainability and Demand Forecasting

Effective supply chain management requires understanding product lifecycles to optimize material inflows and support sustainability. Lifetime-based inflow management focuses on durability and end-of-life processing, influencing procurement, inventory control, and waste reduction strategies [27]. Traditional inventory models, such as EOQ and JIT, emphasize short-term demand and cost minimization [removed: without fully accounting for product longevity] [28]. However, incorporating lifecycle considerations can enhance supply chain resilience by aligning inflows with usage patterns and extending product utility [19]. Moreover, CE principles emphasize closed-loop systems where materials are reused, remanufactured, and recycled, reducing dependence on virgin resources [29].

Several models have addressed inflow management with lifecycle perspectives, but gaps remain in real-world applications. Hybrid models that include reverse logistics and remanufacturing have reduced waste and optimized flows [30], yet they often assume static return rates and overlook real-world usage variability. Although predictive analytics have improved visibility, their use in lifetime-based inflow planning is still limited [31]. Many studies still emphasize traditional performance metrics, neglecting environmental and economic gains from lifecycle-based inflow strategies [32]. Addressing these issues requires combining MFA with advanced forecasting to create supply chains that are both efficient and sustainable.

2.3. Lifetime-Based Inflow Management in Supply Chain Sustainability

Effective SCM requires a comprehensive understanding of product lifecycles to optimize material inflows and ensure sustainability. Lifetime-based inflow management focuses on the durability, reuse, and end-of-life processing of materials, influencing procurement, inventory control, and waste reduction strategies [27]. Traditional inventory management models, such as economic order quantity (EOQ) and just-in-time (JIT), primarily emphasize short-term demand fluctuations and cost minimization without fully accounting for product longevity [28]. However, Hazen and Russo [19] believed that incorporating lifecycle considerations can enhance supply chain resilience by aligning inflows with actual usage patterns and extending product utility. Moreover, CE principles further emphasize closed-loop systems where materials are continuously cycled through reuse, remanufacturing, and recycling, reducing dependence on virgin resources [29].

Several models have attempted to address inflow management with lifecycle perspectives, but significant gaps remain in their implementation. For instance, hybrid inventory models incorporating reverse logistics and remanufacturing have shown promise in reducing waste and optimizing material flows [30]. However, these models often rely on static assumptions about product returns and fail to capture the uncertainties associated with real-world usage patterns. Additionally, while predictive analytics and AI-driven forecasting methods have improved supply chain visibility [31], their application in lifetime-based inflow management is still underdeveloped. Many existing studies focus on traditional supply chain efficiency metrics, overlooking the environmental and economic benefits of integrating lifecycle considerations into inflow planning [32]. Addressing these gaps requires a more holistic approach that combines MFA with advanced forecasting techniques to optimize supply chains for long-term sustainability and resource efficiency.

2.4. Integration of MFA with SC Sustainability: Existing Studies and Limitations

Several studies have explored the integration of MFA with supply chain (SC) sustainability, emphasizing its role in optimizing resource efficiency and reducing environmental impact. Millette and Williams [33] applied MFA to track material consumption, improve CE practices, and support waste reduction in supply chains. Dzhuguryan and Deja [34] showed that MFA can enhance sustainable SCM by identifying inefficiencies in material usage and promoting closed-loop systems. Moreover, Kužmarskytė [2] demonstrated how MFA-based assessments could guide sustainable procurement and recycling strategies by mapping material flows across different life cycle stages. Additionally, integrating MFA into decision-making frameworks has been linked to improved sustainability reporting and compliance with environmental regulations [35]. However, many existing studies focus on macro-level sustainability assessments rather than micro-level supply chain operations, limiting the practical application of MFA in real-time SC management. For example, Chowdhury and Quaddus [36] highlighted the potential of MFA to enhance supply chain resilience by identifying critical material dependencies and mitigating resource shortages.

Technological advancements such as artificial intelligence (AI), the Internet of things (IoT), and blockchain have been increasingly explored to enhance MFA-SC integration, improving data accuracy, transparency, and predictive capabilities. AI-driven models have shown promise in forecasting material demand and optimizing resource allocation [37], while IoT-enabled tracking systems provide real-time visibility into material movements [7]. Blockchain technology further enhances traceability and accountability, ensuring more reliable material flow data [38]. Moreover, the integration of digital twins with MFA enables dynamic simulation of material flows, allowing for proactive adjustments in supply chain operations [39]. These technologies facilitate the implementation of closed-loop supply chains by improving the monitoring and validation of recycled materials. Additionally, Mageto [40] found that combining big data analytics with MFA has been shown to enhance decision-making by identifying patterns in material consumption and waste generation, leading to more sustainable procurement strategies.

2.5. Research Gap and Contribution

While research on MFA and SCM has expanded significantly, their integration for operational-level decision-making, particularly in demand forecasting and lifetime-based inflow management, remains limited. MFA is widely used for analyzing resource efficiency, reducing waste, and supporting CE practices [41]. However, its application in real-time supply chain operations is still lacking. Most existing studies emphasize large-scale material flow assessments, with limited focus on integrating MFA into procurement, production, and inventory planning [42]. Conversely, SCM research often focuses on forecasting and optimization [43], relying heavily on historical sales data without considering real-time material flows [44]. Moreover, conventional SCM models rarely incorporate lifetime-based inflow strategies that account for product durability, reuse, and end-of-life value, resulting in inefficient material planning. This study addresses these limitations by proposing an integrated MFA–SCM framework to enhance demand forecasting and inflow management. By embedding MFA data into supply chain planning, the model improves resource allocation, reduces waste, and strengthens resilience. This research also explores how emerging technologies, such as AI, blockchain, and IoT, can support real-time material tracking within SCM systems [45]. Unlike prior studies treating MFA and SCM as separate fields, this approach bridges them to promote closed-loop systems and optimized secondary inflows, advancing both theoretical knowledge and practical application in sustainable supply chains. Such integration enhances efficiency in dynamic markets, contributing to resource-conscious and adaptive supply chain strategies [46].

3. Research Method

This study adopts an integrated methodological approach that combines MFA with SCM principles to address the challenges of demand forecasting and lifetime-based inflow management in sustainable supply chains. The methodology is designed to bridge gaps identified in previous research by leveraging real-time data tracking, lifecycle assessment, and digital technologies to enhance decision-making processes. It involves mapping material flows across supply chain stages, analyzing inflow patterns in relation to product lifespans, and applying predictive modeling techniques to align resource planning with actual consumption trends. Furthermore, the study incorporates both qualitative and quantitative data sources, supported by digital tools such as machine learning, system dynamics modeling, and Jupyter Notebook ver. 7.2.2, to improve visibility, accuracy, and adaptability in forecasting and material inflow strategies. The following subsections detail the data collection methods, system boundaries, analytical procedures, and tools used to operationalize the integrated MFA-SCM framework outlined in Figure 1.

3.1. Data Collection Methods

The study employs a mixed-method data collection strategy, combining secondary data from environmental agencies, municipal waste records, and retail sales data with primary interviews from supply chain managers and packaging manufacturers. This approach ensures a comprehensive understanding of the material flow of shopping plastic bags from production to end-of-life stages. Historical demand data were collected from leading supermarket chains, while plastic bag composition and degradation rates were sourced from technical datasheets and environmental impact assessments. These datasets support the construction of a reliable MFA and the integration with supply chain models for demand forecasting.

Moreover, stakeholders from various sectors, including retailers, waste management services, and recycling facilities, were interviewed to validate data accuracy and supply chain assumptions. These qualitative inputs helped clarify factors such as consumer reuse behavior, collection rates, and the effectiveness of plastic bag recycling schemes. As emphasized by Cerchione and Morelli [47], capturing both quantitative flows and qualitative insights allows for a more holistic and accurate representation of the material lifecycle. This dual-method approach enhances the study’s robustness, ensuring that the forecasted inflows reflect real consumption patterns and product lifespans.

3.2. System Boundaries

The system boundary for this analysis covers the extraction of raw materials for the production of HDPE (high-density polyethylene) and LDPE (low-density polyethylene) plastic bags for final disposal or recycling. This includes upstream processes (resin production and bag manufacturing), midstream activities (retail distribution and consumer use), and downstream flows (collection, sorting, and disposal). Following the guidelines outlined by Stafford and Russo [48], this cradle-to-grave perspective allows for a complete material balance across each stage of the plastic bag’s lifecycle. This boundary setting ensures the inclusion of both technical flows, like production waste and recycled content, and socio-behavioral flows, such as reuse frequency and improper disposal. In defining these system boundaries, particular emphasis was placed on capturing inflow dynamics over time, segmented by product lifetime categories (e.g., single-use vs. multi-use plastic bags). The approach aligns with [49], who argue for lifetime-based inflow management to improve forecasting and resource efficiency. Additionally, the system explicitly includes leakages to the environment and uncollected waste to assess the sustainability impact more accurately. This level of detail is essential for evaluating CE strategies, including reuse loops and secondary material integration [50].

This study defines the system boundaries in spatial, temporal, and functional terms. Temporally, the system is analyzed over a three-year period (2022–2024). The selection of this timeframe aims to capture contemporary dynamics related to the demand for plastic shopping bags within a short-term horizon, adequately representing changes in usage trends and material management practices within the retail distribution system. Although the time span is relatively short, the analysis aligns with the ISO 14044 [51] guidelines, which allow for study period adjustments based on data availability and the specific research focus on material dynamics during the active use phase.

Functionally, the system focuses on three main components: (1) inflow, referring to the number of new products entering the market each year; (2) in-use stock, which represents the accumulation of products still being used by consumers; (3) end-of-life inflow, denoting the number of products exiting the system after reaching the end of their useful life and entering waste management or recycling processes. Functions beyond this scope, such as upstream raw material production and downstream waste processing outside the study area, are excluded from the system boundaries as they are not directly relevant to the objective of assessing local sustainability performance.

As an integral part of the system, the demand forecasting component plays a critical role in estimating the annual inflow of new products into the system. The forecasting approach incorporates several machine learning models, including hybrid models based on ARIMAX and ensemble methods [52,53]. The application of machine learning in demand forecasting within a sustainability context has been shown to effectively capture non-linear patterns and market uncertainties with greater accuracy than conventional statistical methods [52,53,54]. These demand forecasts serve as a key input for the product lifetime-based inflow model, which estimates the number of products likely to exit the system in the future. Hence, the system boundaries are not merely descriptive but also form the foundation for dynamic quantitative modeling within the MFA–SCM framework.

3.3. Machine Learning for Demand Prediction

In this research, plastic shopping bag demand forecasting is modeled using the extreme gradient boosting (XGBoost) algorithm, a machine learning-based approach designed to handle complex, non-linear relationships in data. Unlike conventional statistical models such as exponential smoothing or linear regression, XGBoost leverages an ensemble of decision trees combined through gradient boosting to improve prediction accuracy and generalization performance. This method is particularly well-suited for capturing sudden fluctuations and non-linear demand patterns influenced by external factors such as policy changes, population shifts, and inflation. By integrating multiple predictors, including temporal variables (e.g., month and seasonality) and relevant exogenous drivers, XGBoost constructs a robust forecasting framework capable of modeling the dynamic nature of plastic bag consumption [52,53].

3.4. Demand Forecasting Model

Model performance evaluation is a crucial stage in ensuring the accuracy, validity, and reliability of forecasting models developed in the integration of MFA and SCM. This evaluation aims to assess the extent to which the model is able to predict demand and material inflows precisely and consistently compared to historical actual data. In this context, various statistical indicators are used to measure prediction accuracy, such as the absolute error rate and the coefficient of determination to explain variations in actual data. As explained by Badulescu and Hameri [55], systematic evaluation enables the identification of model strengths and weaknesses in capturing demand dynamics, thereby reducing the risk of uncertainty in strategic decision-making. Sattar and Dattana [56] also emphasized that in the context of machine learning-based supply chains, model performance is measured through prediction accuracy, stability, and generalizability when tested on new or unexpected data.

In this study, the forecasting model integrates specific time-dependent variables such as seasonal indices (e.g., monthly or quarterly consumption patterns) and exogenous factors, including policy dummy variables (e.g., the introduction of plastic bag bans or subsidies) and macroeconomic indicators (e.g., inflation and consumer spending trends). These additions enable the model to better capture real-world fluctuations and anticipate changes due to external influences. This approach improves demand visibility and responsiveness, particularly in volatile supply environments. Furthermore, the model evaluation considers not only statistical metrics but also the sustainability and resilience dimensions of the supply chain. Koppiahraj and Bathrinath [57] suggest that performance assessment should include resource efficiency and waste reduction based on CE principles, such as closing, slowing, and narrowing material loops. This is supported by Sattar and Dattana [56], who showed that integrating machine learning with real-time analytics enhances logistics efficiency by improving inventory control and demand forecasting accuracy. Therefore, the evaluation framework adopted here is both technically robust and practically grounded, ensuring the MFA-SCM model supports efficient, adaptable, and sustainability-oriented decision-making.

3.5. MFA as a Modeling and Simulation Tool

In this study, MFA is employed as a systematic approach to identify, quantify, and analyze the flow and stock of plastic shopping bags throughout their life cycle. MFA functions as an input-output analysis tool and as a dynamic modeling framework based on the product’s lifetime distribution. This approach focuses on understanding the interactions among three key components in the life cycle system of plastic shopping bags: inflow (the number of new products entering the market), in-use stock (products currently being used by consumers), and end-of-life flow (products reaching the end of their useful life and entering waste management systems). The modeling process is based on the principle of mass balance, in which every material entering the system must correspond to the outflow and the changes in stock over a defined period, as emphasized by Zaghdaoui and Jaegler [46] and reinforced by mass-balance-based MFA practices in supply chain resilience studies [21,58].

To support both quantitative and qualitative interpretations of material flow dynamics, a Sankey Diagram is utilized as the main visualization tool. This diagram represents the distribution of material flows across different stages of the system, with arrow thickness proportional to the magnitude of the flow, as discussed by Syafrudin and Budihardjo [20]. Pernici and Stancu [59] indicated that the use of Sankey diagrams provides several strategic advantages in the context of circular supply chains: first, it helps identify dominant consumption points and critical nodes of waste generation; second, it enables the mapping of systemic relationships among stages such as production, distribution, usage, reuse, recycling, and disposal; third, it offers a visual basis for analyzing dynamic changes, including comparisons of different policy scenarios, such as plastic bag bans or enhanced recycling programs.

Furthermore, this study develops a future stock prediction model based on the lifetime distribution of plastic shopping bags, integrating historical data and return patterns through a life-cycle-based forecasting approach. This model facilitates scenario-based simulations for improved SCM, such as reducing single-use plastic bag consumption or increasing recycling rates. Therefore, MFA in this study acts as a static accounting tool and as a dynamic scenario-based modeling instrument to support strategic decision-making in securing a more sustainable and resilient supply chain for plastic shopping bags.

• Top-Down Polymer-Based MFA Approach
  The study applies a top-down polymer-based MFA framework to map and quantify the flow of plastic shopping bags throughout the supply chain. This approach relies on aggregate sectoral data, which is then disaggregated into specific polymer streams, such as high-density polyethylene (HDPE) and low-density polyethylene (LDPE), using market composition ratios and product-specific material data. By leveraging polymer-level analysis, the study can identify material-specific inefficiencies, recycling bottlenecks, and environmental leakages. The top-down perspective allows the research to capture the entire system scope, from upstream resin production to midstream manufacturing and retail distribution, and ultimately to downstream waste management, recycling, incineration, or disposal. This polymer-focused approach strengthens the analysis by aligning it with international material flow reporting frameworks and ensuring the recommendations are relevant not only to operational practices but also to material efficiency and CE goals.
• MFA Scenario Analysis
  To strengthen the system analysis, the study develops MFA-specific scenarios that simulate the impact of material-system interventions on the overall mass balance, circularity, and sustainability performance. These scenarios complement the operational SCM scenarios by focusing directly on the physical material dynamics and polymer flows.

The key MFA scenarios evaluated include the following:

• The polymer substitution scenario, where a defined percentage (e.g., 20%) of virgin HDPE or LDPE inputs is replaced with recycled or bio-based alternatives, assessing the potential reduction in upstream resource extraction and dependency;
• The closed-loop recycling enhancement scenario, modeling an increase in domestic recycling yields (e.g., from 10% to 25%), thereby reducing the demand for virgin resin inputs and lowering end-of-life waste generation;
• The collection and sorting improvement scenario, simulating the impact of higher mechanical sorting and collection efficiencies, especially for thin-film plastics, which are typically under-captured in conventional waste streams;
• The combined intervention scenario, integrating multiple levers (substitution, recycling enhancement, and improved collection) under varying production growth or reduction assumptions, to project a best-case pathway toward circularity goals.

Each scenario is quantitatively evaluated by adjusting the relevant input parameters in the MFA system, recalculating material balances, and visualizing the impacts through updated Sankey diagrams and performance metrics such as recycling rates, landfill diversion, mismanaged waste reduction, and overall system yield. This scenario-driven MFA framework provides empirical, actionable insights for policymakers, industry actors, and waste management stakeholders seeking to transition toward more circular and resource-efficient plastic value chains.

3.6. Impact Analysis and Recommendations

This stage aims to evaluate the outcomes of integrating MFA with SCM in the context of sustainable supply chain operations for plastic shopping bag products. The impact analysis is conducted through several approaches:

• Environmental Impact Evaluation
  Using MFA outputs, the volume of plastic entering, being used, and ending up as waste is calculated. This analysis identifies the contribution of plastic shopping bags to waste accumulation and the potential for impact reduction through enhanced reuse and recycling practices. The evaluation is based on principles of industrial ecology and the CE framework [60].
• Operation Supply Chain Impact Evaluation
  In this study, supply chain performance is evaluated across three critical dimensions: efficiency, stability, and adaptability. Efficiency is measured by assessing the material flow through the input–output ratio, which reflects how effectively resources are converted into final products [61]. Stability pertains to the consistency of in-use stock levels and the associated risks of product scarcity, highlighting the importance of maintaining adequate inventory to ensure service continuity [62]. Lastly, adaptability refers to the supply chain’s responsiveness to demand fluctuations, particularly in alignment with predictions generated by the forecasting model, which is crucial for mitigating the impact of market volatility [63]. These dimensions collectively provide a comprehensive framework for analyzing supply chain performance under dynamic conditions.
• Supply Chain Resilience Evaluation
  Supply chain resilience in this study is evaluated through sensitivity analysis using simulation scenarios that incorporate demand fluctuations and plastic-related policy changes. This approach assesses the system’s capacity to sustain operations under shifting market conditions, providing insights into its adaptive robustness [64].

The recommendations developed from this analysis are intended to enhance sustainability, operational efficiency, and supply chain resilience within the company.

To ensure a comprehensive understanding of the findings, this section transitions from the conceptual and simulation-based evaluation of environmental and operational impacts to a more technical assessment of model performance. The integration of forecasting models and MFA requires validation through appropriate metrics to ensure accuracy, reliability, and practical applicability.

Table 1. Mathematical models and performance metrics for forecasting and material flow evaluation.

Name	Mathematical Formula	Description	Source
Model XGBoost	$\hat{\mathrm{Y}}={\displaystyle \sum _{K=1}^K}{f}_k({\mathrm{x}}_{\mathrm{i}})$	Ensemble regression model using decision trees with gradient boosting; supports temporal and non-linear patterns; robust to multicollinearity and missing values	[53]
Mean Squared Error (MSE)	$\mathrm{MSE}={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}{({\mathrm{y}}_{\mathrm{i}}-{\hat{\mathrm{y}}}_{\mathrm{i}})}^2$	Mean of squared differences between actual and predicted values; lower MSE indicates better performance	[65,66]
Mean Absolute Error (MAE)	$\mathrm{M}\mathrm{A}\mathrm{E}={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}|{\mathrm{y}}_{\mathrm{i}}-{\hat{\mathrm{y}}}_{\mathrm{i}}|$	MAE measures the average absolute difference between actual and predicted values; lower values indicate higher model accuracy	[65,66]
Root Mean Squared Error (RMSE)	$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{1}{n}}}{\displaystyle \sum _{i=1}^n}{({\mathrm{y}}_{\mathrm{i}}-{\hat{\mathrm{y}}}_{\mathrm{i}})}^2$	The square root of MSE; more sensitive to large errors or outliers than MAE	[65,66]
Koefisien Determinasi (R2)	${\mathrm{R}}^2=1-{\displaystyle \frac{SS Eror}{SS Total}}=1-{\displaystyle \frac{{\sum }_{i=1}^n{({\mathrm{y}}_{\mathrm{i}}-{\hat{\mathrm{y}}}_{\mathrm{i}})}^2}{{\sum }_{i=1}^n{({\mathrm{y}}_{\mathrm{i}}-\overline{\mathrm{y}})}^2}}$	Indicates the proportion of variation in the dependent data explained by the model, with values close to 1 indicating better prediction	[67]
Mass Balance Principle	$\sum Input=\sum Output+∆Stoct$	Used in MFA to ensure that the entire material flow is calculated consistently	[68]

presents the mathematical formulas and performance indicators used in this study, providing a foundation for evaluating predictive accuracy and supporting the formulation of actionable recommendations.

• where
• 5. Model XGBoost:
• $\hat{\mathrm{Y}}$ is the predicted value for observation i;
• $K$ is the total number of trees;
• $f_k$ is regression tree $K$, a member of the function space $f$;
• ${\mathrm{x}}_{\mathrm{i}}$ is the input feature vector.
• 6. Mean Squared Error (MSE):
• ${\hat{\mathrm{y}}}_{\mathrm{i}}$ is the predicted value for the i-th observation;
• ${\mathrm{y}}_{\mathrm{i}}$ is the actual (true) value for the i-th observation;
• $n$ is the total number of observations.
• 7. Mean Absolute Error (MAE):
• 8. ${\hat{\mathrm{y}}}_{\mathrm{i}}$ is the predicted value for the i-th observation;
• 9. ${\mathrm{y}}_{\mathrm{i}}$ is the actual (true) value for the i-th observation;
• 10. $n$ is the total number of observations.
• 11. Root Mean Squared Error (RMSE):
• RMSE is derived from the MSE formula by taking its square root;
• Useful for interpreting errors in the same units as the original data.
• 12. R-squared (R²):
• ${\hat{\mathrm{y}}}_{\mathrm{i}}$ is the predicted value;
• ${\mathrm{y}}_{\mathrm{i}}$ is the actual value;
• $\overline{\mathrm{y}}$ is the mean of the actual values;
• Measures the proportion of variance explained by the model.
• 13. Mass Balance Principle:
• Input refers to all material flows entering the system (e.g., raw materials and intermediate goods).
• Output includes all material flows exiting the system (e.g., products, emissions, and waste).
• $∆Stoct$ or accumulation represents the net change in material stock within the system over a defined period. It can be positive (stock increase) or negative (stock depletion).

4. Results and Discussion

To initiate the results and discussion section, a comprehensive dataset was constructed to support material flow forecasting for the year 2025. This dataset, based on monthly records from January 2023 to December 2024, comprises 24 observations and includes key variables such as resin imports, domestic resin production, export volumes, resin consumption, plastic bag production and distribution, post-consumer waste outcomes, and end-of-life treatment. The integration of these variables enables a detailed analysis of material flow dynamics within the plastic bag supply chain. This data-driven approach serves as a foundational element for identifying trends, evaluating system performance, and informing strategic decisions aimed at improving resource efficiency and sustainability [69].

Table 2. Monthly average values of historical material flows (2023–2024).

Flow Component	Value (tons)
Raw Material Import	24.74
Raw Material Production	11.46
Raw Material Export	2.59
Resin Used Production tons	36.20
Production Losses tons	1.10
Plastic Bag Production	32.51
Distribution	32.51
Consumption	32.51
Waste	14.67
Recycling	4.09
Incinerated	32.51

presents the monthly average values of historical material flows for the years 2023 to 2024, offering a detailed overview of key stages in the plastic bag lifecycle. The data indicates that an average of 24.74 tons of raw material was imported monthly, supplemented by 11.46 tons of domestic production, while 2.59 tons were exported. This results in an overall resin input of approximately 36.20 tons used in production, with production losses accounting for 1.10 tons per month. Consequently, the average monthly output for plastic bag production, distribution, and consumption was consistent at 32.51 tons, suggesting a streamlined flow from manufacturing to end use. However, the waste generated averaged 14.67 tons monthly, of which only 4.09 tons were recycled, while the majority (32.51 tons) was incinerated. These figures highlight inefficiencies in post-consumer waste management, particularly the limited recycling rate, underscoring the need for improved recovery strategies to enhance circularity within the plastic supply chain.

4.1. Forecasting of Plastic Bag Material Flows in 2025

A comprehensive MFA of plastic bag systems for 2025 was enabled by forecasting a complete set of annual flow data using a machine learning-based approach. Leveraging historical monthly data from 2023 to 2024, XGBoost regression models were successfully trained for each material flow variable. The modeling process incorporated temporal features, including month encoding and average product lifespan, to account for seasonal fluctuations and post-consumer behavior. A key assumption in the model was that plastic bags have an average useful life of three months, consistent with empirical findings on single-use plastic packaging [70]. This assumption was operationalized by introducing a time lag to the waste generation variable, whereby projected waste quantities in 2025 were derived from consumption data three months prior. This approach effectively simulates the delayed transition from usage to disposal, enhancing the temporal accuracy of the material flow projections. The forecasted annual material flows across the plastic bag value chain are presented in

Table 3. Forecasted annual material flows for plastic bags in 2025.

Flow Component	Value (tons)
Raw Material Import	299.65
Raw Material Production	137.26
Raw Material Export	33.36
Resin Used Production (tons)	438.63
Production Losses (tons)	13.16
Plastic Bag Production	390.39
Distribution	390.39
Consumption	390.39
Waste	390.39
Recycling	39.10
Incinerated	38.37
Landfill	77.69
Loss	235.23

.

Table 3 presents the forecasted annual material flows for plastic shopping bags in 2025, derived through a combination of demand forecasting using the extreme gradient boosting (XGBoost) algorithm and a top-down mass balance approach based on material flow analysis (MFA). The demand forecast model integrated historical consumption data, seasonal indices, and exogenous drivers such as population growth, inflation, and regulatory interventions to project total plastic bag usage. XGBoost was selected due to its capability to model complex non-linear relationships and incorporate multiple predictive features, enabling the generation of reliable consumption estimates [52]. The predicted consumption for 2025 was then used as the core anchor point for the material flow calculations.

From the estimated consumption value (390.39 tons), the upstream and downstream flows were calculated using conversion ratios and loss factors derived from industry benchmarks, technical reports, and national plastic manufacturing statistics. For instance, raw material inputs, both imported (299.65 tons) and domestically produced (137.26 tons), were determined based on resin-to-product yield rates, adjusted for typical production losses of around 3% (13.16 tons) and raw material exports (33.36 tons), resulting in a total resin input of 438.63 tons. This input was reconciled against outputs to ensure a closed mass balance loop, as recommended in MFA frameworks [68].

Post-consumption waste flows were mapped assuming a short average use phase for shopping bags, leading to full disposal within the forecast year. Waste management allocations (recycling (39.10 tons), incineration (38.37 tons), landfill (77.69 tons), and uncollected losses (235.23 tons)) were estimated using national waste treatment ratios and expert-validated assumptions about collection efficiency, recovery rates, and leakages in the system [71,72]. The consistency across input and output values reinforces the integrity of the flow model, while the material losses emphasize persistent system inefficiencies. These findings offer actionable insights for policymakers aiming to design more effective interventions for circularity and plastic waste reduction.

4.2. Model Performance and Accuracy Evaluation

The reliability and robustness of the forecasting process were assessed by evaluating the machine learning model’s performance across all key material flow variables. This evaluation utilized two widely accepted statistical metrics: mean squared error (MSE) and mean absolute percentage error (MAPE). MSE captures the average of the squared differences between predicted and actual values, emphasizing larger errors and offering insight into the overall predictive accuracy of the model. In contrast, MAPE provides a relative measure by expressing prediction errors as a percentage, making it especially useful when comparing variables measured on different scales [73]. Both metrics are frequently employed in time-series forecasting and machine-learning research to assess model accuracy and reliability [74]. Their application in this study supports the validity of the XGBoost regression model in predicting material flow patterns within the plastic bag system, enhancing confidence in the 2025 forecast results.

4.3. Evaluation Metrics

Model performance was quantitatively assessed using two widely adopted evaluation metrics: mean squared error (MSE) and mean absolute percentage error (MAPE). MSE calculates the average of the squared differences between predicted and actual values, placing greater weight on larger errors and thus making it particularly effective in identifying significant deviations and outliers within the prediction results [65]. This sensitivity helps in evaluating the overall precision of the model, especially when accurate forecasting of extreme values is critical. In contrast, MAPE measures the average absolute percentage deviation between predicted and observed values, offering a scale-independent and intuitive interpretation of relative prediction accuracy. Due to its percentage-based nature, MAPE is especially useful for comparing model performance across variables with different units or magnitudes [73]. In this study, both MSE and MAPE were calculated for each forecasted material flow variable to assess the predictive accuracy of the model and to identify potential areas of uncertainty, thereby ensuring a robust evaluation of the model’s forecasting capability.

4.4. Model Performance Across Material Flow Components

A detailed evaluation of model performance across individual material flow components highlights the overall robustness and accuracy of the forecasting framework. By applying the mean squared error (MSE) and mean absolute percentage error (MAPE) to each predicted variable, the analysis enables a nuanced understanding of the model’s predictive capabilities and its ability to capture the dynamics of plastic bag material flows.

Table 4. Model performance metrics for each forecasted variable.

Flow Component	MSE	MAPE
Raw Material Import	0.05	0.72%
Raw Material Production	0.02	0.67%
Raw Material Export	0.02	3.37%
Resin Used Production (tons)	0.06	0.58%
Production Losses (tons)	0	0.75%
Plastic Bag Production	0.12	0.84%
Distribution	0.12	0.84%
Consumption	0.12	0.84%
Waste	0.12	0.84%
Recycling	0.35	15.33%
Incinerated	0.25	11.81%
Landfill_	0.1	3.76%
Loss_	0.07	1.20%

presents the error metrics for each forecasted component, illustrating that the model performed particularly well for core variables such as raw material import (MSE = 0.05 and MAPE = 0.72%), resin used in production (MSE = 0.06 and MAPE = 0.58%), and plastic bag production, distribution, and consumption (each with MSE = 0.12 and MAPE = 0.84%). These low error values reflect a high degree of precision and minimal deviation between predicted and actual trends, reinforcing the suitability of the XGBoost model for time-series forecasting in MFA [74]. However, slightly higher errors were observed in variables related to end-of-life processes, such as recycling (MAPE = 15.33%) and incineration (MAPE = 11.81%), suggesting greater uncertainty in forecasting post-consumer behavior and treatment outcomes. These findings underscore the importance of incorporating more granular data or contextual factors in future modeling efforts to improve prediction accuracy in waste management components.

The performance evaluation presented in Table 4 reveals a distinct disparity in predictive accuracy between upstream and downstream material flow components. For upstream flows, including raw material imports (MAPE = 0.72%), domestic resin production (MAPE = 0.67%), resin usage in production (MAPE = 0.58%), and plastic bag production (MAPE = 0.84%), the XGBoost model demonstrated high predictive accuracy, as indicated by consistently low MAPE values below 1%. This strong performance is likely due to the structured and regulated nature of supply chain operations, which tend to follow established production schedules and are less susceptible to fluctuations driven by consumer behavior [75]. In contrast, downstream variables such as recycling (MAPE = 15.33%), incineration (MAPE = 11.81%), and landfill (MAPE = 3.76%) exhibited higher error margins, reflecting greater uncertainty in forecasting end-of-life outcomes. These downstream processes are heavily influenced by variable consumer disposal patterns, collection inefficiencies, and policy enforcement, making them inherently more difficult to model with high precision [71]. Thus, while the model is effective in capturing the upstream dynamics of the plastic bag life cycle, its performance diminishes slightly in the more volatile post-consumer phase, highlighting areas for future model refinement.

Intermediate flow components, such as raw material exports (MAPE = 3.37%) and waste stockpiling (MAPE = 3.76%), were forecasted with reasonably strong accuracy, indicating that the model effectively captured their month-to-month variability. While these flows are subject to fluctuations due to market demand or logistical factors in waste handling, their relative temporal consistency enables the model to generalize well from historical patterns. In contrast, downstream flows associated with waste treatment presented more challenges for accurate prediction. Notably, recycling and incineration recorded higher MAPE values of 15.33% and 11.81%, respectively. These elevated error rates can be attributed to several factors, including inconsistent consumer disposal behaviors, variable collection efficiency, seasonal effects, and the often-underreported activities of the informal recycling sector [73]. Despite this variability, the model demonstrated robust accuracy in key production-related variables critical to mass balance calculations, such as resin input and finished plastic output, indicating the reliability of the dataset for MFA applications. This finding supports previous research, which acknowledges that predictive models tend to exhibit reduced performance in post-consumer stages due to higher uncertainty and data heterogeneity [40]. Overall, the model’s balanced performance across upstream, intermediate, and downstream flows affirms its value as a forecasting tool within MFA frameworks, providing a solid foundation for evaluating material circularity, quantifying losses, and simulating the impact of policy or system-level interventions.

4.5. Material Flow Analysis (MFA) of Shopping Plastic Bags

This section presents a comprehensive top-down MFA for shopping plastic bags in the year 2025, utilizing forecasted flow data derived from an XGBoost-based machine learning model. MFA is a systematic approach widely used to quantify the flow and stock of materials within a defined system, enabling researchers and policymakers to evaluate resource efficiency, environmental impacts, and opportunities for circularity [68]. The model in this study integrates historical material flow data with advanced forecasting techniques to construct a detailed representation of the plastic bag life cycle.

The system boundary covers the entire value chain, from the importation and domestic production of raw materials (primarily polyethylene resins), through manufacturing, distribution, and consumption, to end-of-life (EOL) stages, including recycling, incineration, landfill, and losses to the environment. This cradle-to-grave perspective is essential for capturing the interdependencies across the supply chain and accurately identifying material losses and recovery potentials [76,77]. Figure 2 illustrates the Sankey diagram representing the forecasted material flows for plastic bags in 2025, offering a visual depiction of input-output relationships and the proportional magnitude of flows across each stage. This analytical framework lays the groundwork for subsequent discussions on material circularity, system inefficiencies, and policy implications.

The Sankey diagram presented in Figure 2 offers a comprehensive visualization of the material flow system for plastic bags in 2025, revealing critical inefficiencies and imbalances along the value chain. This diagram breaks down material inputs, outputs, and losses across the plastic bag life cycle, from production and consumption to disposal, making it possible to identify pressure points and intervention opportunities. One of the most striking insights from the diagram is the disproportionate volume of mismanaged waste, which accounts for approximately 235.23 tons, equivalent to 60% of total post-consumer plastic bag waste. This figure significantly overshadows the comparatively small quantities of waste that are either recycled (39.10 tons) or incinerated (38.37 tons), emphasizing the limited effectiveness of existing end-of-life management strategies.

These findings underscore the inadequate material recovery infrastructure, particularly in developing countries such as Indonesia, where systemic inefficiencies lead to high rates of environmental leakage. Compared to studies by Jerie et al. [78], who also identified high levels of mismanaged plastics in Zimbabwe’s informal settlements, our results show an even greater proportion of leakage in the Indonesian context. Similarly, ref. [79] reported only 25–35% of post-consumer plastics being properly managed in other Southeast Asian countries, suggesting that the 60% figure in our analysis reflects an especially urgent case.

The visual representation underscores a pressing issue in plastic waste governance: the inability of current infrastructure and behavioral practices to divert substantial amounts of used plastic bags away from environmentally detrimental pathways such as open dumping, leakage, or informal burning [80]. These findings align with broader global concerns, where single-use plastics frequently escape formal waste collection systems due to inadequate recycling infrastructure, limited public awareness, and weak policy enforcement [79]. The dominance of mismanaged waste in the system represents a lost opportunity for resource recovery and poses serious environmental threats, including land and marine pollution, greenhouse gas emissions, and public health risks. The comparative analysis with earlier studies strengthens the case for urgent, context-specific interventions. The Sankey diagram, therefore, does more than illustrate flows, it tells a critical story of systemic failure that requires urgent multi-stakeholder intervention through improved waste segregation, recycling incentives, and the implementation of CE principles to reduce leakage and enhance sustainability outcomes.

4.6. Overview of the Plastic Bag System

This section presents an integrated overview of the plastic bag system for the year 2025, employing a top-down MFA approach to trace and quantify the life cycle of plastic bags from raw material input to post-consumer waste outcomes. The analysis is grounded in forecasted data generated through machine learning techniques, particularly the XGBoost regression model, which was trained on historical data from 2023 to 2024 (see Section 4.1 and Section 4.2). This predictive modeling, according to Smith and Bilec [81], enables a data-driven estimation of material flows within clearly defined system boundaries—encompassing raw material importation and domestic resin production, resin transformation into plastic bags, distribution and use, and end-of-life (EOL) waste pathways including recycling, incineration, landfilling, and unregulated losses.

The Sankey diagram in Figure 2 visually represents the mass balance of the plastic bag system and highlights the linear and loss-prone nature of the current lifecycle. Total resin input, comprising imported and domestically produced raw materials, was projected at 436.91 tons. After accounting for production inefficiencies such as losses (13.16 tons) and raw material exports (33.36 tons), the volume of plastic bags produced stood at 390.39 tons. Notably, this figure aligns precisely with the volume of waste generated at end-of-life, reflecting the assumed average product lifespan of three months [70]. This lag accounts for the delay between consumption and disposal, enhancing the temporal realism of the system dynamics.

Such a balanced mass flow confirms the internal consistency of the forecasted data and underscores the accuracy of the machine learning model in simulating real-world conditions [73]. However, despite this balance, the diagram reveals a worrying reality: a significant portion of plastic bag waste (over 60%) ends up as mismanaged waste, far exceeding the fractions allocated to controlled treatment methods like recycling or incineration. This imbalance points to systemic weaknesses in waste management infrastructure, public engagement, and policy implementation, which continue to hamper efforts toward sustainable plastic use [82]. As such, the MFA quantifies flows and identifies intervention points for improving the environmental performance of plastic bag systems.

Table 5. Key MFA indicators for plastic bags (2025).

Indicators	Value
Production Yield (%)	89%
Loss Rate (%) Production Stage	3%
Export Share (%)	8%
Recycling Rate (%)	10%
Incineration Share (%)	10%
Landfill Share (%)	20%
Loss Rate (%) or End of Life (Mismanaged Waste)	60%

presents the key performance indicators derived from the 2025 MFA of plastic bags, offering a concise snapshot of system efficiency and end-of-life outcomes. The production yield, calculated as the ratio of finished plastic bags to total resin input, was estimated at 89%, indicating a relatively efficient transformation process from raw material to final product. This figure reflects minimal inefficiencies during manufacturing, as further emphasized by the modest 3% production-stage loss rate. Such performance aligns with industry norms for polymer processing, where yields above 85% are often considered acceptable due to technical constraints in the extrusion and forming stages [83].

Example Calculations for Material Flow Analysis Indicators

• $Production Yeild\left(\mathrm{\%}\right)={\displaystyle \frac{Plastic Bag Production}{Resin Used Production}}\times 100\mathrm{\%}={\displaystyle \frac{390.39}{438.63}}\times 100\mathrm{\%}=89;$
• $\mathrm{L}\mathrm{o}\mathrm{s}\mathrm{s} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\left(\mathrm{\%}\right)={\displaystyle \frac{Productio{n}_{Losse{s}_{tons}}}{Resin Used Production}}\times 100\mathrm{\%}={\displaystyle \frac{13.16}{438.63}}\times 100\mathrm{\%}=3\mathrm{\%};$
• $\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t} \mathrm{S}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{e}\left(\mathrm{\%}\right)={\displaystyle \frac{Raw Material Export}{(Raw Material Import+Raw Material Production)}}\times 100\mathrm{\%}={\displaystyle \frac{33.36}{(299.65+137.26)}}\times 100\mathrm{\%}=8\mathrm{\%};$
• $\mathrm{R}\mathrm{e}\mathrm{c}\mathrm{y}\mathrm{c}\mathrm{l}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\left(\mathrm{\%}\right)={\displaystyle \frac{\mathrm{R}\mathrm{e}\mathrm{c}\mathrm{y}\mathrm{c}\mathrm{l}\mathrm{i}\mathrm{n}\mathrm{g}}{Waste}}\times 100\mathrm{\%}={\displaystyle \frac{39.10}{390.39}}\times 100\mathrm{\%}=10\mathrm{\%};$
• $\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{i}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{S}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{e}\left(\mathrm{\%}\right)={\displaystyle \frac{\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{i}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{Waste}}\times 100\mathrm{\%}={\displaystyle \frac{38.37}{390.39}}\times 100\mathrm{\%}=10\mathrm{\%};$
• $\mathrm{L}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{f}\mathrm{i}\mathrm{l}\mathrm{l} \mathrm{S}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{e}\left(\mathrm{\%}\right)={\displaystyle \frac{\mathrm{L}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{f}\mathrm{i}\mathrm{l}\mathrm{l}}{Waste}}\times 100\mathrm{\%}={\displaystyle \frac{77.69}{390.39}}\times 100\mathrm{\%}=20\mathrm{\%};$
• $\mathrm{L}\mathrm{o}\mathrm{s}\mathrm{s} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e} \left(\mathrm{\%}\right) \mathrm{E}\mathrm{n}\mathrm{d} \mathrm{o}\mathrm{f} \mathrm{L}\mathrm{i}\mathrm{f}\mathrm{e}={\displaystyle \frac{\mathrm{L}\mathrm{o}\mathrm{s}}{Waste}}\times 100\mathrm{\%}={\displaystyle \frac{235.23}{390.39}}\times 100\mathrm{\%}=60\mathrm{\%}.$

However, the post-production indicators reveal more critical systemic challenges. Notably, 8% of raw materials were exported, limiting the domestic capacity to control the full life cycle of the material. More concerning is the fate of plastic bags at the end of life. The recycling rate stands at just 10%, with an additional 10% managed through incineration. While these figures are consistent with global recycling averages for flexible plastics, which remain under 15% due to material complexity and contamination [84], they underscore persistent barriers to circularity in the plastic bag value chain.

The most alarming statistic is the 60% mismanaged waste rate, representing materials that are neither formally collected nor treated through proper waste channels. This includes littering, open dumping, and unmanaged landfilling, all of which contribute directly to environmental degradation, particularly in low- and middle-income regions where waste governance remains weak [85]. In contrast, only 20% of plastic bags are sent to landfills under formal waste management systems, highlighting a stark disparity between production efficiency and end-of-life stewardship. Collectively, these indicators paint a picture of a system that performs well in terms of production efficiency but falls short in managing post-consumer waste. Strengthening regulatory frameworks, expanding recycling infrastructure, and promoting behavioral change among consumers are critical steps needed to improve downstream outcomes and transition toward a more sustainable plastic bag system.

Figure 3 illustrates the key performance indicators derived from the 2025 MFA of plastic bags, highlighting a stark contrast between upstream efficiency and downstream inefficiency. On the production side, the system demonstrates strong performance, with a high production yield and minimal losses during manufacturing, indicators that reflect a mature and technically optimized industrial process. This level of operational efficiency is typical for standardized plastic production, where controlled environments and streamlined supply chains minimize resource wastage [83]. However, the post-consumer phase tells a very different story. The majority of plastic bags are not captured by formal waste management systems, leading to a 60% mismanaged waste rate, which is alarmingly high. This figure suggests a severe breakdown in waste collection and treatment systems, particularly in urban areas where plastic bag consumption is highest.

In comparison to Geyer and Jambeck [84], who reported that only 9% of global plastic waste was recycled and 12% incinerated, the present study shows even lower rates of proper disposal in the Indonesian context. Similarly, Widayat, Praharjo [80] found that packaging plastics in developing countries often face collection inefficiencies due to insufficient infrastructure and a lack of incentives. Our analysis confirms and extends these findings by quantifying the imbalance through key indicators, offering empirical support for the argument that production efficiency alone is insufficient to drive sustainability.

The data underscores a critical bottleneck: while industries have largely succeeded in optimizing production, end-of-life management remains fragmented and under-resourced. This imbalance reveals the systemic challenge of closing the material loop and emphasizes the need for integrated policy measures, improved recycling infrastructure, and consumer engagement to reduce leakages. By comparing our results with earlier global and regional assessments, we reinforce the argument that addressing downstream inefficiencies is central to any serious CE strategy. As plastic bags continue to play a significant role in consumption habits, especially in emerging economies, prioritizing effective waste governance becomes essential for achieving both environmental sustainability and CE goals.

4.7. Sectoral and Flow-Specific Analysis

The MFA for plastic bags in 2025 (see

Table 6. Material flow balance for plastic bags (2025).

Material Flow	Value (tons)
Raw Material Import	299.65
Raw Material Production	137.26
Raw Material Export	33.36
Resin Used Production (tons)	438.63
Production Losses (tons)	13.16
Plastic Bag Production	390.39
Distribution	390.39
Consumption	390.39
Waste	390.39
Recycling	39.10
Incinerated	38.37
Landfill (tons)	77.69
Loss (tons)	235.23

and Figure 2) highlights a striking imbalance between upstream efficiency and downstream management challenges. The system is heavily reliant on raw material imports, with 299.65 tons feeding into a total resin input of 438.63 tons when combined with domestic production. This results in a robust plastic bag production volume of 390.39 tons after accounting for minor production losses and exports, indicating a well-functioning and relatively efficient manufacturing process. However, this operational strength stands in sharp contrast to the system’s post-consumer phase, where end-of-life handling falters significantly. Alarmingly, 60% of the waste generated (equivalent to 235.23 tons) is lost to the environment, predominantly due to insufficient collection infrastructure and reliance on informal disposal pathways. Such high leakage rates align with global observations, particularly for lightweight and single-use plastic packaging, which is often overlooked by formal waste management systems and more likely to become fugitive waste [86]. This systemic inefficiency at the waste management stage undermines upstream gains, pointing to a critical need for targeted interventions in collection, recycling, and consumer education to achieve a more circular and sustainable plastic bag system.

4.8. Scenario Analysis

To explore viable pathways for enhancing the circularity of the plastic bag value chain, this study develops and evaluates a set of simplified future scenarios grounded in recent European Union policy ambitions [87] and supported by contemporary industry research [88]. These scenarios are strategically designed to examine the potential impacts of targeted interventions at critical points within the system, specifically waste export reduction, collection system improvements, and recycling performance enhancement. The overarching goal is to assess how these combined measures can contribute toward meeting, or even surpassing, established recycling targets for 2025, moving the plastic bag sector beyond its current recycling baseline of approximately 10%.

Scenario A, termed reduced waste export, hypothesizes a substantial cutback in the volume of plastic waste exported internationally, shifting focus toward strengthening domestic recycling infrastructure and capacity. By minimizing dependence on often volatile global waste markets, this approach retains more post-consumer plastic bags within the national system. This retention enhances feedstock availability for local recycling facilities, reduces uncertainties associated with fluctuating export policies, and aligns with calls for more resilient and self-sufficient CE strategies [88]. In effect, this scenario envisions a more localized value chain that could foster innovation and investment in recycling technologies tailored to the specific characteristics of plastic bag waste streams.

Scenario B, termed improved collection, emphasizes enhancements to waste segregation and sorting systems, particularly targeting thin-film plastics like plastic bags, which notoriously evade capture in conventional mixed waste streams. This scenario models a 20% reduction in mismanaged waste, reflecting gains achievable through improved public awareness campaigns, expanded separate collection programs, and upgraded sorting technologies. Such measures would increase the proportion of plastic bag waste channeled into recycling or controlled landfilling, thereby curbing environmental leakage and improving material recovery [89]. This scenario highlights the critical role of upstream collection improvements as a foundational step toward closing material loops, reinforcing findings from global waste management studies that demonstrate the necessity of effective source separation [90].

In Scenario C, termed enhanced recycling performance, attention shifts to optimizing the operational efficiency and technological sophistication of recycling facilities themselves. This involves adopting advanced mechanical and emerging chemical recycling processes capable of increasing recycling yields by approximately 10% while concurrently reducing the share of waste directed to incineration. The scenario reflects ongoing technological advancements that promise to improve the quality and quantity of recycled outputs, addressing common challenges related to contamination and polymer degradation [88]. Domenech and Borrion [91] affirmed that enhanced recycling can mitigate landfill and incineration pressures and align with CE principles advocating for higher resource recovery rates and improved material value retention.

Scenario D, the combined scenario with production variations, integrates the interventions from Scenarios A to C within three distinct production trajectories: F1, representing a 10% increase in plastic bag production relative to the 2025 baseline, reflecting potential market growth or policy leniency; F2, a 10% reduction prompted by regulatory restrictions or shifting consumer behaviors; F3, a scenario of production stagnation with no change from baseline levels. This comprehensive approach captures the interplay between production volumes and circularity measures, illustrating how different industry dynamics can affect overall system performance. While the EU’s aggregate target aims for 8.8 million tonnes of recycled plastics across all sectors, these plastic bag-specific scenarios serve as a microcosm to benchmark sectoral progress and demonstrate the feasibility of doubling or tripling recycling rates within a defined timeframe [92]. Achieving these ambitious targets will require not only technical and operational advancements but also coordinated policy support, infrastructure investments, and shifts in both consumer and industry practices. Although the plastic bag sector’s quantitative contribution to national recycling targets may be relatively small, improvements in its circularity hold substantial environmental and economic value within its niche. Furthermore, Kinn [93] believed that success in this domain can serve as a replicable model for managing other flexible plastic packaging materials, which face similar challenges worldwide. Together, these scenarios provide a structured framework to assess the potential effectiveness of different strategic interventions, emphasizing the multifaceted nature of transitioning toward a more sustainable and circular plastic bag system. They underscore the importance of systemic change encompassing supply chain actors, technological innovation, and consumer engagement to address persistent issues of plastic waste leakage and resource inefficiency.

Although the projected recovery rate in the combined scenario could potentially reach 35%, the practical viability of such interventions depends on their economic feasibility. In particular, investments in advanced sorting technologies and the associated operating costs need to be assessed to ensure long-term sustainability. Studies using technical and economic analysis (TEA) from similar contexts, such as Indonesia and Thailand, have shown that the payback period for investing in plastic film sorting lines can range from 4 to 7 years, depending on input volume and material purity [94]. Incorporating a TEA framework would enable stakeholders to evaluate not only environmental gains but also return on investment, which is crucial for informing public and private sector decisions.

4.9. Theoretical Implications and Integration

The findings from the scenario analyses provide significant theoretical implications for advancing the understanding of CE applications within flexible plastic packaging, particularly plastic bags. Firstly, the study reinforces the importance of integrating multi-dimensional interventions across the entire value chain, from production to end-of-life management, to effectively enhance circularity [88]. The demonstrated need for reducing waste exports, improving collection systems, and enhancing recycling performance supports systems theory perspectives, which emphasize the interconnectivity and interdependence of supply chain actors and processes [95]. By incorporating multiple feedback loops, such as retaining waste domestically to increase local recycling feedstock and improving sorting efficiencies to minimize environmental leakage, this research contributes to expanding the theoretical framework of MFA and CE strategies. Additionally, the inclusion of production volume variations highlights how external factors such as policy shifts or consumer behavior changes can modulate circular system outcomes, aligning with contingency theory that stresses the need for adaptable management strategies under varying environmental conditions [96]. These insights extend the theoretical discourse by illustrating the dynamic and multi-scalar nature of CE transitions in plastic packaging sectors, emphasizing that no single intervention is sufficient in isolation.

Furthermore, the integration of scenario modeling with policy ambitions and industry realities bridges the often theoretical conceptualizations of circularity with practical, actionable pathways, enriching the theoretical integration between CE frameworks and sustainable supply chain management [19,97]. This study demonstrates the critical role of aligning technological innovation, infrastructure development, and stakeholder engagement to overcome persistent barriers such as mismanaged waste and low recycling rates, as documented in prior empirical research [86,90]. Cimpan and Bjelle [98] argued that by situating the plastic bag sector within broader environmental and economic contexts, the findings underscore the value of sector-specific circularity benchmarks, moving beyond aggregate targets to recognize the unique challenges and potentials of flexible plastic packaging. This approach contributes to the theoretical conversation around differentiated CE strategies that are context-sensitive and tailored to material-specific flows and lifecycle characteristics. Ultimately, these results advocate for a holistic, multi-pronged theoretical model that integrates environmental policy, technological capacity, and behavioral dimensions to effectively drive circular transitions in complex waste management systems.

5. Conclusions

The MFA of shopping plastic bags in 2025 reveals a complex and nuanced system characterized by strong upstream production efficiency but significant downstream challenges in waste management. While the manufacturing process demonstrates a high yield with minimal losses, the end-of-life phase paints a less optimistic picture. A large proportion of plastic bag waste (over 60%) is mismanaged, highlighting critical deficiencies in collection infrastructure, recycling capacity, and public participation. This imbalance underscores a systemic failure to close the material loop and points to the urgent need for integrated strategies that combine regulatory reforms, improved waste segregation, and enhanced recycling technologies. The forecasted flows and mass balance data provide a reliable foundation to inform decision-makers about where interventions can be most effective in reducing environmental leakage and promoting circularity.

Looking forward, the scenario analysis offers promising pathways for improving the sustainability of the plastic bag system. Strategies that reduce waste exports, improve collection and sorting, and enhance recycling performance collectively show the potential to substantially increase recycling rates and reduce environmental impacts. These interventions, especially when combined with adaptive production scenarios, highlight the interconnectedness of policy, technology, and consumer behavior in shaping future outcomes. Strengthening policy frameworks should involve the implementation of regulatory tools such as extended producer responsibility (EPR) schemes, which shift waste management costs to producers, tax incentives to promote eco-design and recycled content, and levies or bans on single-use plastics to discourage excessive consumption. Moving toward a CE for plastic bags will require concerted efforts across multiple sectors and stakeholders, emphasizing innovation, infrastructure investment, and public engagement. Success in this area could not only mitigate the environmental risks posed by plastic waste but also serve as a scalable model for managing other challenging plastic packaging materials globally. Future research could explore the integration of emerging biodegradable materials into the existing plastic bag system and assess their potential impacts on material flow and waste management. Additionally, investigating consumer behavior dynamics and their influence on collection and recycling participation could provide valuable insights to design more effective awareness and incentive programs.