A Full-Process Carbon Footprint Assessment of Online and Offline Apparel Sales: Integrating Return Logistics

Abstract

This study develops a comprehensive carbon footprint assessment model that integrates forward and reverse logistics to evaluate and compare greenhouse gas emissions from online and offline apparel sales channels in China, with a particular focus on high return rates. The model quantifies emissions from transportation, packaging, storage, and operations, incorporating return and exchange logistics. The system boundary is limited to enterprise-controllable sales-phase activities and excludes consumer travel. Three sales models are compared: factory-to-consumer (F2C), traditional business-to-consumer (B2C) e-commerce, and brick-and-mortar retail (BMR). Within this defined boundary, BMR exhibits the lowest carbon footprint (0.296 kg CO₂e/item), followed by F2C (0.408 kg CO₂e/item) and B2C (0.602 kg CO₂e/item). Packaging dominates online emissions (55–57%), whereas store operations are the main contributor to offline emissions (43%). Return rates are identified as a decisive factor, accounting for over 31% of e-commerce emissions and potentially increasing them by 171.3% under extreme scenarios. Sensitivity analysis reveals that trunk line distance (factory to warehouse) has a greater impact on emissions than last-mile return route optimization. Relocating the factory closer to consumers reduces B2C transport emissions by 72.3%, whereas replacing conventional packaging with recycled plastic reduces total B2C emissions by 46.0%. These findings provide channel-specific sustainability strategies: return reduction and packaging innovation for online channels, and energy efficiency improvements for physical stores. These results are conditional on the defined system boundary. If consumer travel by private car were included, the relative advantage of offline channels would diminish or could reverse.

1. Introduction

The fashion industry faces increasing global scrutiny due to the environmental impacts of its supply chain operations [1]. The textile and apparel sector accounts for approximately 10% of global carbon emissions—more than the combined total of international aviation and maritime shipping [2]. Under a business-as-usual scenario, greenhouse gas (GHG) emissions from the global textile and apparel industry are projected to reach 1.588 billion tonnes of CO₂ equivalent by 2030 [3]. According to the IPCC’s 2018 report, limiting global warming to 1.5 °C above pre-industrial levels requires a 45% reduction in CO₂ emissions by 2030 [4]. Applying this target proportionally to the textile and apparel sector would require its emissions to decrease from 1.025 billion tonnes in 2019 to 564 million tonnes of CO₂ equivalent by 2030 [5]. The rapid growth in textile production and fashion consumption—epitomized by the rise of fast fashion—results from low-cost manufacturing, high consumption frequency, and the short lifespan of garments. If current population growth and demand trends continue, global textile and apparel consumption is expected to triple by 2050, thereby exhausting the carbon budget aligned with the 2 °C global warming limit [6]. As documented by Qiuxia et al. [7], carbon footprint research has experienced rapid growth across multiple sectors, including energy, industry, and consumption. The environmental footprint of specific industries, including textiles and apparel, has received growing research attention [8]. However, systematic comparative assessments across different retail channels remain limited.

China’s apparel e-commerce market continues to expand under the dual drivers of consumption upgrading and technological innovation. This makes it the world’s largest online apparel consumer market. The viral reach of social commerce, the immersive experience of live-streaming e-commerce, and the application of intelligent recommendation technologies are collectively reshaping consumer decision-making pathways [9]. As e-commerce evolves rapidly, apparel—a key category in online retail—has seen its sales models transition from traditional brick-and-mortar stores to multi-channel and omnichannel retailing. Fast fashion brands launch tens of thousands of new styles annually, resulting in product homogenization and inconsistent sizing on e-commerce platforms. Consequently, consumers often purchase multiple items intending to keep only one, which drives significantly higher returns and exchange rates. Apparel has consistently ranked highest in return rates among all product categories on Chinese e-commerce platforms. According to data from Taobao, the return rate for women’s apparel on traditional e-commerce channels generally ranged between 50% and 60% in 2024, with some merchants reporting rates as high as 75%. During the Double Eleven promotional period, the return rate for live-streaming e-commerce exceeded 80%, and in some cases, approached 90%. These figures are substantially higher than the return rates for categories such as home appliances, digital products, and food. Consequently, high return rates have become a major constraint on the sustainable development of apparel e-commerce [10]. In this context, a key scientific question urgently needs to be addressed: what are the differences in the full-process carbon footprints between online and offline apparel sales channels, how do their emission structures differ, and how does the return rate affect total carbon emissions? Answering these questions is of significant theoretical and practical importance for apparel companies in developing science-based low-carbon sales strategies.

In recent years, comparative studies on the carbon footprints of online and offline retail models have grown, with scholars exploring this issue across different industries and from various perspectives. Early research primarily focused on “last-mile” delivery or emissions from individual stages, and conclusions often varied depending on the research context. Zheng et al. [11] studied fresh produce in China and developed a full-process carbon emission assessment framework covering transportation, storage, operations, and packaging. They found that online just-in-time delivery models had lower carbon emissions than traditional offline store models, but noted that packaging emissions were significantly higher than those of offline retail. Van Loon et al. [12] examined fast-moving consumer goods in the UK and identified basket size, delivery method, and return rate as key determinants of the carbon footprint. They emphasized that simply comparing channels without considering consumer behavior may lead to misleading conclusions. In the field of apparel retail, Wiese et al. [13] compared transport-related CO₂ emissions of online and offline shopping using data from a German multi-channel apparel retailer. They found that online retail had lower emissions in most scenarios. However, when consumers travelled short distances, the offline channel had an advantage. Mangiaracina et al. [14] developed an activity-based carbon emission model for the Italian apparel industry. Their results showed that logistics activities dominate emissions in both online and offline channels, accounting for 99% and 62% of total emissions, respectively, with consumer residential location being the most critical parameter affecting online channel emissions. He [15] analyzed the carbon footprint of tomato supply chains, focusing on the impact of different transportation modes during distribution. The study revealed that production-phase emissions predominate in sea and rail transport. Carbonstop [16] reported that transportation constitutes the largest share of carbon emissions in both online and offline shopping scenarios. Lu and Bai [17] systematically mapped the apparel sales process from factory to consumer, analyzing greenhouse gas emission sources in both traditional and online retail models. Rosas et al. [18] compared the carbon footprint of omnichannel and pure-player fashion e-commerce models in Spain, finding that in-store pickup and return models generate significantly lower emissions than home delivery. Their work provides a foundational analysis of the carbon emission structure in the apparel sales phase.

From a broader perspective, some scholars have attempted to incorporate consumer travel behavior into the analytical framework. Seebaue et al. [19] conducted a case study in Austria and quantified the carbon footprints of different shopping scenarios, including neighborhood stores, city centers, discount stores, shopping malls, and mail order or online sales. They found that private car use for consumer travel is the main source of emissions in offline channels. Moreover, even under policy scenarios that promote online sales, carbon emissions still show an increasing trend. Hischier [20] performed a simplified life cycle assessment based on the Swiss context, comparing the carbon emissions of online and offline clothing purchases. The study confirmed that the environmental impact of offline shopping is highly dependent on travel mode and distance. It also found that transportation and packaging materials are the main contributors to online shopping emissions. Return logistics represents a critical pain point in apparel e-commerce. Long and Liu [21] used proprietary data from two major U.S. apparel retailers and applied life cycle assessment to quantify the carbon emissions of the return process. They found that transportation distance is the primary driver of emissions, accounting for over 90% of emissions in a centralized return system. Their results also showed that different logistics network designs have significant effects on emissions. Yu et al. [22] systematically reviewed urban road traffic CO₂ emission models and identified transport distance as a key determinant of carbon emissions. In terms of packaging sustainability, Park and Waqar [23] conducted case studies based on the Canadian market. Their findings indicate that the environmental advantages of reusable courier bags are highly dependent on transportation distance and the number of reuse cycles.

Although the above studies provide important theoretical foundations and empirical data for understanding the carbon footprint of apparel sales, several critical research gaps remain that warrant the attention of this study. First, existing literature lacks a systematic integration of reverse logistics into a theoretical model of the environmental performance of channel structures. Most studies treat returns as an exogenous variable or ignore them entirely, failing to incorporate the return rate as a core design parameter in a full-process carbon footprint assessment framework. Second, a dynamic assessment framework tailored to the high return rates characteristic of the apparel industry has not yet been established. Existing models are mostly static and particularly lack sensitivity analyses and scenario simulations for the extreme return rates observed in China’s e-commerce market, where return rates can exceed 80% in live-streaming channels. Third, current research predominantly relies on data from Western markets or focuses on specific product categories. The applicability of these findings to China’s rapidly growing apparel e-commerce market remains to be validated. This market features a unique logistics structure, including electric tricycles for last-mile delivery and a dense express delivery network. Fourth, existing comparative studies do not systematically deconstruct the contributions of the four major emission sources in the sales stage: transportation, packaging, storage, and operations. In particular, the literature lacks a cross-comparative analysis of three typical sales models in the Chinese context, namely factory-to-customer (F2C), business-to-customer (B2C) e-commerce, and traditional brick-and-mortar retail (BMR).

To address the above research gaps, this study builds on existing literature and develops a full-process carbon footprint assessment model for apparel sales that integrates both forward and reverse logistics. The model is then applied to an empirical analysis within the context of China’s e-commerce market. The main contributions and innovations of this study are reflected in three aspects. First, theoretical contribution. Drawing on sustainable supply chain management theory [24], this study constructs an analytical framework based on three dimensions: channel structure, operational characteristics, and carbon emission performance. The framework reveals how the interaction between channel structure and operational characteristics influences carbon emissions. Channel structure includes factory-to-customer (F2C), centralized business-to-consumer delivery (B2C), and traditional brick-and-mortar retail (BMR). Operational characteristics focus particularly on the return rate. This extends the application of sustainable supply chain management theory to the field of retail channel selection. Second, methodological innovation. Unlike previous studies that mostly employ static and single-stage emission estimations, this study is the first to systematically integrate the full-process reverse logistics of returns and exchanges into an apparel sales carbon footprint model. A dynamic weighted transportation distance function is constructed, which enables the model to sensitively capture the nonlinear amplifying effect of changes in the return rate on total carbon emissions. Third, practical value. This study compares the full-process carbon footprints of three typical sales models: factory-to-customer (F2C), business-to-customer (B2C) e-commerce, and traditional brick-and-mortar retail (BMR). The comparison identifies the key emission sources for each channel and proposes differentiated carbon reduction strategies for different channel types. In particular, this study quantifies the marginal contribution of return rates to carbon emissions, addressing the practical challenge of high return rates in China’s apparel e-commerce market. These findings provide quantitative decision-making support for e-commerce platforms and apparel companies in formulating return management policies, optimizing packaging design, and planning regional return networks.

2. Materials and Methods

This study is structured according to the principles of ISO 14040 and ISO 14067 for Life Cycle Assessment (LCA) [25,26]. The quantification and integration of indicators in this model follow the core principles of product carbon footprint accounting outlined in ISO 14067. It should be noted that this study does not constitute a full ISO-compliant LCA. Rather, it adopts the core principles of the ISO standards methodologically, with a focus on enterprise-controllable sales-phase activities. Specifically, based on global warming potential, greenhouse gas emissions from all stages are uniformly converted into carbon dioxide equivalent values for aggregation and comparison. This section first defines the objective and scope of the analysis, clearly delineating the system boundaries for both online and offline sales models. It then presents the model architecture in detail, specifying the functional unit used to enable comparative analysis, along with the data sources and underlying assumptions.

2.1. Goal and Scope Definition

Amid growing global pressure on the fashion industry to reduce emissions and rising environmental concerns over China’s apparel e-commerce sector—largely driven by high return rates—this study addresses key gaps in the current literature. The specific research objectives and scope are as follows. (1) This study develops a full-process carbon footprint assessment model for the apparel sales phase. The model covers the forward logistics process from regional central warehouses to end consumers, including transportation, packaging, storage, and operational activities. Notably, it fully integrates the reverse logistics generated by high return and exchange rates, a critical dimension of the analysis. This integration includes all additional transportation, repackaging, and handling activities associated with returns and exchanges within the defined system boundary. (2) The developed sales-phase model is applied to quantify the average carbon emissions per garment sold through two distinct sales models in the Chinese e-commerce context: (a) online sales (e.g., factory-to-customer or business-to-customer) and (b) traditional offline brick-and-mortar retail. The analysis specifically focuses on how return and exchange behaviors differentially affect the total carbon footprint of each sales model. (3) This study systematically identifies the key stages and emission sources of the carbon footprint in both online and offline channels. The findings provide a scientific foundation for formulating targeted carbon reduction strategies in apparel retail.

2.2. System Boundary

Figure 1 illustrates the processes involved in the apparel sales phase under the online and offline models examined in this study. Currently, online apparel sales in China fall into two primary categories. The first is factory-to-customer (F2C), where garments are packed directly at the factory and shipped straight to consumers, bypassing central warehouses and retail stores to provide a service model tailored to individualized demand [24]. The second is classic business-to-customer (B2C) e-commerce, where garments are transported from factories through a distribution network that includes central and regional warehouses for storage and allocation before final delivery to online customers. Offline sales follow the traditional brick-and-mortar retail (BMR) model, in which garments are shipped from factories to physical stores, where consumers make in-person purchases. Both types of sales models include four operational stages: transportation, packaging, storage, and operations. All of these stages fall within the system boundary of this study. Consumer travel is excluded from the system boundary due to its high variability, which would otherwise compromise cross-channel comparability. Although excluding consumer travel is a deliberate choice to focus on enterprise-controllable activities and ensure cross-channel comparability, it may lead to an overestimation of the relative carbon advantage of offline channels.

2.3. Model Architecture

This study models encompasses the full life-cycle processes for three distinct retail models: factory-to-customer (F2C), classic business-to-customer (B2C) e-commerce, and traditional brick-and-mortar retail (BMR). These processes cover transportation, packaging, storage, and operations. To enable a consistent and quantifiable comparison across these sales models, the functional unit is defined as “one standard apparel item successfully sold to a consumer”.

2.3.1. Model Architecture from Transportation

According to the International Energy Agency [27], transportation within supply chains is the second-largest source of carbon emissions, after thermal power generation. In the transportation segment, emissions arise not only from primary forward logistics but also from multiple rounds of reverse logistics triggered by returns and exchanges. The geographically weighted average transport distance is calculated based on order distribution ratios, as shown in Equation (1).

$$D_{i,fc/wc/wr}={\displaystyle \frac{\sum {Ordercount}_{ij}\times D_{ij,fc/wc/wr}}{{Ts}_i}}$$

(1)

In the notation used, $D_{i,fc}$ denotes the weighted average transport distance from the factory to the consumer under sales model i (factory-to-customer, denoted as i = 1); $D_{i,wc}$ represents the weighted average transport distance from the central warehouse to consumers under sales model i (classic business-to-customer e-commerce, denoted as i = 2); and $D_{i,wr}$ represents the weighted average transport distance from the central warehouse to physical retail stores under sales model i (brick-and-mortar retail model, denoted as i = 3). Here, ${Ordercount}_{ij}$ is the number of items shipped to region j under sales model i during the statistical period; $D_{ij,fc/wc}$ is the transport distance from the factory or central warehouse to consumers in region j under model i; $D_{ij,wr}$ is the transport distance from the central warehouse to retail stores in region j under model i; and ${Ts}_i$ is the total quantity of goods sold under sales model i.

Currently, there is no universally standardized definition for return and exchange rates in the literature. In this study, the return rate is defined as the proportion of items returned by consumers to the seller within a given time period relative to the total number of orders. The exchange rate is defined as the proportion of items for which consumers request a replacement with another item of the same category within the same period, also relative to the total number of orders. The total quantity of goods sold $({Ts}_i$) refers to the total number of items successfully transacted during the period, including all sold items, regardless of whether they are subsequently returned or exchanged.

The return rate and exchange rate are calculated according to Equation (2) and Equation (3), respectively.

$$R_{1i}={\displaystyle \frac{B_i}{{Ts}_i}}$$

(2)

Here, $R_{1i}$ denotes the return rate under sales model i, where i = 1 represents the factory-to-customer (F2C) model, i = 2 the classic business-to-customer (B2C) e-commerce model, and i = 3 the brick-and-mortar retail model (BMR). $B_i$ is the number of returned items under sales model i, and ${Ts}_i$ is the total quantity of successfully transacted items during the period under the sales model i.

$$R_{2i}={\displaystyle \frac{C_i}{{Ts}_i}}$$

(3)

Here, $R_{2i}$ denotes the exchange rate under sales model i, where i = 1 represents the factory-to-customer (F2C) model, i = 2 the classic business-to-customer (B2C) e-commerce model, and i = 3 the brick-and-mortar retail model (BMR). $C_i$ is the number of exchanged items under the sales model i.

This study considers only a single return or exchange event per order. Both returns and exchanges are explicitly incorporated into the transportation model by adjusting the number of transport trips accordingly. The dynamic weighted average transport distance is derived by accounting for these additional trips induced by returns and exchanges, and is expressed as a function of the return rate ($R_{1i}$) and exchange rate ($R_{2i}$). The formulations for the dynamic weighted average transport distances under the factory-to-customer (F2C), classic business-to-customer (B2C), and brick-and-mortar retail (BMR) models are presented in Equation (4), Equation (5), and Equation (6), respectively.

$$D_{average,1}=D_{1,fc}\times {\displaystyle \frac{1}{1-R_{11}-R_{21}}}+D_{1,cw}\times {\displaystyle \frac{R_{11}+R_{21}}{1-R_{11}-R_{21}}}$$

(4)

Here, $D_{average,1}$ denotes the dynamic weighted average transport distance under sales model 1 (factory-to-customer, F2C). $D_{1,fc}$ represents the weighted average distance from the factory to the consumer in model 1, while $D_{1,cw}$ refers to the weighted average distance from the consumer back to the central warehouse (or designated return center) in the same model. $R_{11}$ is the return rate under sales model 1, and $R_{21}$ is the exchange rate under the sales model 1.

$$D_{average,2}=D_{fw}\times {\displaystyle \frac{T_2}{N_2}}+D_{2,wc}\times \left[1+{\displaystyle \frac{2\times \left(R_{12}+R_{22}\right)}{1-R_{12}-R_{22}}}\right]$$

(5)

Here, $D_{average,2}$ denotes the dynamic weighted average transport distance under sales model 2 (classic business-to-customer e-commerce, B2C). $D_{fw}$ represents the fixed transport distance from the factory to the central warehouse. $T_2$ is the total quantity of goods shipped from the factory to the central warehouse under model 2, and $N_2$ is the number of items successfully sold to consumers under the same model. $D_{2,wc}$ refers to the weighted average transport distance from the central warehouse to consumers in model 2. $R_{12}$ and $R_{22}$ denote the return rate and exchange rate, respectively, under sales model 2.

$$D_{average,3}=D_{fw}\times {\displaystyle \frac{T_3}{N_3}}+D_{3,wr}\times \left[{\displaystyle \frac{{1+R}_{13}+R_{23}}{1-R_{13}-R_{23}}}\right]$$

(6)

Here, $D_{average,3}$ denotes the dynamic weighted average transport distance under sales model 3 (brick-and-mortar retail model, BMR). $D_{fw}$ represents the fixed transport distance from the factory to the central warehouse. $T_3$ is the total quantity of goods shipped from the factory to the central warehouse under model 3, and $N_3$ is the number of items successfully sold to consumers through physical stores in this model. $D_{3,wr}$ refers to the weighted average transport distance from the central warehouse to retail stores under model 3. $R_{13}$ and $R_{23}$ denote the return rate and exchange rate, respectively, under sales model 3. Notably, consumer travel to and from retail stores is excluded from the system boundary and thus not included in the calculation of $D_{average,3.}$

The difference in the coefficients for the return- and exchange-related terms between Equations (5) and (6) originates from the fundamental difference in the logistics pathways for returns and exchanges between the two models. In the B2C model, there is no physical interaction node between the consumer and the warehouse. Therefore, returns and exchanges must be completed through courier services for the entire round trip. Consequently, one return or exchange event generates two transport trips: from the consumer to the warehouse and from the warehouse to the consumer. In contrast, in the BMR model, the physical store serves as an intermediary for returns and exchanges. Consumers complete product handover and replacement on-site, eliminating the need for a separate forward delivery. Only a one-way reverse transport trip from the store to the warehouse is required. This difference captures the moderating effect of different terminal formats on reverse logistics carbon emissions in multi-channel retailing.

During transportation, different vehicle types and load capacities are employed depending on the transport segment. For instance, heavy-duty trucks are typically used for trunk-line (long-haul) transport, while light-duty vehicles or vans serve regional distribution (branch-line) routes. For the last-mile delivery, electric three-wheelers are predominantly used in China’s urban e-commerce logistics, supplemented occasionally by conventional fuel-powered vehicles. The carbon footprint of the target apparel item during the transportation phase is thus formulated as shown in Equation (7).

$${CF}_{i,transportation}=\sum _k{\displaystyle \frac{\frac{D_{average,i}}{100}\times P_{i,k}\times {AD}_{i,k}\times {EF}_{i,k}\times \frac{m_{i,k}}{M_{i,k}}}{S_{i,k}}}$$

(7)

Here, ${CF}_{i,transportation}$ denotes the carbon footprint (in kg CO₂e) of transporting one unit of the target apparel item under sales model i. $D_{average,i}$ is the dynamic weighted average transport distance for model i. $P_{i,k}$ represents the proportion of total transport distance attributed to vehicle type k in model i. ${AD}_{i,k}$ is the energy consumption intensity of vehicle type k (expressed in L/100 km or kWh/100 km), and ${EF}_{i,k}$ is the corresponding carbon emission factor for that energy source (in kg CO₂eq/L or kg CO₂eq/kWh). $m_{i,k}$ denotes the total weight of the target apparel items carried by vehicle type k under model i, while $M_{i,k}$ is the total weight of all goods transported by that vehicle type. Finally, $S_{i,k}$ represents the number of units of the target apparel item shipped via vehicle type k in sales model i.

2.3.2. Model Architecture from Packaging

The carbon footprint of the packaging phase is calculated as shown in Equation (8), with the total emissions from all packaging materials proportionally allocated on a per-unit basis to each successfully sold item of the target apparel product.

$${CF}_{i,packaging}={\displaystyle \frac{(\sum {AD}_{ij}\times {EF}_{ij})\times T_i}{N_i}}$$

(8)

Here, ${CF}_{i,packaging}$ denotes the carbon footprint (in kg CO₂e) of the packaging phase for one unit of the target apparel product under sales model i. ${AD}_{ij}$ represents the amount of activity data associated with packaging material or process type j per unit of product under model i (e.g., mass of plastic bag, length of tape, or area of cardboard used). ${EF}_{ij}$ is the corresponding carbon emission factor (in kg CO₂eq per unit of activity) for packaging input j under model i. $T_i$ is the total number of items transacted (including sold, returned, and exchanged units) under sales model i, while $N_i$ is the number of items successfully sold under the same model. The total packaging-related emissions are allocated evenly across the $N_i$ successfully sold units to yield the per-unit packaging carbon footprint.

2.3.3. Model Architecture from Storage

In online sales models, apparel products under the factory-to-consumer (F2C) model are stored in on-site factory warehouses, whereas those under the classic business-to-customer (B2C) e-commerce model are held in a centralized warehouse. In the offline retail model, garments are stored directly in physical store inventories. The carbon footprint of the storage phase is calculated as shown in Equation (9).

$${CF}_{i,storage}={\displaystyle \frac{v_i\times {time}_i\times 4}{V_{i,storage}\times {Time}_{i,storage}}}\times {AD}_i\times {EF}_i$$

(9)

Here, ${CF}_{i,storage}$ denotes the carbon footprint of the storage phase for one unit of the target apparel product under sales model i. $v_i$ is the physical volume of a single item under model i (in m³). ${time}_i$ represents the average storage duration of the product in the warehouse (in weeks). The factor of 4 accounts for the storage volume multiplier, reflecting that each item typically occupies approximately four times its own physical volume due to aisle space, racking, and handling requirements. $V_{i,storage}$ is the total usable volume capacity of the warehouse under model i, and ${Time}_{i,storage}$ is the total operational period of the warehouse during the analysis timeframe (in weeks). ${AD}_i$ denotes the total energy consumption of the warehouse during the period when the target product is stored, and ${EF}_i$ is the corresponding carbon emission factor for electricity or energy used in the storage operations (in kg CO₂eq/kWh).

2.3.4. Model Architecture from Operation

The activity data for the online sales model include electricity consumption from equipment used in order processing, such as computers, printers, and barcode scanners. For the offline brick-and-mortar retail model, operational electricity use includes energy consumed by point-of-sale (POS) systems, audio equipment, and other in-store devices. The carbon footprint of the operations phase is calculated as shown in Equation (10).

$${CF}_{i,operation}={\displaystyle \frac{(\sum {AD}_{ij}\times {EF}_{ij})\times T_i}{N_i}}$$

(10)

Here, ${CF}_{i,operation}$ denotes the carbon footprint (in kg CO₂e) of the operation phase for one unit of the target apparel product under sales model i. ${AD}_{ij}$ represents the amount of activity data associated with operational input type j per unit sold under model i. ${EF}_{ij}$ is the corresponding carbon emission factor (in kg CO₂eq per unit of energy or activity) for input type j under sales model i. $N_i$ is the total number of successfully sold units under model i, and the total operational emissions are allocated evenly across these $N_i$ units to yield the per-unit sales-phase carbon footprint.

2.4. Data Sources and Assumptions

This study conducts a case analysis by applying the proposed model using literature-derived and secondary data. A simulated case based on the geographical structure of a representative apparel company, parameterized using secondary and industry-average data. The logistics network model is constructed based on the actual geographical layout of Chinese apparel company A’s supply chain, including the locations of its factory, central warehouse, and retail stores, to ensure real-world representativeness. However, the core activity data employed in the model—such as transport distances, energy consumption, return and exchange rates, and packaging materials—are not proprietary to company A. Instead, they are drawn from publicly available sources, including industry reports, national statistical yearbooks, authoritative databases such as the China Products Carbon Footprint Factors Database [28], and peer-reviewed academic literature. This data strategy aims to reveal generalizable patterns and trends within a realistic structural framework rather than reflect the performance of a single enterprise. All key assumptions, including distribution channel shares, vehicle load factors, and reverse logistics pathways, are explicitly stated in this analysis. According to publicly available information, company A’s manufacturing facility is located in Wenzhou, Zhejiang Province. The central warehouse is modeled after Jingdong Logistics’ central warehouse in Shanghai. Consumer and retail locations are selected within the warehouse’s service radius, specifically in Hangzhou (Zhejiang), Suzhou (Jiangsu), and Shanghai itself, as illustrated in Figure 2. According to the report on total greenhouse gas emissions of China’s e-commerce enterprises [16], the primary vehicle for trunk-line freight in China’s express logistics network is the 17.5 m diesel-powered heavy-duty truck. For branch-line distribution, 4.2 m fuel-powered vans dominate, with a small share of electric vans. The assumed ratio of fuel to electric vehicles is 9:1. For last-mile delivery, courier companies primarily rely on electric three-wheelers, supplemented by a minor share of fuel-powered vehicles, with an assumed ratio of 1:9 (fuel to electric). The analysis considers three distribution models: factory-to-customer (F2C), classic business-to-customer (B2C) e-commerce, and brick-and-mortar retail (BMR). It is assumed that company A operates all three channels. Data for a specific apparel item—denim jeans—over a single quarter are collected for modeling purposes. The allocation of sales volume across the three channels follows the Product Environmental Footprint Category Rules [29]. This study assumes that under the three distribution models, all returned and exchanged products are sent back to the central warehouse for processing. After quality inspection, all returned products are resold as new items rather than being discarded or downgraded. This study also assumes that each order involves at most one return or one exchange. Multiple returns or exchanges for the same order are not considered. One exchange process involves two shipments. The first shipment is the reverse logistics process, in which the consumer sends the original product back to the warehouse. The second shipment is the forward logistics process, in which the new product is delivered from the warehouse to the consumer. As shown in Figure 3, the return and exchange paths under different sales models, this study assumes that the exchange process follows the same reverse logistics path as the return process. That is, the consumer sends the original product back to the warehouse. The delivery of the new product follows the same path as the original order delivery. Therefore, in the model calculation, the transportation emissions generated by one exchange equal the sum of emissions from one return and one normal delivery. This study only considers a single return or exchange event and assumes that the reverse logistics path for returns and exchanges is consistent with the forward logistics path. This study assumes that the original packaging of a returned product is discarded during the return process. For an exchange, the new product shipped to the consumer uses entirely new packaging materials. This assumption is based on the following industry observations. In apparel e-commerce practice, the original packaging of a returned product is often damaged during the unboxing process and cannot be restored to a reusable condition. Even if the original packaging remains intact, companies rarely reuse it for hygiene and brand image considerations. This study defines the return rate and the exchange rate as mutually exclusive events. That is, the same order cannot be counted as both a return and an exchange. In the model calculation, returns and exchanges trigger separate reverse logistics activities. When both occur simultaneously, meaning the consumer requests an exchange rather than a return, the model processes the transaction as an exchange only. Return emissions are not double-counted. The main simplifying assumptions of the model and their justifications are summarized in Table S1 in the Supplementary Materials.

3. Case Study

3.1. Basic Data for the Case Study

This study constructs a simulated case based on the geographical distribution of the Chinese apparel company A. The case uses the spatial configuration of company A’s factory, central warehouse, and consumer locations as a realistic logistics network framework to analyze the carbon footprints of different sales models under this specific structure. Operational parameters, including return rates, vehicle energy consumption, and storage electricity use, are assigned industry averages or representative values from the literature. This approach ensures sector-wide relevance and enables fair cross-channel comparisons. The analysis focuses on denim jeans sold by company A over a single quarter, calculating the carbon footprint per successfully sold unit during the sales phase. According to industry reports, company A produced 1 million pairs of jeans in the quarter, distributed across three sales channels: factory-to-customer (F2C), classic business-to-customer (B2C) e-commerce, and brick-and-mortar retail (BMR). The allocation across these channels follows the Product Environmental Footprint Category Rules (PEFCR). Industry data indicate that return rates for traditional B2C e-commerce range from 15% to 25%, while live-streaming e-commerce exhibits significantly higher rates of 30% to 50% [30]. Based on the market share ratio of approximately 60:40 between traditional B2C and live-streaming e-commerce reported in the 2024 China Online Retail Market Data Report [31], and adopting conservative estimates (15% for traditional B2C and 30% for live-streaming), the weighted average return rate is calculated as 21%. In the absence of standardized exchange rate data, the exchange rate is conservatively assumed to be half the return rate. For the F2C model, which primarily serves personalized direct-to-consumer orders, and the BMR model, where customers try on garments in-store, return and exchange rates are substantially lower. Therefore, default values from the PEFCR are applied [29]. Background data for the sales-phase modeling are summarized in

Table 1. Background data for the sales phase.

Parameter	F2C (i = 1)	B2C (i = 2)	BMR (i = 3)
One million pairs of jeans were manufactured within a single quarter.
Allocation ratio	3.8%	34.2%	62%
Total quantity of goods sold (Tsi)	38,000	342,000	620,000
Return rate (R1i)	1.2%	21%	4.5%
Exchange rate (R2i)	0.6%	10.5%	2.3%
Number of successfully sold items (Ni)	37,316	234,270	577,840

Note: N_i = T_si − T_si $\times$ (R_1i + R_2i).

. Transport volumes are estimated based on both volumetric occupancy and weight capacity. For example, a 17.5 m diesel truck has an internal cargo volume of approximately 130 m³. Assuming each pair of jeans occupies 0.004 m³ and the vehicle operates at 50% of its maximum payload capacity, the effective shipment quantity per trip is calculated as 16,250 units. Detailed transport parameters are provided in

Table 2. Data inventory for the transport stage.

Sales Model	Type of Transportation Vehicle (k)	Energy Consumption	Weighted Average Transport Distance	Dynamic Weighted Average Transport Distance
F2C (i = 1)	17.6 m diesel truck	47 L/100 km	324.350 km (Dfc)	332.510 km
	4.2 m diesel truck	18.6 L/100 km
	electric tricycle	10 kwh/100 km
B2C (i = 2)	17.6 m diesel truck	47 L/100 km	120.835 km (Dwc)	944.377 km
	4.2 m diesel truck	18.6 L/100 km
	electric tricycle	10 kwh/100 km
BMR (i = 3)	17.6 m diesel truck	47 L/100 km	109.100 km (Dwr)	640.665 km
	4.2 m diesel truck	18.6 L/100 km

.

The packaging inventory is defined per single pair of jeans. Primary packaging includes a transparent polyethylene (PE) bag and a hangtag. Secondary packaging consists of a gray courier mailer. Additional materials include adhesive tape and shipping waybills. Key parameters include the volume occupied by one pair of jeans, warehouse floor area, and total storage volume. According to the National Bureau of Statistics of China, there are 23,526 specialized retail outlets for textiles, apparel, and daily necessities, with a total floor area of 5.854 million m²—yielding an average store size of approximately 249 m² per outlet. Based on guidelines from the Chinese Academy of Sciences [32], the minimum clear indoor height for spaces with regular human occupancy should be no less than 2.0 m, allowing estimation of the average retail store volume as 498 m³ (249 m² × 2.0 m). Energy consumption data indicate that chain retail stores consume approximately 121.56 kWh/m² annually [33], while warehouses consume about 40 kWh/m² per year [29]. Operational electricity use primarily covers order processing and printing activities. Publicly reported data show that Jing Dong handled 595.829 billion items in 2019, with its data centers consuming 1,210,490.91 MWh of electricity [16]. This implies an average electricity consumption of approximately 0.203 kWh per item processed. Detailed packaging parameters are provided in

Table 3. Data inventory for packaging, storage, and operations.

Packaging
Category	Primary packaging	Secondary packaging	Others
Packaging type	transparent PE bag and swing tag	courier bag	packing tape courier label
Weight	0.021 kg	0.035 kg	0.0075 roll/item
			0.004 kg
Operation
Category	Data-center order processing		Printer
Data	0.03 kWh/item		0.018 kWh/100 sheets
Storage
Category	F2C (i = 1)	B2C (i = 2)	BMR (i = 3)
product volume	0.006 m3
warehouse volume	4800 m3	4000 m3	498 m3
energy consumption	40 kwh/m2 * year	40 kwh/m2 * year	121.56 kwh/m2 * year

.

3.2. Analysis Results

3.2.1. Carbon Footprint Results Analysis

Based on the theoretical analysis model and underlying inventory data, the carbon footprint of “one successfully sold pair of jeans” during the sales phase can be calculated for each distribution model. This calculation is performed within the defined system boundary. As shown in Figure 3, under the factory-to-customer (F2C) model, the carbon footprint amounts to 0.408 kg CO₂e per unit. Transportation accounts for 9%, storage 5%, packaging 57%, and operations 29%. Under the classic business-to-customer (B2C) e-commerce model, the footprint increases to 0.602 kg CO₂e per item. This is composed of 13% transportation, 3% storage, 55% packaging, and 28% operation. In contrast, the brick-and-mortar retail (BMR) model yields the lowest footprint at 0.296 kg CO₂e per item, with contributions of 16% from transportation, 20% from storage, 22% from packaging, and 43% from operation. These findings challenge the common assumption that online retail is inherently more environmentally friendly. Specifically, the brick-and-mortar retail (BMR) model produces the lowest sales-phase emissions, followed by factory-to-customer (F2C) sales, while the classic business-to-customer (B2C) e-commerce model generates the highest carbon footprint. This outcome is primarily due to classic e-commerce’s reliance on secondary packaging, elevated return rates, and resulting reverse logistics. From a sustainable supply chain management perspective, the B2C channel’s centralized storage and high-frequency, small-batch delivery model significantly exacerbates its environmental disadvantages. This is particularly true when these features are coupled with the operational characteristic of high return rates. In comparison, the brick-and-mortar retail (BMR) model uses minimal packaging and experiences low return rates. The F2C model avoids intermediate storage but remains packaging-intensive.

However, these differences are mainly attributed to the significant variations in research contexts, key parameters, and system boundaries. Firstly, industry characteristics and key parameters lie at the core of these discrepancies. Unlike the high return rates characteristic of the apparel sector focused on in this study, Zheng [11] investigated fresh produce and examined fast-moving consumer goods, where return rates were not a dominant factor. In this study, the business-to-customer (B2C) e-commerce model has a substantial return rate of 21%. This significantly increases transport distances and packaging consumption through reverse logistics. As a result, the carbon footprint from transportation is nearly twice as high as that of the factory-to-customer (F2C) model (Figure 4). Secondly, the definition of system boundaries has a decisive impact on the results. For instance, Wiese et al. [13] showed that consumer travel accounted for the majority of total emissions, making any online alternative appear superior by comparison. This study excluded consumer travel to enhance comparability and focus on enterprise-controllable aspects. This approach more clearly reveals the impacts of internal logistics activities, particularly returns and exchanges. Including complete reverse logistics, the carbon advantage of online models in the high-return-rate apparel sector is substantially diminished or even reversed. Thirdly, there exists a significant difference in the carbon contribution of packaging across different channels. In online models (both F2C and B2C), packaging constitutes the largest source of carbon emissions, accounting for 57% and 55%, respectively. This is due to the need for additional courier bags, cartons, and other materials to ensure product integrity during individual shipments. In contrast, offline retail packaging (22%) focuses primarily on primary packaging, with a significantly lower carbon footprint than online models. This highlights the critical issue of over-packaging in assessing the environmental benefits of apparel e-commerce. Notably, the storage stage in the brick-and-mortar retail (BMR) model generates a carbon footprint approximately three times higher than in F2C and B2C models (Figure 5). This disparity primarily arises from differences in electricity-related emissions between retail stores and warehouses. These differences are driven by their distinct operational patterns, energy consumption structures, and operational intensity levels. Retail stores serve as customer-facing environments. They have highly concentrated energy expenditures aimed at creating a comfortable shopping experience. These include high-power HVAC systems running around the clock, intensive lighting for product displays, and numerous electronic devices such as screens and speakers. Warehouses are designed primarily for storage. They have simpler energy demands focused on basic preservation, including limited ventilation and temperature control, minimal lighting activated only when needed, and intermittent operation of material handling equipment.

3.2.2. Greenhouse-Gas Emission Source Analysis

Figure 6 illustrates the sources of greenhouse gas (GHG) emissions under online and offline sales models. In the factory-to-customer (F2C) model, the supply chain structure concentrates emissions heavily at the logistics endpoint. The courier bag is the largest source of GHG emissions during the jeans’ sales phase, accounting for 41.4% of the product’s total carbon footprint. The second-largest contributor is electricity consumption from electric tricycles used in last-mile delivery and from warehouse operations, which together represent 36.5% of the total. Transparent PE film contributes 13% of packaging emissions, and other consumables, including diesel fuel, packing tape, and courier labels, account for 9.1%. The emission profile of the B2C model closely resembles that of the F2C model but exhibits greater environmental pressure overall. The courier bag remains the primary emission source (40.3%), followed by electricity consumption (35.3%), transparent PE film packaging (12.6%), and various auxiliary materials (11.9%), all of which yield higher carbon footprints than in the F2C model. This is primarily driven by its more complex supply chain network (involving a central warehouse and regional warehouses) and the additional transportation and packaging associated with its high return rate. These findings corroborate the earlier modeling results, confirming that the return rate is a key sensitivity factor that amplifies the carbon footprint of the B2C model. In stark contrast, the BMR’s emission structure is dominated by operational energy use in physical facilities. Electricity consumption for in-store storage and operations constitutes the overwhelming majority of emissions, at 62.5%. This highlights that enhancing energy efficiency in brick-and-mortar stores is central to achieving deep decarbonization in offline channels. Key measures include adopting LED lighting and high-efficiency heating, ventilation, and air conditioning (HVAC) systems. Although offline sales involve significantly less packaging than online models, the transparent PE film still accounts for 18.9% of emissions, highlighting the importance of adopting sustainable packaging solutions even in physical retail. Transportation contributes 15.5% of emissions, primarily from 17.6 m diesel trucks. Other consumables account for a minimal share (3%). These figures reflect the inherent advantages of offline retail, such as bulk transportation and simplified packaging.

4. Uncertainty and Sensitivity Analysis

Although this study has developed a comprehensive carbon footprint model that incorporates the full reverse logistics process, including returns and exchanges, and uses reliable, representative data sources, the results still carry inherent uncertainties. These uncertainties primarily stem from model assumptions, the selection of key parameter values, and the origins of the input data.

4.1. Uncertainty Analysis

This study has several limitations that should be noted when interpreting the results. The model relies on industry average data and secondary data rather than company-specific operational data. This may lead to deviations between the model results and the actual carbon emissions of a particular company. The model assumes that all returns are sent back to a central warehouse for processing and that each order involves at most one return or exchange. In actual operations, companies often adopt regional return center strategies, and multiple returns or exchanges for a single order do occur. The logistics network model in this study is constructed based on the geographic locations of Company A’s factories, warehouses, and stores. Supply chain layouts vary across different companies. This may affect the absolute values of carbon footprints, but the relative comparison trends among the sales models are expected to be generalizable. This study focuses only on a single product category, namely jeans. Different apparel categories differ in terms of weight, volume, packaging requirements, and return and exchange rates. Therefore, generalization of the findings to other categories, such as outerwear, knitwear, and underwear, should be done with caution. This study uses the national average electricity emission factor and does not account for differences in carbon intensity across regional power grids. There are significant differences in emission factors among regional power grids in China, including those in East China, Central China, and South China. This may affect the operational emission estimates for warehouses and stores located in different regions. The model does not consider individual differences in consumer behavior, such as shopping frequency, return propensity, and travel mode choice. Such behavioral heterogeneity may affect the actual comparison of carbon emissions between online and offline channels and represents an important direction for future research.

Furthermore, the model systematically excludes consumer travel. While this exclusion aimed to reduce variability and enhance cross-channel comparability, it omits potentially significant emissions from consumers’ drives to physical stores. This omission may lead to an underestimation of the brick-and-mortar retail (BMR) model’s carbon footprint. To assess the potential impact of this exclusion on the conclusions, this subsection presents a scenario analysis incorporating consumer travel into the BMR model. Based on emission factors from the China Products Carbon Footprint Database (CPCD, 2024), assuming a one-way travel distance of 10 km, we consider two categories of travel modes: public transport (light rail, bicycle, diesel bus, electric bus) and private car (gasoline car/taxi, electric taxi) [28]. The baseline carbon footprint of the BMR model is 0.296 kg CO₂e/item (excluding consumer travel). The results are shown in

Table 4. Analysis of the Sensitivity of Consumer Travel to the BMR.

Travel Mode	Consumer Travel CF (kg CO2e/Trip)	BMR Total CF (kg CO2e/Item)	Change from Baseline (%)
Public transport
Bicycle	0	0.296	0
Diesel bus	0.150	0.446	+50.7%
Electric bus	0.400	0.696	+135.1%
Light rail	0.136	0.432	+46.0%
Private car
Gasoline car/taxi	0.410	0.706	+138.5%
Electric car/taxi	0.170	0.466	+57.4%

. Under the private car (gasoline) scenario, consumer travel generates 0.410 kg CO₂e per trip (based on an emission factor of 0.041 kg CO₂e/person·km), and the total carbon footprint of the BMR model increases to 0.706 kg CO₂e/item, an increase of 138.5%. Under this scenario, the carbon footprint of the BMR model (0.706 kg CO₂e/item) exceeds the baseline B2C value (0.602 kg CO₂e/item), implying that the relative carbon advantage of offline channels would be reversed if consumer car travel is included. Even under a relatively low-carbon travel mode such as light rail (0.0136 kg CO₂e/person·km), the total carbon footprint of the BMR model increases to 0.432 kg CO₂e/item. Although bicycle travel generates zero carbon emissions (0 kg CO₂e/km), its generalizability is limited by distance and seasonal constraints. Under the diesel bus (0.015 kg CO₂e/person·km) and electric bus (0.04 kg CO₂e/person·km) scenarios, the total carbon footprint of the BMR model increases to 0.446 kg CO₂e/item and 0.696 kg CO₂e/item, respectively.

This analysis demonstrates that the finding that “the BMR model exhibits the lowest carbon footprint” is strictly dependent on the system boundary assumption that excludes consumer travel. In practice, consumer travel mode choice has a decisive impact on the carbon footprint of offline channels. If consumers primarily travel by private car, the carbon footprint of offline channels will be significantly higher than that of online channels; if consumers travel by public transport or non-motorized modes, offline channels may still maintain a certain carbon advantage. Therefore, the conclusions of this study should be interpreted within the defined system boundary. It should also be noted that the model assumes a fixed relationship between return and exchange rates (the exchange rate is set at half the return rate) and allows at most one return or exchange event per order. These simplifying assumptions may affect the generalizability of the model to scenarios with different return-exchange patterns or higher rates of multiple returns. Future research should validate the model under such conditions to further assess its external validity.

4.2. Sensitivity Analysis

4.2.1. Return Rates’ Impact on the Carbon Footprint During the Sales Stage

The findings of this study are highly sensitive to the return rate parameter. The model employs a 21% online return rate, reflecting the industry average. In practice, however, return rates vary significantly across retailers and promotional campaigns. They can reach as high as 70% in contexts such as live-streaming e-commerce. Moreover, due to a lack of empirical data on exchange rates, this study assumed the exchange rate to be half of the return rate, which introduces potential bias. To quantify the impact of return rates, we conducted a sensitivity analysis using the business-to-customer (B2C) model as a baseline while holding all other parameters constant. The carbon footprint per pair of jeans was simulated under varying return rates (see

Table 5. Sensitivity of carbon footprint to return rate under the classic B2C e-commerce mode.

Scenario	Return Rate (%)	Carbon Footprint (kg CO2e/Item)	Change Relative to Baseline	Change Relative to the BMR Model
Excluding the return rate	0	0.415	−31.1%	+40.2%
Low-return scenario	10	0.487	−19.1%	+64.5%
Baseline scenario	21	0.602	0%	+103.4%
High-return scenario	35	0.863	+43.4%	191.6%
Extreme scenario	50	1.633	+171.3%	+451.7%

). As illustrated in Figure 7, a near-exponential relationship exists between return rates and carbon footprint. When the return rate increases from the baseline of 21% to 35%, the carbon footprint rises by 43.4%. Under an extreme scenario with a 50% return rate, the carbon footprint surges by 171.3%. This demonstrates that reverse logistics, which includes additional transportation, repackaging, and handling due to returns, imposes a significant marginal environmental burden. In an idealized scenario with zero returns, the B2C model’s carbon footprint would be only 0.415 kg CO₂e per item. Compared to the baseline scenario (0.602 kg CO₂e per item), the difference of 0.187 kg CO₂e per item is entirely attributable to the 21% return rate. This implies that, under baseline conditions, over 31% of the total carbon footprint stems solely from returns. Notably, even in the complete absence of returns, the B2C model’s carbon footprint (0.415 kg CO₂e) remains 40.2% higher than that of the BMR (0.296 kg CO₂e), primarily due to its inherent packaging and fragmented last-mile delivery structure. As the return rate escalates, the environmental disadvantage of online channels intensifies dramatically. Under the baseline scenario, the B2C model already generates more than double the emissions of the BMR model. In the extreme 50% return rate scenario, its carbon footprint exceeds that of the BMR model by a factor of 4.5. These results provide robust evidence that, given the high return rates in China’s apparel e-commerce sector, the environmental impact of online sales becomes increasingly severe as return rates rise.

4.2.2. The Impact of Transportation Distance on the Carbon Footprint of the Transportation Process

Transport distance is a key parameter affecting the carbon footprint of the transportation stage. To comprehensively assess the impact of transport distance changes across different logistics segments on carbon emissions, this study conducts sensitivity analyses from two dimensions. The first dimension is the reverse logistics endpoint, specifically the regional return center. The second dimension is the upstream trunk transport endpoint, specifically the factory location. It should be noted that this analysis uses the original model as the baseline scenario, keeping all other parameters unchanged. In the baseline scenario, this study assumes that all returns are sent back to a central warehouse for processing. This is a simplifying assumption designed to ensure comparability of return processing across the three sales models. However, in actual operations, many e-commerce companies adopt a regional return center strategy. Under this strategy, returned products are first sent to a regional warehouse close to the consumer. After quality inspection, the subsequent handling method is determined. To assess the impact of this practice on the carbon footprint, this study designs three supplementary scenarios. These scenarios assume that the distance from the consumer to the regional return center is 40%, 20%, and 10% of the originally assumed distance from the consumer to the central warehouse.

The results show that the establishment of regional return centers has a very limited impact on the carbon footprint of the transportation stage for each sales model. As shown in

Table 6. Analysis of the Sensitivity of Distribution Areas to the Carbon Footprint of the Transportation Process.

Scenario	Return Center Distance (% of Original)	Sales Model	Dynamic Weighted Average Distance (km)	Transport-Stage CF (kg CO2e/Item)
Baseline	100%	F2C	332.510	0.028
		B2C	944.377	0.080
		BMR	329.550	0.047
Scenario 1	40%	F2C	331.181	0.028
		B2C	911.037	0.077
		BMR	643.849	0.047
Scenario 2	20%	F2C	330.738	0.028
		B2C	899.924	0.076
		BMR	642.257	0.046
Scenario 3	10%	F2C	330.517	0.028
		B2C	894.367	0.075
		BMR	641.461	0.047

, the emission reduction effect for the F2C and BMR models is nearly zero. This is related to the logistics structure of these two models. The F2C model has no central warehouse as an intermediate node. In the BMR model, the processing path for store returns is already short. Thus, the marginal benefit of further optimizing the final return route is extremely low. The B2C model is the only channel that shows an observable emission reduction effect. However, the magnitude of change in the carbon footprint of the transportation stage is very limited. Under the 40% scenario, the carbon footprint value is 0.077 kg CO₂e per item, a reduction of about 3.8% compared to the baseline scenario. Even under the 10% scenario, the reduction is only about 6.3%. The main reason for this result is that trunk transport from the factory to the central warehouse accounts for an absolutely dominant share of total transport distance, exceeding 80%. The contribution of last-mile delivery and reverse logistics is extremely limited. Therefore, even significantly shortening the final return distance has a negligible marginal improvement effect on total transport emissions.

This study further examines the impact of factory location, which is the trunk transport distance, on the carbon footprint of the transportation stage. In the baseline scenario, the factory of the case company, Company A, is located in Wenzhou City, Zhejiang Province, approximately 488 km from the central warehouse. To assess the impact of changes in trunk transport distance, this study designs two supplementary scenarios. Scenario B assumes the factory is located in Songjiang District, Shanghai, about 50 km from the central warehouse. Scenario C assumes the factory is located in Baiyun District, Guangzhou, about 1200 km from the central warehouse, simulating a long-distance allocation model. The calculation results show that the factory location has a significant impact on the carbon footprint of the transportation stage. As shown in

Table 7. Analysis of the Sensitivity of Factory Location to the Carbon Footprint of the Transportation Process.

Scenario	Location	Sales Model	Dynamic Weighted Average Distance (km)	Transport-Stage CF (kg CO2e/Item)
Baseline	Wenzhou	F2C	332.510	0.028
		B2C	944.377	0.080
		BMR	329.550	0.047
Scenario 1	Shanghai	F2C	98.956	0.008
		B2C	304.961	0.026
		BMR	178.668	0.013
Scenario 2	Guangzhou	F2C	1385.514	0.118
		B2C	2328.318	0.198
		BMR	1665.793	0.121

, under the F2C model, when the factory is moved from Wenzhou to Shanghai, the dynamic weighted average transport distance decreases from 332.51 km to 98.96 km. The carbon footprint of the transportation stage decreases from 0.028 kg CO₂e per item to 0.008 kg CO₂e per item, a reduction of 71.4%. Conversely, when the factory is moved to Guangzhou, the transport distance increases to 1385.51 km. The carbon footprint of the transportation stage increases to 0.118 kg CO₂e per item, an increase of 321.4%. Under the B2C model, Scenario Shanghai has a transportation stage carbon footprint of 0.026 kg CO₂e per item, a decrease of 67.5% compared to the baseline scenario of 0.080 kg CO₂e per item. Scenario Guangzhou has a transportation stage carbon footprint of 0.198 kg CO₂e per item, an increase of 147.5%. Under the BMR model, the transportation stage carbon footprint decreases from 0.047 kg CO₂e per item to 0.013 kg CO₂e per item in Scenario Shanghai, a reduction of 72.3%. In Scenario Guangzhou, it increases to 0.121 kg CO₂e per item, an increase of 157.4%.

These results reveal a key finding. Changes in trunk transport distance have a significant amplifying effect on the carbon footprint of the transportation stage. Taking the B2C model as an example, moving the factory from Wenzhou to Shanghai shortens the trunk distance by about 295 km. This leads to a 67.5% reduction in the carbon footprint of the transportation stage. Conversely, moving the factory to Guangzhou increases the trunk distance by about 850 km. This leads to a 147.5% increase in the carbon footprint. This sensitivity is much higher than the effects observed in the previous regional return center sensitivity analysis. This comparison suggests that for apparel e-commerce companies, optimizing the upstream supply chain layout is a more effective carbon reduction strategy. Specifically, shortening the trunk transport distance between the factory and the central warehouse is more effective than optimizing the final return route.

4.2.3. Sensitivity Analysis of the Carbon Footprint of the Packaging Process Based on Packaging Weight and Packaging Type

Given that packaging accounts for the highest proportion of total emissions in online channels, specifically 57% for the F2C model and 55% for the B2C model, this study further examines the impact of packaging material weight and packaging material type on the model results. All analyses are based on a baseline scenario that uses traditional plastic packaging with a baseline packaging weight. The changes in total carbon footprint are calculated separately for the F2C, B2C, and BMR sales models.

Using the packaging weight in the baseline scenario as a reference, this study examines scenarios in which packaging weight is reduced by 20% and 30% and increased by 20% and 30%. As shown in

Table 8. Sensitivity analysis of packaging parameters on packaging process carbon footprint.

Parameter	Value	F2C	B2C	BMR
Baseline	—	0.231	0.331	0.065
Packaging weight
	−30%	0.162	0.232	0.046
	−20%	0.185	0.265	0.052
	+20%	0.277	0.397	0.078
	+30%	0.300	0.430	0.085
Packaging material
	Recycled plastic [34]	0.035	0.050	0.038
	PLA [35]	0.038	0.054	0.042
	Paper package [28]	0.106	0.152	0.139

, changes in packaging weight have a consistently positive correlation with the carbon footprint of the packaging stage across all sales models. However, the magnitude of the impact varies by model. Taking the B2C model as an example, when packaging weight is reduced by 30%, the packaging carbon footprint decreases from 0.331 kg CO₂e per item to 0.232 kg CO₂e per item, a reduction of 29.9%. When the packaging weight is increased by 30%, the packaging carbon footprint increases to 0.430 kg CO₂e per item, an increase of 29.9%. However, because the share of packaging emissions in total emissions differs across sales models, the marginal impact of packaging weight changes on total carbon footprint also varies. Under the F2C model, packaging accounts for 57% of total emissions. A 30% reduction in packaging weight reduces the packaging carbon footprint by 30%, thereby reducing total carbon footprint from 0.408 kg CO₂e per item to 0.329 kg CO₂e per item, a reduction of 19.4%. Under the B2C model, packaging accounts for 55% of total emissions. A 30% reduction in packaging weight reduces total carbon footprint from 0.602 kg CO₂e per item to 0.502 kg CO₂e per item, a reduction of 16.6%. Under the BMR model, packaging accounts for 22% of total emissions. A 30% reduction in packaging weight reduces total carbon footprint from 0.296 kg CO₂e per item to 0.277 kg CO₂e per item, a reduction of 6.4%. These results show that packaging weight reduction has a significant effect on reducing the total carbon footprint of online channels but has a relatively limited effect on offline channels. This directly reflects the differences in the share of packaging emissions across the models.

This study further examines the impact of different packaging material types on total carbon footprint. In the baseline scenario, the packaging material is conventional plastic. To assess the emission reduction potential of alternative materials, this study selects three alternative materials. These are recycled plastic, bio-based PLA, and paper packaging. As shown in Table 7, under the B2C model, replacing conventional plastic with recycled plastic reduces the packaging carbon footprint from 0.331 kg CO₂e per item to 0.050 kg CO₂e per item, a reduction of 84.9%. This reduces total carbon footprint from 0.602 kg CO₂e per item to 0.325 kg CO₂e per item, a reduction of 46.0%. Replacing conventional plastic with PLA reduces total carbon footprint to 0.328 kg CO₂e per item, a reduction of 45.5%. Replacing with paper packaging reduces the total carbon footprint to 0.451 kg CO₂e per item, a reduction of 25.1%. Under the F2C model, the emission reduction effect of material substitution is even more significant. Replacing conventional plastic with recycled plastic reduces total carbon footprint from 0.408 kg CO₂e per item to 0.202 kg CO₂e per item, a reduction of 50.5%. Replacing with PLA reduces the total carbon footprint to 0.205 kg CO₂e per item, a reduction of 49.8%. Replacing with paper packaging reduces the total carbon footprint to 0.321 kg CO₂e per item, a reduction of 21.3%. Under the BMR model, the impact of material substitution on total carbon footprint is relatively limited. Replacing conventional plastic with recycled plastic reduces total carbon footprint from 0.296 kg CO₂e per item to 0.269 kg CO₂e per item, a reduction of 9.1%. Replacing with PLA packaging reduces the total carbon footprint to 0.273 kg CO₂e per item, a reduction of 7.8%. Low-carbon substitution of packaging materials has a significant emission reduction effect for online channels, especially for the F2C and B2C models, but has a relatively limited effect for offline channels. This difference mainly arises from the different packaging usage patterns across the sales models.

It is worth noting that in practical applications, the feasibility of alternative materials also needs to consider factors such as functionality, including moisture resistance and tear resistance, supply chain availability, cost differences, and waste disposal methods. For example, although paper packaging has a lower carbon emission factor, it may face challenges related to insufficient moisture resistance in apparel e-commerce applications. Although PLA packaging is derived from renewable resources, industrial composting facilities are not yet widely available in China. If PLA packaging ends up in landfills or incineration, its carbon advantage may be weakened. Comparing the results of the above two sets of sensitivity analyses reveals differences in the emission reduction efficiency of different packaging optimization strategies. Under the B2C model, a 30% reduction in packaging weight reduces total carbon footprint by about 30%, while replacing conventional plastic with recycled plastic reduces total carbon footprint by about 85%. This comparison shows that low-carbon substitution of packaging materials has higher emission reduction efficiency than packaging weight reduction. However, from the perspective of implementation difficulty, packaging weight reduction, such as optimizing packaging dimensions and reducing fillers, is often easier to achieve in the short term than material substitution, which involves supply chain adjustments, cost changes, and functionality validation. Therefore, it is recommended that e-commerce companies first reduce packaging material consumption through design optimization and then gradually introduce low-carbon alternative materials to achieve deeper carbon emission reductions.

5. Conclusions

This study develops a comprehensive carbon footprint assessment model integrating forward and reverse logistics and applies it to the Chinese apparel e-commerce context. The system boundary focuses on enterprise-controllable sales-phase activities, including transportation, packaging, storage, and operations. Consumer travel behavior is excluded. Within this boundary, three sales models are compared: factory-to-consumer (F2C), traditional business-to-consumer (B2C) e-commerce, and brick-and-mortar retail (BMR).

The BMR model shows the lowest carbon footprint at 0.296 kg CO₂e per item, followed by F2C at 0.408 kg CO₂e per item and B2C at 0.602 kg CO₂e per item. Packaging is the main source of online emissions, accounting for 57% of F2C emissions and 55% of B2C emissions. Store operations contribute 43% of offline emissions. This conclusion depends on the defined system boundary. If consumer travel by private car were included, the relative advantage of offline channels would decrease or even reverse. The return rate is a key amplifier of carbon emissions. At the baseline return rate of 21%, return-related emissions account for more than 31% of total e-commerce emissions. When the return rate rises to 50%, the total carbon footprint of the B2C model increases by 171.3%. Transport distance sensitivity analysis shows that the trunk line distance from the factory to the warehouse has a much greater impact on the carbon footprint than last-mile return route optimization. Packaging sensitivity analysis reveals that replacing conventional plastic with recycled plastic reduces total B2C emissions by 46%. Using PLA reduces emissions by 45.5%, and using paper reduces emissions by 25.1%. A 30% reduction in packaging weight lowers total B2C emissions by 16.6%.

Based on these findings, differentiated carbon reduction strategies are proposed for different channels. For online channels, three actions are recommended. First, return management should be a core priority. Companies can reduce return rates at the source through AI-powered virtual fitting, standardized sizing systems, and enriched product pages. The sensitivity analysis shows that for every one %age point reduction in the return rate, carbon emissions under the B2C model decrease by approximately 0.01 kg CO₂e per item. Second, packaging innovation is the most effective lever. First, a 10 to 15% reduction in packaging material can be achieved by optimizing packaging dimensions and reducing fillers. Subsequently, the introduction of low-carbon materials such as recycled plastic or bio-based PLA can achieve a deeper emission reduction of more than 45%. Companies should first optimize package size and reduce fillers, then introduce low-carbon materials such as recycled or reusable alternatives to achieve deep emission reductions. Third, logistics network optimization should take priority over last-mile optimization. Companies should move inventory forward to regional warehouses close to major consumer markets. For offline channels, improving store energy efficiency is the top priority. Companies should upgrade to LED lighting, adopt variable-frequency air conditioning systems with top energy efficiency ratings, and establish energy use intensity benchmarks measured in kilowatt-hours per square meter. Encouraging consumers to use public transport or non-motorized travel can also significantly reduce the overall carbon footprint of offline shopping. For policymakers, governments are encouraged to use tax incentives to support e-commerce companies in establishing regional return processing centers in major consumer cities. Metrics such as express packaging reduction and the share of reusable packaging should be integrated into green evaluation systems for e-commerce platforms. For consumers, companies can raise environmental awareness through carbon information transparency. This can be done by showing the estimated carbon emissions of an order and the additional carbon cost of a return on the order page. Companies may also offer green point rewards to consumers with low return rates.

In summary, by integrating sustainable supply chain management and life cycle assessment, this study not only quantifies and compares the carbon footprints of apparel sales channels but also reveals the underlying structural and operational drivers. The proposed channel-specific strategies provide decision support with both theoretical depth and practical relevance, offering evidence-based guidance for apparel companies to develop precise, low-carbon transition pathways. It should be reiterated that these findings are derived within the defined system boundary, which focuses on enterprise-controllable sales-phase activities and excludes consumer travel. Therefore, caution should be exercised when generalizing these results to real-world shopping scenarios that include consumer travel behavior. Future research should incorporate consumer travel behavior, cover more product categories and regions, collect primary company-level operational data, and refine the model using regional return center assumptions.