How perceived sustainability influences consumers’ clothing preferences

Sustainable fashion is characterized by its basis on environmental, social, and economic impacts, achieved through the adoption of environmentally friendly production and processing methods, as well as the use of renewable, recyclable, or biodegradable materials¹. This type of clothing aims to reduce dependence on natural resources, minimize negative environmental impacts, and promote social responsibility. Common sustainable fashion materials include organic cotton, recycled polyester, linen, and wool. In today’s fashion industry, sustainability has become a key driving force for development, guiding businesses and consumers in exploring new paths towards sustainable fashion².With increasing environmental awareness, consumers are paying more attention to the sustainability of products, sparking reflections within the apparel industry^3,4. The rise of sustainable fashion signifies a shift in the fashion industry’s development, transcending mere aesthetics and style to embrace responsibilities towards the environment, social welfare, and sustainability. Sustainable fashion is no longer just a trend but an inevitable direction for future development. Consumers are becoming increasingly aware of the environmental impact of their purchasing choices, thus demanding higher sustainability standards for products⁵.

In manufacturing sustainable fashion, product design improvement is one of the primary steps towards achieving sustainability goals. Product improvement design involves adjustments, optimizations, or innovations to enhance existing products’ performance, functionality, quality, user experience, or other features. This design approach aims to align products more closely with market demands, meet consumer expectations, and adapt to rapidly changing market environments⁶.

As the global fashion industry endeavors to address the necessities of sustainable development, understanding the dynamics of consumer behavior becomes increasingly crucial. According to data from Kings Research⁷, the global sustainable fashion market was valued at $70.706 billion in 2023 and is projected to reach $135.139 billion by 2031, with a compound annual growth rate (CAGR) of 8.58% from 2024 to 2031. Real-time data and forecasts highlight shifts in consumer behavior and market dynamics, reflecting the fashion industry’s growing commitment to sustainability. As consumer environmental awareness strengthens, the demand for apparel with minimal ecological impact continues to rise, influencing emerging and established markets globally.

However, looking at detailed regional data, in 2023, the European sustainable fashion market accounted for approximately 36.09% of the global market share, worth $25.518 billion. Europe’s dominant position in the sustainable fashion sector is primarily attributed to its consumers’ prioritization of eco-friendly products. Meanwhile, the Asia-Pacific region is expected to grow significantly during the forecast period, with a CAGR of 10.77%. Due to various factors, including the large and growing populations of countries like India and China, the Asia-Pacific is anticipated to dominate the sustainable fashion market throughout the forecast period (2024–2031). Current consumer awareness of environmental issues in the Asia-Pacific region could be much higher. Still, as one of the world’s largest fashion markets, China’s consumer behavior and preferences influence regional trends and impact global fashion brands and supply chains. Therefore, understanding the factors influencing Chinese consumers’ decisions to purchase sustainable apparel and analyzing how to integrate these into their buying habits is critical to advancing the global sustainable fashion industry.

Online reviews, as a form of consumer feedback, provide researchers with avenues to explore public perceptions and behaviors. These reviews reflect consumers’ direct experiences with products and reveal their views on sustainability issues. For example, Zhao et al.⁸ predicted overall customer satisfaction by analyzing online text reviews of hotels, demonstrating the role of review analysis in understanding consumer behavior. Sun et al.⁹ explored the different impacts of online reviews on searching for and experiencing products, with consumer reviews proven to be an effective method for understanding market trends and consumer needs. However, advanced data analysis techniques are required to extract useful information due to the large volume of reviews and their typically unstructured nature. In the Internet era, consumers share their experiences and opinions on products and services through online reviews. These reviews not only influence the purchasing decisions of other consumers but also provide valuable market insights for businesses. However, due to the large volume and diverse structure of reviews, especially analyzing them, mainly extracting helpful information, has become an important research area¹⁰. As an intuitive feedback mechanism, online reviews demonstrate outstanding potential in measuring customer satisfaction. The linguistic characteristics of online reviews can effectively predict customer satisfaction. The linguistic features of reviews are considered a direct reflection of customer experience and a key indicator of measuring customer satisfaction¹¹. In this process, the language used in reviews, emotional tones, and level of detail are meticulously analyzed to explore their correlation with customer satisfaction¹². Through in-depth analysis of consumer feedback, businesses can better understand customer needs and preferences, optimize products and services, and enhance customer satisfaction. Thus, this research also provides valuable insights into using online reviews to improve business decisions and marketing strategies.

At present, due to unsustainable production and consumption practices, the fashion industry is one of the leading contributors to global environmental degradation. The traditional “take-make-waste” fashion model generates pollution and landfill overload at every step of its process, accelerating the ecological crisis. With the fast expansion of the fashion market and increasing consumer awareness, there is an urgent need to adopt sustainable practices to reduce the fashion industry’s environmental footprint significantly. This study aims to facilitate the transition to sustainable fashion practices, which are crucial not only for environmental sustainability but also for the long-term viability of the fashion industry itself.

Therefore, this study focuses on assessing the current sustainable practices in the Chinese fashion industry through data from China’s largest e-commerce platform, exploring innovative solutions, and developing feasible strategies for stakeholders. Fundamental questions drive this research: What are the current sustainable practices and their effectiveness? What innovative practices can be implemented to enhance sustainability? And what strategies can industry stakeholders adopt to promote sustainable practices? This study aims to leverage direct feedback from online communities to gain insights into the current state of sustainable fashion design and propose innovative product optimization strategies. We extensively explore consumer perceptions of sustainable fashion, thoroughly analyze their evaluations of existing products, and identify viable directions for improvement. This comprehensive approach addresses immediate environmental challenges and contributes to the long-term survival of the global fashion industry.

The association between sustainable fashion and online reviews

In contemporary society, consumers recognize sustainable fashion’s significant role in the broader environmental movement, acknowledging its ecological benefits. Moreover, there is a growing awareness of the importance and value of consuming sustainable apparel. Research indicates that environmental awareness is strengthening. Consequently, more consumers are opting for sustainable fashion choices^13,14. Innovative practices in sustainable fashion design encompass various aspects. Fatimah and Kim¹⁵ highlight the enhancement of these designs by manipulating fabrics and creating intriguing surface designs to minimize environmental impact.

On the other hand, designers also focus on creating visually appealing garments sustainably, rejuvenating discarded clothing through artistic and technical means. In discussing sustainable fashion development, attention is given to the importance of sustainability within the fashion industry and how it influences product design and consumer behavior. Sustainable fashion typically refers to products that consider environmental protection, social justice, and economic feasibility throughout the design, manufacturing, distribution, and consumption processes. This form of fashion underscores the importance of environmental protection, social responsibility, and economic development, calling for a shift in the fashion industry from traditional production and consumption patterns to practices that prioritize sustainability¹⁶. Key points in this transformation include the use of recyclable materials, waste reduction, and increased transparency in the production process¹⁷.

Consequently, consumer acceptance and purchasing behavior regarding sustainable apparel are influenced. Analyzing consumer motivations for choosing sustainable fashion products, such as environmental awareness, social impact, or personal health considerations, is crucial for understanding market trends¹⁸. Ikram¹⁹ suggests that the development of new technologies not only supports an environmentally friendly consumer experience but also revolutionizes the fashion industry’s environmental impact by significantly reducing waste through improvements in materials and product sustainability. Vishwakarma et al.²⁰ explore the barriers to sustainable practices in the textile and apparel industry. They emphasize the need to overcome waste management challenges and adopt modern technologies to enhance industry efficiency. Furthermore, their research highlights the importance of technological innovation and strategic focus on consumer behavior, suggesting that the fashion industry can move towards a more environmentally friendly direction. This is crucial for understanding the environmental impacts of fashion production and consumption and their solutions. Camacho-Otero et al.²¹ delve into user acceptance and adoption of sustainable products, particularly their response to sustainable features and designs. Additionally, there needs to be more unified standards in terms and contexts within sustainability research, rendering the concept of sustainable fashion complex and multifaceted. Understanding consumer values and behavioral motivations becomes critical as sustainability evolves within the fashion industry. This study aims to precisely grasp the core elements of sustainable apparel product design and reveal critical factors influencing consumer choices, such as environmental consciousness, social responsibility, personal values, and practical considerations. Through thorough and meticulous research, we aim to provide more straightforward and practical guidance for the development of sustainable fashion. Online reviews play a significant role in the realm of sustainability. Abbate et al.²² conducted a systematic literature review highlighting the critical role of circular economy initiatives in implementing sustainability within the textile, apparel, and fashion industries. Their study reveals that corporate social responsibility and changes in consumer behavior are critical drivers in advancing sustainable fashion. Kawaf and Istanbulluoglu²³ reveal how online reviews influence consumer purchasing decisions.

Moreover, research by Bigne et al.,²⁴; Ventre and Kolbe²⁵ demonstrates that positive online reviews significantly enhance consumer trust in products or brands, increasing their willingness to purchase. Remarkably, the role of emotional expression in reviews profoundly impacts consumer purchasing decisions. This study also focuses on utilizing deep learning to mine consumer reviews.Zhang et al.²⁶ illustrate how deep learning is used to extract innovative ideas and product improvement suggestions from consumer reviews, emphasizing the value of deep learning in uncovering insights within large-scale review data. Therefore, mining and analyzing online reviews can better understand how consumers influence sustainable apparel products’ design and market trends, revealing differences in consumer behavior concerning environmental issues to better meet consumer needs in new product development.

This section comprehensively summarizes recent studies that have contributed to understanding sustainable fashion and its interaction with online consumer behavior. The Table 1 below integrates critical research findings, each offering unique insights into different aspects of sustainable development in the fashion industry. These studies have been carefully selected to illustrate the diversity of methods and perspectives in this field, highlighting how digital marketing, consumer psychology, and environmental attitudes influence consumer choices and brand strategies. This compilation helps identify current trends and methodologies and emphasizes the key factors driving global consumer acceptance and adoption of sustainable fashion practices. The purpose of this table is to systematically present how various researchers approach the topic of sustainable fashion, identifying their focus areas, data sources, theoretical frameworks, and principal conclusions. This enables us to better understand the current state and emerging trends in sustainable fashion research.

Development and application of latent Dirichlet allocation

LDA (Latent Dirichlet Allocation) originated from the field of information retrieval, aiming to address the challenge of extracting meaningful patterns from large volumes of text. Despite the initial success of early models like Latent Semantic Analysis (LSA), they needed help capturing the latent structure and topics in textual data. The exploration of these early models laid the foundation for developing LDA. Since its proposal by David in 2003, LDA has made significant strides in text mining and natural language processing fields. LDA has become critical in analyzing and understanding vast online consumer reviews. In social media analysis, LDA is used to uncover topics and trends in user comments, tweets, and social media posts²⁷. With the rapid advancement of machine learning and deep learning technologies, integrating LDA with these advanced techniques has opened up new research directions. For example, research combining LDA with neural networks has shown great potential in text classification and sentiment analysis²⁸. This integration not only expands the application scope of LDA but also enhances its capability to handle complex textual data. Therefore, combining LDA and deep learning technologies offers new possibilities and approaches in natural language processing. Here are detailed descriptions of four analytical methods in the LDA model:

1. 1.

   Document similarity analysis Using the LDA model, document similarity can be evaluated by comparing the topic distributions of documents. This analysis is valuable in information retrieval, document recommendation systems, and automatic classification. Relevant articles or research papers can be recommended based on topic similarity. Document similarity can be determined by calculating the similarity of topic distributions, such as cosine similarity or Jaccard similarity. Similar documents exhibit higher overlap in topic distributions²⁹.

2. 2.

   Trend analysis and time series LDA can be applied to time series data, such as continuous news reports or social media posts, to identify and track the temporal evolution of topics. This analysis is valuable for understanding changes in public opinion, tracking the development of news events, or analyzing social media trends. By applying the LDA model at different time points, the popularity of specific topics over time can be observed, revealing changes in trends and patterns³⁰.

3. 3.

   Sentiment analysis integration Although LDA does not directly conduct sentiment analysis, it can be combined with sentiment analysis techniques to identify topics associated with specific sentiments³¹. This approach is helpful in brand monitoring, market research, and social media analysis, helping businesses understand consumer emotional responses to products or services. Initially, LDA identifies topics in the text, and then sentiment analysis tools are applied to assess the sentiment tendencies associated with these topics, such as positive, negative, or neutral sentiments³².

4. 4.

   Topic discovery and allocation LDA identifies latent topics by analyzing document collections and assigns these topics to each document. In this process, each document is generated randomly from a set of latent topics, while each topic is composed of a specific set of vocabulary. This method is widely used in text mining, such as news classification and academic literature analysis, aiding in automatically identifying and labeling the topics of documents³³.

The LDA model has undergone several optimizations and extensions. Zhou et al.³⁴ proposed an optimization method for Spark news text topic clustering based on the TF-IDF algorithm. Kandukuri and Haragopal³⁵ emphasized topic repetitiveness using topic modeling techniques and identifying key ideas from popular monthly radio programs like Mann Ki Baat. Li et al.³⁶ introduced an SDMM (Statistical Dirichlet Multinomial model) keyword extraction model to extract keywords from short-term financial comments. Gupta et al.³⁷ proposed an LDA-based topic modeling approach that merged COVID-19 case data and news articles into a generic LDA to obtain new features. In recent years, text mining, machine learning, and sentiment analysis have made significant advancements, enabling researchers to extract valuable information more timely and accurately from social platforms. For instance, Yadav et al.³⁸ first introduced topic modeling methods to text mining in sustainable fashion literature. The study conducted an exhaustive literature review using the Scopus database. It analyzed 658 articles using text mining and topic modeling (LDA) techniques to identify key trends and themes from manufacturing to marketing sustainably. The application of LDA was extended to analyze sustainable fashion trends, identify consumer behaviors, and formulate strategies for relevant businesses. This aligns with our research objective of using LDA for a more comprehensive and objective data analysis, thereby supporting product improvements and business decisions within sustainable development. By categorizing thematic content into dimensions that reflect user needs and potential improvements, our study contributes to the technical literature on topic modeling. It provides actionable insights for evolving industry practices, particularly in integrating sustainability with technological innovation.

Therefore, this study uses the LDA model to categorize topic content into corresponding dimensions based on user needs and improvement directions. It aims to provide a more comprehensive, objective design interpretation and support for evolutionary paths and visualization. This research aims to expand the applications in data mining and product design improvement.

Data source

The core objective of this study is to explore innovative approaches to sustainable fashion design by systematically analyzing consumer online reviews. This aims to understand consumer attitudes towards sustainable fashion and gather their expectations and suggestions for product design. Through qualitative and quantitative methods, the study seeks to reveal directions for improving sustainable clothing design and develop effective design strategies. Therefore, user review data from consumers of sustainable clothing were collected from the JD.com website. These reviews provide insights into consumer feedback and perspectives on sustainable clothing. Subsequently, the Latent Dirichlet Allocation (LDA) method was employed to analyze the textual comments. The methodology used in this study combines user requirements with data analysis, enabling businesses to comprehensively understand consumer perceptions of sustainable clothing and develop effective design strategies, providing strong support for developing the sustainable clothing industry. Additionally, we collected data was from one of China’s largest e-commerce platforms (jd.com), which has a large consumer base. Therefore, this platform’s vast amount of user reviews can offer rich information to analyze consumer attitudes and demands for sustainable clothing design.

Data was collected on October 24, 2023. The search term used was “sustainable clothing,” the platform collected the top 12 products with the highest number of comments. Consumer reviews on JD.com are publicly visible to all users, providing consumers with product details for reference. The comments section also allows consumers to communicate with each other and exchange product information. Therefore, the data used in this study comes from a legitimate source.

Data processing

The data processing workflow is illustrated in Fig. 1. First, this study utilized Python to crawl the comments of JD.com’s top twelve sustainable clothing products on JD.com. There were a total of 23,544 initial data entries. The raw data included user IDs, user names, product review content, review timestamps, etc. To protect user privacy, only the review content was used for subsequent research analysis.

Methodology flow.

Full size image

During the cleaning process, 67.16% of the comments were deleted. After preliminary review, comments were removed based on the following criteria: blank entries, repeated automated comments such as “This user thought it was perfect” and other non-user comments, and content that was too vague, such as single-word replies or non-specific comments like “okay,” “so-so,” or “not bad.” Specifically, 5.57% of the comments were deleted due to being blank, 33.83% were repetitive automated comments, and 27.76% were too vague for meaningful analysis. Consequently, the effective comments amounted to 32.84% of the initial data, reducing the total to 7,730 comments. This significant reduction in data volume due to our stringent cleaning process is essential for enhancing the quality and relevance of the analyzed comments. Such extensive cleaning is crucial for improving the model’s accuracy, ensuring the remaining data’s relevance, and providing meaningful insights into consumer behavior.

Then, the complete text information needed to be segmented into keywords through tokenization for topic modeling. This study used the HanLP library for text segmentation. HanLP is a third-party Chinese word segmentation tool library that accurately calculates the probability of association between Chinese characters using its unique basic Chinese vocabulary library. HanLP can intelligently form word combinations based on these high association probabilities to obtain accurate tokenization results³⁹.

Next, we used the LDA model to identify topics in the text. This helped us understand the concepts and trends in the textual data. Further topic clustering was implemented by profoundly analyzing the topic distribution of documents and extracting relevant topic features. In this process, the importance of each topic’s keywords was determined based on Term Frequency-Inverse Document Frequency (TF-IDF) weighted data. This step ensured an accurate understanding of the true intent of user comments, avoiding errors based on single words. In this step, we determined the appropriate number of topics and selected appropriate model parameters to ensure interpretable and meaningful topics. Subsequently, we obtained three topics and interpreted and analyzed the meanings implied by the keywords in the topics through existing literature in order to determine the dimensions of the topics.

To validate the results of the topic analysis, we consulted ___domain experts or professionals in related industries to obtain professional opinions and suggestions on sustainable clothing design. We adopted the Delphi method, an expert survey technique to collect and integrate expert opinions and judgments on a particular issue or topic⁴⁰. Through the subjective judgments of experts, we obtained relatively objective information, opinions, and insights. A user-demand-based sustainable clothing design strategy was developed by interpreting the topics, comparing product features, and exploring user needs.

Finally, we conducted a questionnaire survey to assess whether there were significant differences in perceived quality, perceived value, purchasing behavior, and purchasing intention and to validate the effectiveness of the sustainable clothing design strategy. We used Multivariate Analysis of Variance (MANOVA) to analyze the questionnaire data to obtain objective statistical results to validate the effectiveness of the sustainable clothing design strategy. Through these methods, we could comprehensively understand consumer perceptions of sustainable clothing and propose effective design strategies based on data and expert opinions, providing strong support for developing the sustainable clothing industry.

Determination of the number of topics

Perplexity, coherence score, and PyLDAvis are commonly used tools in determining the optimal number of topics in LDA. Perplexity is a metric for evaluating the performance of probability models, particularly language models and topic models like LDA. It measures the predictive capability of the model on new data (test set). In the context of the LDA model, a lower perplexity indicates a more substantial predictive capability of the model on unseen documents, typically implying better model quality⁴¹. Previous studies have explored the impact of perplexity on determining the number of topics in the LDA model, and the results of this research have been validated and confirmed in various scenarios, demonstrating high reliability and effectiveness⁴². Perplexity is calculated based on the likelihood function of the model. In the LDA model, perplexity is usually calculated as follows: (1)

M is the number of texts. Detest is the document collection. Nd: represents the length of the d-th text, i.e., the number of words in the text. p(wd) is the probability of the term wd occurring given the document d.

Previous research has indicated the central role of perplexity in determining the optimal number of topics in LDA models and has empirically validated its effectiveness across various contexts⁴³. However, while perplexity primarily focuses on the model’s performance in language modeling tasks, coherence scores are more concerned with the overall coherence and logic of the generated text. These metrics are typically used with other metrics to comprehensively evaluate the model’s performance⁴¹. The formula for the coherence score is shown in (2–4).

W: a set of the top N most likely words for a topic, W = {W1…WN}. Si: Each word W’ ∈ W is paired with all other words W* ∈ W. S: the set of all defined pairs, such as S = {(W’, W*) | W’ = {wi}; ∀wi ∈ W; W* = W}. P(wi): probability of a single word. P(wi, wj): joint probability of two words. ϵ: used to interpret the logarithms of zero and γ to impose more weight on higher NPMI values.

PyLDAvis is a Python library for visualizing topic models, providing an intuitive way to explore and interpret the results of topic modeling. It combines LDA topic modeling with interactive data visualization techniques. With PyLDAvis, an LDA model can be fitted to text data, and then the results of the model can be explored using an interactive visualization tool. Research by Gottfried et al.⁴⁴ utilized LDA topic modeling to extract emerging topics from OGD and used PyLDAvis to visualize the topics for interpretation and business intelligence purposes. PyLDAvis offers a dynamic, interactive visualization interface that allows users to intuitively understand relationships between topics, the distribution of words within topics, and the key words for each topic^45,46. The use of PyLDAvis adds a visual dimension to the model evaluation process, making it easier to understand the relationships between different topics and the terms that comprise them. Perplexity, coherence score, and PyLDAvis each play important roles in LDA topic modeling. Perplexity serves as a statistical measure of the model’s predictive ability, providing a quantitative indicator of model quality. Coherence score focuses on the interpretability of topics, making topics more human-understandable. PyLDAvis, on the other hand, allows us to visually inspect topics and their relationships. By integrating these tools, this study aims to enhance the interpretability of the LDA model while ensuring its accuracy, thereby improving the overall quality of topic modeling.

Topic modeling and visualization of user comments

We collected a total of 23,544 comments. After removing preset content, blank comments, and comments containing only one word, 7,730 comments were used for text analysis. Descriptive statistical analysis was conducted on the review sample. After tokenizing the sample, we removed irrelevant characters, stopwords, links, and special symbols, focusing on crucial nouns related to clothing to ensure consistency of vocabulary and reduce noise in the topic model. The average length of comments in the sample was approximately 53.5 characters. The most common comment length (mode) was 27 characters. The total number of comments was 7,730. The standard deviation was approximately 42.3 characters. The significant variation in review lengths could be attributed to JD.com’s method of collecting consumer feedback after product use to facilitate evaluations. After purchase, consumers are typically invited to review and rate specific aspects such as usage experience, product quality, merchant service, and delivery service. Based on this review model, shorter reviews tend to provide less description of the product, while longer comments focus more on the value and attributes of sustainable fashion. It is important to note that the length of a comment does not directly indicate a positive or negative review. Reviews exceeding 100 characters are usually based on detailed evaluations of each attribute so that individual sentences can contain positive and negative feedback, such as, “I am satisfied with the quality, but not with the price.”

To establish an appropriate number of topics, this study manually adjusted the number of topics and used two indicators: model perplexity and coherence score. Using the pyLDAvis toolkit for visualization analysis, we evaluated LDA models with topics ranging from 1 to 10. The results showed that the perplexity values of the text ranged from 65 to 80, increasing with the number of topics, indicating that too many topics may decrease the clarity of the model. According to the perplexity data, it was understood that the model has 70 possible choices for each word when processing a segment of text. Whether this is considered “good” depends on a comparison with other models or results from the same task. In context, these perplexity values are within the expected range for the text data. As Ma et al.⁴⁷ indicate, whether it is “good” also depends on comparisons with other models or results from similar tasks in complex language processing tasks. Through such assessments, we can better understand the impact of model configurations on predictive capabilities and adjust the number of topics accordingly to optimize the model’s overall performance.

The coherence scores fluctuate between 0.4 and 0.6, reflecting the semantic similarity between high-scoring words within each topic. Scores within this range are generally considered moderate, suggesting that while the identified topics are meaningfully related, there is room for improvement regarding topic clarity and distinctiveness. In our preliminary evaluation, as shown in Chart 2, we followed the principle of minimizing perplexity and maximizing coherence scores and chose a relatively lower and higher coherence score. After carefully considering model complexity and interpretability, we decided to select three topics definitively. This choice represents an optimal balance where topics remain distinct and comprehensible without overfitting noise in the data. The evolution of perplexity and coherence scores for the LDA model is visualized in Fig. 2. Additionally, visual feedback from pyLDAvis further supported this decision, as it showed clearer topic clusters when using three topics, which helps intuitively understand the underlying topic structure, as shown in Fig. 3.

Confusion values and coherence scores for sustainable clothing discussions. Note: Perplexity measures the accuracy with which a probability model predicts a sample (the lower the score, the better), indicating the model’s simplicity and generalization ability. The coherence score assesses the degree to which kay words in each topic coherently convey meaningful themes (the higher the score, the better), reflecting the model’s interpretive usefulness.

Full size image

The pyLDAvis results of the sustainable clothing discussion.

Full size image

Topic model and labels

This study employs LDA to conduct topic modeling on online comments related to clothing products. The topic modeling results and the proportion of keyword weights for the top 10 keywords in each topic are presented below.

Table 2 displays the main constituent words for each topic. Each topic is sorted based on TF-IDF values, showing the top 10 words ranked by importance. Notably, certain words appear in multiple topics, indicating their significant contribution to each topic. Considering previous studies that used Latent Dirichlet Allocation (LDA) topic modeling where the same word appeared in multiple topics, it suggests that topics may influence each other. This indicates a good correlation between topics and provides insights into the importance of these terms. According to our analysis, six words appear simultaneously in the top ten list of topics: “quality” appears in topics 1 and 2. “fit” appears in topics 1 and 3. “comfortable” appears in topics 1 and 3. “size”: appears in topics 1 and 3. “material”: appears in topics 1 and 2. “softness”: appears in topics 1 and 3.

As indicated above, we have noted that six key terms—‘quality,’ ‘fit,’ ‘comfort,’ ‘size,’ ‘material,’ and ‘softness’—appear across multiple topics, highlighting their multifaceted importance. To better understand the significance of these terms in various discussions, we calculated the percentage of comments in which each term appeared. The term “quality” was mentioned in 72.33% of the comments, reflecting its primary importance in discussions related to consumers’ perceptions of sustainable fashion. Similarly, “fit” appeared in 20.74% of comments, where discussions typically focus on body fit and garment comfort. “Comfort” was mentioned in 14.13% of comments, again emphasizing its cross-topic importance, particularly in conversations about expressing subjective consumer feelings. “Size” appeared in 25.12% of comments, highlighting its relevance in garment fit and personal comfort discussions. “Softness” was mentioned in 14.21% of comments, indicating its significant but more focused role in specific aspects of product discussions. “Material” appeared in 5.31% of comments, further demonstrating its specific but essential role in discussions related to garment attributes. These percentages underscore the influence of these terms in shaping the dimensions within the topic model, with “quality” being an essential attribute of clothing consumers discuss. The extent to which these keywords appear across different topics proves their relevance and indicates topic overlap, suggesting that these terms play a crucial role in shaping consumer dialogue in a sustainable fashion.

Analysis of text topics

Topic 1: “Perceived Quality”

When analyzing the distribution of content in Topic 2, it is crucial to consider the TF-IDF of each keyword within the discussed topic. The keyword “quality” in Topic 1 has the highest TF-IDF value at 0.112, emphasizing its role as the primary descriptor for topic identification. This high TF-IDF value underscores the critical nature of measurable standards in assessing apparel attributes. Following “quality,” the terms “warmth,” with a TF-IDF score of 0.070, and “fit,” with a score of 0.040, also show significant representativeness, highlighting their relevance to consumers’ perceptions of comfort and fit, which are crucial for perceived quality. Lower-weighted terms like “softness” (0.029), “comfort” (0.023), and “size” (0.023) contribute to the subjective dialogue about factors influencing consumer satisfaction and preferences. These keywords are weighted to represent their frequency and relevance, illustrating a layered understanding of quality.

The actual result of quality consists of 2 dimensions, i.e. objective quality which refers to measurable and verifiable aspects of the garment according to predetermined quality standards, and perceived quality where the consumer’s subjective judgement of the product quality is influenced by the individual situation and personal assessment. Objective quality refers to product attributes that can be directly measured and verified by scientific methods according to predefined quality standards. Perceived quality, on the other hand, is based on the consumer’s personal situation and assessment, and is a subjective judgement that involves factors such as material, comfort, style, etc., reflecting the consumer’s overall perception of product quality. The boundary that distinguishes the two is that objective quality can be assessed quantitatively, whereas perceived quality is a combination of personal experiences and preferences⁴⁸.

We found that the key words in Theme 1 are closely related to the two dimensions of quality. The keyword “quality” in Theme 1 explains objective quality as a core term due to the highest distribution rate in the theme. As a physical characteristic of clothing, “quality” can be quantitatively assessed by durability and manufacturing process. The term “material”, “thickness” and “detail” thus explains in more detail a range of physical properties of garments that can be measured quantitatively in objective quality. Examples include colour fastness, breaking strength and elongation, stretch and recovery of fabrics containing elastic fibres, seam slippage, pilling resistance and dimensional stability⁴⁹. “soft” “fit” “comfort” “size” “warmth “Style” appears in the vocabulary, which is influenced by factors such as expectations, experience and personal preferences, while the perception of the quality of clothing is a subjective judgement made by the consumer, which may be different from the actual objective quality of the product, so we consider this part of the psychological and physiological perception of the clothing, which is part of the perceived quality. It belongs to perceived quality. For example, fabric construction, composition and surface treatment also play a role in the sensory and thermal comfort of a garment. Parameters such as fabric drape, coefficient of friction, heat and evaporation resistance affect the comfort of a garment. Thermoregulation is a key comfort factor that should be considered in the design of garments, with ‘warmth’ performance becoming a key consideration. Similarly, for products that come into frequent contact with the skin, ‘soft’ properties ensure greater comfort and a better user experience⁵⁰). The correct ‘size’ is not only related to comfort, but sizing is a sign of attention to detail and has a direct impact on user satisfaction⁵¹. The overview of style, on the other hand, reflects a consumer’s expectation of the style and design of the garment, and attention to style is crucial in fashion analysis, as it provides a more nuanced understanding of clothing trends and preferences⁵². It is important to note that the themes ‘comfort’ and ‘fit’ are intuitively communicated by consumers to express satisfaction with quality attributes and do not appear as essential attributes of clothing quality.

Topic 2: “Perceived Value”

While exploring Topic 2, it is evident that the keyword “quality” holds a TF-IDF score of 0.140, making it the core vocabulary of Topic 2. Although it is the highest core keyword and overlaps with Topic 1, its meaning within this theme is redefined by other keywords in the topic. The sequential TF-IDF scores of “price,” “cheap,” and “discount” underscore the dominant presence of these weighty keywords, highlighting the distinction of Topic 2’s essence from Topic 1. The distribution of keyword weights explains that overlapping core keywords do not necessarily indicate overlapping dimensions.

Perceived value is consumers’ subjective evaluation of the value or benefits they expect from a product or service. It is a multidimensional process of overall consumer evaluation⁵³. In clothing, consumers’ perceived value evaluation indicators consist of five dimensions. However, the perceived value of clothing typically covers multiple dimensions, generally summarized as functional value (such as durability, practicality), emotional value (such as personal preference, aesthetics), social value (such as brand image, social recognition), cognitive value (such as price, cost-effectiveness), and situational value (value in specific situations, such as wearing on special occasions). These dimensions collectively determine consumers’ overall perception of the value of clothing⁵⁴. Notably, although the keyword “quality” has the highest distribution probability in Topic 2, the keywords related to price predominate in this topic, indicating that the quality factor describes price. Functional value can be defined as consumers’ consideration of high- and low-quality factors in the fluctuation of prices when making purchases⁵⁵. One of the factors that influences functional value is consumers’ consideration of “quality” when making purchases. Cognitive value refers to consumers’ subjective attitudes after making judgments. In the market environment, consumers can intuitively identify the most favorable price, which is undoubtedly one of the most critical factors. Since price is the first factor consumers pay attention to, rather than product quality or model, low price becomes one of the best choices for consumers. Therefore, in selling products, price becomes one of the essential variables determining purchasing behavior⁵⁶.

Emotional value can be achieved through various sensory experiences of consumers, such as visual, olfactory, and tactile experiences. This refers to clothing-related factors, such as color, fabric, smell, and touch⁵⁷. Social value refers to how consumers can create an identity through clothing. Clothing can also enhance existing identities, making it essential to establish professional status. Clothing is crucial in establishing professional status and identity⁵⁸. Situational value is the subjective clothing awareness of specific groups of people. Consumers will choose their local cultural clothing for activities on specific holidays and special occasions⁵⁹. Based on the results of clothing topics, we conclude that the word “quality” in the topic corresponds to functional value. “Price,” “cost-effectiveness,” “cheap,” and “discount” may be related to cognitive value.

Additionally, “color” is associated with emotional value. It is noteworthy that words such as “expectation,” “feature,” “material,” and “color fading” seem to have influences from two to multiple dimensions. Therefore, the vocabulary in Topic 2 refers to the concepts of functional value, cognitive value, and emotional value.

Topic 3: “Sensory Comfort”

In our exploration of Topic 3, we quantify the importance of specific terms using their TF-IDF scores to illustrate their relevance in the discussion context. The term “fabric” scores a TF-IDF of 0.146, underscoring its significant role within the topic model. This high score reflects the term’s central position in this theme, especially in discussions that help specifically define the dimensions of Topic 3. Similarly, other terms like “breathability” (0.088) and “size” (0.079) also exhibit TF-IDF values, highlighting their importance in further clarifying the dimensions.

The dimensions of sensory comfort typically include thermal comfort (such as physical comfort), aesthetic comfort (such as confidence in wearing and aesthetic satisfaction), and tactile comfort (such as the suitability of clothing and freedom of movement). These three dimensions interact, collectively influencing individuals’ overall perception of clothing comfort⁶⁰. Our research results in Topic 3 indicate that the vocabulary is closely related to the definition of sensory comfort.

Physical comfort usually involves clothing attributes that directly contact the human body, such as temperature regulation, breathability, humidity control, and fabric softness. These factors directly affect consumers’ bodily sensations when wearing clothes, which is crucial for enhancing individual comfort and overall wearing experience⁶¹. The breathability of clothing is related to its thermal comfort. Thermal protective performance indicates that with increased recycled cotton fiber content in clothing, fabrics’ conductivity, and radiation resistance also increase⁶². In Topic 3, breathability is an essential indicator for evaluating thermal comfort, especially in products that directly contact the skin.

Similarly, the clothing material is also one factor that directly affects sensory perception. For example, fabrics with poor breathability can cause discomfort, excessive sweating, and skin irritation. On the other hand, breathable fabrics allow moisture and heat to escape, keeping the wearer cool and dry and enhancing thermal comfort⁶³. Words such as “fabric” and “pure cotton” reflect consumers’ attention to the primary materials used in clothing, which directly affect the product’s appearance, texture, and durability. As a fundamental attribute, the type of fabric (such as pure cotton, polyester, etc.) directly affects breathability and moisture absorption. This indicates that consumers expect clothing to have high quality and good performance⁶⁴.

Tactile comfort refers to a garment fitting to a certain extent and matching the wearer by considering the user’s body size and the corresponding parts of the garment. As an essential dimension for evaluating clothing comfort, it directly affects wearing experience and satisfaction. Softness is the primary factor in measuring the tactile comfort of clothing. Soft fabrics can reduce skin irritation and friction, providing a more comfortable wearing experience. Correct sizing affects the appearance of clothing and directly relates to the comfort and functionality of wearing⁶⁵.

Aesthetic comfort principles indicate that our clothes should be comfortable and reflect our style and identity. This concept goes beyond the physical attributes of clothing, such as softness, breathability, and fit, to include the psychological effects of clothing on self-awareness and social interaction. For example, wearing fashionable clothes suitable for specific occasions can significantly increase personal confidence and comfort⁶⁶, where style becomes essential. This balance between form and function is crucial in determining individuals’ overall satisfaction and confidence in their chosen clothing.

Based on these three themes-perceived quality, perceived value, and sensory comfort-we can gain a more comprehensive understanding of consumers’ psychological and behavioral aspects in the clothing purchase decision process. Each theme reveals how consumers evaluate clothing based on different criteria and personal preferences, including but not limited to the physical characteristics of clothing, price, brand image, and subjective perceptions of comfort and aesthetics.

Perceived quality emphasizes the distinction between objective quality and perceived quality, where objective quality involves quantifiable product attributes such as durability and manufacturing processes, while perceived quality is based on consumers’ personal circumstances and evaluations, such as material, comfort, and style. This indicates that consumers rely not only on physical attributes when assessing clothing quality but also on personal emotions and preferences. Understanding this is crucial for clothing companies to design and market strategies effectively to meet consumer demands.

Perceived value reflects how consumers evaluate the overall value of clothing based on multiple dimensions, including functional, emotional, social, cognitive, and situational values. Consumer purchase decisions are influenced by price and quality and personal emotions, social recognition, and value in specific situations. This multidimensional evaluation reveals the importance of increasing product value in the clothing industry by meeting consumers’ needs and expectations.

Sensory comfort emphasizes the importance of thermal, aesthetic, and tactile comfort in clothing selection. These dimensions influence consumers’ overall perception of clothing comfort, including physical comfort, such as temperature regulation and breathability, and aesthetic comfort, such as confidence in wearing and personal style. Clothing companies need to consider these factors and design comfortable and aesthetically pleasing products to increase satisfaction and loyalty.

In summary, clothing companies need to consider consumers’ perceived quality, perceived value, and sensory comfort comprehensively in product development and marketing strategies. By gaining a deeper understanding of these dimensions, companies can more effectively meet consumer needs, thereby enhancing their products’ market competitiveness.

Development of a user-centric clothing design index system

Collecting online reviews of clothing products and utilizing LDA modeling to identify the themes of the review texts, we integrate the textual information to map it to user needs, and developing sustainable clothing design strategies based on user requirements. We invited four product design experts (Table 3), aged between 29 and 37, with an average industry experience exceeding eight years, to integrate and elaborate on these themes and related requirements.

In this study, we meticulously organized and presented the analysis of the LDA topic model. We provided a clear display of keywords and topic labels, accompanied by an explanation of the significance of each topic through a literature review approach. This method enhanced the accuracy and readability of the analysis results, providing experts and readers with an intuitive understanding and facilitating a deeper interpretation of the underlying connections within the analysis.

In the “Discussion” section, we delved into various aspects of the analysis, including examining the identified topic dimensions. These core questions served as fundamental guidelines, ensuring the discussion was comprehensively and profoundly identified and analyzed each topic.

Furthermore, we conducted in-depth interviews with four experts with extensive experience in fashion and clothing-related fields. Before the interviews, we shared the discussion topics and relevant background materials with the experts to ensure they were adequately prepared to participate effectively in the analysis and discussion process. Before engaging in discussions with the experts, we presented them with the results of the LDA analysis to provide them with a basic understanding. We asked them to integrate and translate these analysis results into specific product design requirements based on their professional knowledge and experience.

When synthesizing the discussion results into specific product requirements, we used keywords such as “fabric” and “quality” as examples to guide the experts in concretizing these abstract concepts. This step ensured precise categorization of each topic, laying the foundation for refining subsequent requirements. In the explanations provided by the experts for the requirements, we emphasized thoroughness, including detailed content, potential implementation strategies, and possible challenges associated with each requirement. This approach ensured that each requirement underwent comprehensive deliberation and discussion, providing detailed information for formulating design metrics.

Finally, based on the formulated strategy and leveraging the topic model and research background, we differentiated the strategic framework for sustainable clothing in manufacturing and environmental aspects (Table 4). We translated the previously abstracted requirements into specific, actionable indicators, providing clear direction for optimizing and improving products. For example, under the “perceived quality” indicator, we specified designing through environmentally friendly production processes and durable materials to reduce clothing waste.

Indeed, these strategies ensure that clothing meets fashion trends in terms of aesthetics and maintains sustainability while meeting users’ needs in functionality and comfort. In terms of perceived quality, objective quality requirements consider durability, manufacturing processes, and material selection to ensure environmentally friendly, durable clothing. Perceived quality demands focus on comfort, proper sizing, and design details to enhance the wearing experience and appearance. Regarding perceived value, functional value emphasizes design durability and practicality; emotional value focuses on personalized design and aesthetics, and social value emphasizes brand social responsibility and sustainable partnerships. Regarding sensory comfort, physical comfort focuses on breathability, temperature and humidity regulation, and material softness, while aesthetic comfort emphasizes the balance between fashion and functionality, identity expression, and social interaction. Sustainability is then addressed through insights derived from the research background and the overall dimensions. Overall, this design strategy emphasizes that businesses need to consider multiple factors such as quality, value, comfort, and social responsibility when designing and manufacturing clothing to meet the comprehensive needs of modern consumers.

Validation of sustainable clothing design strategy

At the current research stage, we are exploring whether the developed sustainability clothing design indicator system is consistent with consumer evaluations. To achieve this goal, based on the 12 sustainable clothing brand strategies previously proposed and referencing relevant studies, we invited an expert group (Table 5) to select the top 35 products from the JD.com website according to the sustainable clothing design strategies. To eliminate any bias that may arise from products of different levels, we selected a pair of products from these 35 that were products that were relatively close in price and functionality. One product (Sample 1 in Table 4) aligns with our developed sustainable clothing design strategy, while another (Sample 2 in Table 4) aligns with lower sustainability. Furthermore, to ensure fairness in the evaluation, we processed the product images and added some textual descriptions to help respondents focus on evaluating product features and usability while avoiding the interference of factors such as brand and color.

The Appendix A shows the proportion of items used for validation. The questionnaire covers four aspects: perceived quality, value, purchase behavior, and purchase intention. The questionnaire for this study was revised based on mature scales validated in previous studies^67,68,69,70. The questionnaire format adopted the Likert five-point scale. Before the formal survey, we randomly selected 6 participants who met the research criteria for questionnaire pretesting. During the pre-survey, participants were asked to assess their understanding of the questionnaire items and provide feedback to optimize the wording for improved readability. In the formal survey stage, all respondents confirmed having relevant product purchasing experience and thoroughly understood the functionality and application of the two samples before filling out the questionnaire. The questionnaire survey was initiated in February 2024, and 657 questionnaires were collected. To ensure the effectiveness of respondent feedback, we set reverse questions and evaluated their attention level based on response time. After excluding invalid samples, this study finally obtained 475 valid questionnaires. The demographic characteristics of the respondents are presented in Table 6.

To validate the effectiveness of sustainable clothing design strategies, we conducted data analysis using SPSS 27 to assess whether there were significant differences in perceived quality, perceived value, purchase behavior, and intention to purchase sustainable clothing. In this study, Cronbach’s α test was used to evaluate the reliability of the questionnaire data. The results showed that the Cronbach’s α value for the dimension of perceived quality was 0.869, for perceived value was 0.855, for purchase behavior was 0.879, and for intention to continue using was 0.880. The Cronbach’s α values for each dimension were more significant than 0.85, and deleting any item within the scale did not increase the Cronbach’s α value, indicating that the data in this study were reliable and suitable for subsequent analysis. As Jansen et al.⁷¹ demonstrated, an alpha value greater than 0.7 is considered sufficient. Alpha values exceeding the 0.85 thresholds confirm that our questionnaire effectively captured the intended dimensions. This high reliability aligns with the findings from psychometric studies in other fields, thereby emphasizing the high standards for scale reliability, ensuring the scale consistently reflects the specific constructs it intends to measure, unaffected by random measurement errors or external variables. Furthermore, in the context of sustainability research, high alpha values indicate that the questionnaire reliably captured dimensions such as perceived quality and value, which are subjective and may vary among respondents. Establishing this level of reliability is crucial for laying a solid foundation for subsequent analyses and ensuring that the research results can effectively inform practice and policy.

Next, we prepared for multivariate analysis of variance (MANOVA). Before the analysis, we tested the assumptions of the method. Kline skewness and kurtosis displayed normal distributions, with skewness values ranging from 0.071 (CI) to 0.002 (FE) and kurtosis ranging from 1.035 (CH) to 0.917 (FE). Therefore, the total skewness of the data was less than 3.0, and the total kurtosis was less than 8.0, indicating univariate normality.

Then, we measured boxplots and identified any outliers. In multicollinearity, the Pearson correlation coefficients between the four dependent variables ranged from 0.28 to 0.45, indicating a mild correlation and no multicollinearity (|r| < 0.9). When conducting Mahalanobis distance analysis, since there were four dependent variables in this case, the corresponding critical value was 18.47. The maximum Mahalanobis distance in this case was 13.45685, which is less than 18.47, indicating the absence of multivariate outliers.

This study used Levene’s Test and Box’s M test to determine whether the data distribution met the assumptions of MANOVA. Table 7 shows that the p-values for both tests were more significant than the significance level (typically 0.05), indicating homogeneity in variance and covariance. Thus, the data met the prerequisites for MANOVA. Therefore, we proceeded with the multivariate analysis of variance using the data.

The analysis of variance results showed that the p-values for samples 1 and 2 in all four dimensions were less than 0.05, indicating that the effects of samples 1 and 2 on all dimensions were significant (Table 8). The analysis revealed differences between the samples in perceived quality (F = 82.642, p < 0.05), perceived value (F = 9.427, p < 0.05), purchase behavior (F = 9.808, p < 0.05), and purchase intention (F = 8.383, p < 0.05). The Partial Eta squared values were 0.149, 0.02, 0.02, and 0.017 for perceived quality, perceived value, purchase behavior, and purchase intention, respectively. These values indicate low to moderate effect sizes for the differences observed between the samples across the four dimensions. Perceived quality had the most significant impact on the samples (0.149), suggesting its significant role, possibly related to environmental awareness. Perceived value and purchase behavior had similar, albeit much lower, values than perceived quality (0.02), indicating some influence on the samples. Purchase intention had the lowest value, suggesting a more minor sample impact.

Based on the Partial Eta squared values mentioned, 0.149 for “perceived quality” indicates a moderate effect size, accounting for 14.9% of the sample variance. This suggests that perceived quality can significantly impact consumer reactions to sustainable fashion. The relatively high percentage indicates that factors such as garment durability, fabric quality, and overall structure are considered critical quality indicators, playing an essential role in shaping consumers’ attitudes and behaviors. This is particularly important in the context of sustainable fashion, as quality not only affects immediate satisfaction but also impacts perceptions of long-term value and sustainability. When Partial Eta squared values are at 0.02 for perceived value and purchasing behavior, these dimensions explain about 2% of the variance. Although small, these effects are still significant, indicating that while these factors influence consumer decisions, their impact is less pronounced than perceived quality. This suggests that in sustainable purchasing, consumers prioritize the product’s intrinsic quality over pricing or immediate monetary value. Similarly, the effect sizes for purchasing behavior suggest that factors such as purchase convenience, availability, and consumer support play a supporting rather than a primary role in influencing sustainable purchasing decisions.

The minimal Partial Eta squared value of 0.017 for purchase intentions indicates a relatively small impact, covering only 1.7% of the variance. This lower level of influence suggests that while the willingness to purchase sustainable fashion is affected by the factors studied, they are likely more significantly driven by external or other variables not captured in this study, such as marketing influences or personal environmental beliefs.

Table 9 shows differences between samples 1 and 2 in perceived quality, perceived value, purchase behavior, and purchase intention. For example, in perceived quality, sample 1 (marked as 1) was significantly higher than sample 2 (p < 0.05). The same trend was observed across other dimensions. Sample 1 was rated higher than sample 2 across all dimensions, consistent with the evaluation results based on sustainable clothing design strategies by experts.

The significant difference in perceived quality between Sample 1 and Sample 2 (t = 9.088, p < 0.05) suggests that consumers consider Sample 1 significantly superior. This could be due to Sample 1 using higher quality materials or communicating its sustainability attributes more effectively. For instance, if Sample 1 utilizes environmentally friendly materials known for durability and performance and highlights these features in its marketing campaigns, it could significantly enhance consumers’ perception of quality. This starkly contrasts Sample 2, which may have yet to use similarly perceived high-quality materials or effectively communicate its sustainability attributes, resulting in a lower perceived quality score.

Although smaller than the differences in perceived quality, the differences in perceived value are statistically significant (t = 3.073, p < 0.05). This discrepancy suggests that even if Sample 1 might be priced higher, the perceived benefits (such as durability and ethical production practices) are considered to justify the cost. Sample 2’s lower perceived value might not align with consumers’ expectations of cost versus benefits, thereby impacting its lower valuation.

Similarly, variations in purchasing behavior and intentions further highlight the differences between the two samples. Consumers are more likely to buy and intend to purchase Sample 1 due to its higher ratings in quality and value. Sample 1 may more clearly align with consumer values regarding sustainability and product quality compared to Sample 2. These differences indicate that Sample 1’s marketing strategy may highlight the long-term benefits of sustainable purchasing more effectively than Sample 2.

Based on the LDA thematic model shows that this study constructs a sustainable clothing strategy with four core dimensions and twenty-six guiding strategies. The proposed model can be used as a basis for subsequent research, and through textual analyses and literature review, we have been able to clearly define the concept of the apparel industry in terms of sustainability. Specifically, apparel quality includes durability, conformability, fit and size, materials and manufacturing processes, and production consistency. The strategy aims to provide apparel companies with an in-depth understanding of consumer perspectives on sustainability issues and the importance they place on them, and to provide designers with valuable market insights to grasp changes in consumer demand.

With the growing public interest in sustainability issues, this study closely aligns with the 2030 Agenda for Sustainable Development proposed by the United Nations in 2015^72,73. By combining thematic modelling with user comments, the study explores ‘sustainable design’ as a goal for the apparel industry. The study focuses on e-commerce consumers and aims to drive progress in research and development activities. With the popularity of e-commerce platforms, many consumers are willing to shop and express their intuitive feelings about products on websites⁷⁴. The study found that sustainable apparel design strategies play a crucial role in raising consumer awareness and concern about environmental issues. These strategies include increasing transparency in the industry, using environmentally friendly materials, being mindful of the environment during the manufacturing process, and providing education about sustainable practices. Integrating personal norms and corporate social responsibility (CSR) expectations can influence eco-friendly purchase intentions and close the gap between attitudes and behaviours in sustainable clothing decisions⁷⁵. Furthermore, the use of natural materials such as organic household wool and encouraging recycled designs proves the acceptability of eco-friendly products⁷⁶. This is in line with our sustainable clothing design strategy. Notably, sustainable fashion also influences the decision-making process by emphasising criteria such as sustainable materials and eco-friendly brand image⁷⁷. In terms of environmental impact, sustainable apparel design strategies significantly reduce resource consumption and environmental pollution in the fashion industry. esearch has shown that designer-led product development can minimise resource entry and waste, improve emissions and energy use, and contribute to the circular economy⁷⁸. Reusing and recycling discarded fashion items reduces energy use and textile waste⁷⁹ and plays an important role in contributing to waste reduction, resource transformation, and controlling pollution through practices such as reuse, recycling, and remodelling of garments⁸⁰. The shift from a simple to a systemic approach to sustainable textiles and fashion is also key⁸¹. In summary, these theories suggest that the sustainable clothing design strategies developed in this study could potentially benefit companies or consumers across multiple dimensions. Furthermore, it is important to note that although more and more consumers are becoming interested in sustainable fashion, when it comes to actual purchasing, they are still more focused on factors such as product quality, fashionability, and price, which is exemplified by Indian consumers⁸².

Therefore, even if companies implement sustainable apparel design strategies, they may not be able to attract a sufficient number of consumers. In the past, most consumers’ impressions of apparel products were based on the style and quality of the product, as well as the feel of the product, and not all generations of consumers had a specific idea of how sustainable apparel could be⁸³. However, our study argues against this, based on the data surveyed the population indicated that consumers of all ages responded positively to products with sustainable attributes. In contrast, the opposite is true for products with fewer sustainable attributes. This means that almost all age groups have a favourable attitude towards products made with sustainable clothing design strategies. It should be noted that the population we studied was mainly from Chinese e-commerce platforms, which may only represent the attitudes of Chinese consumers.

The research findings indicate that to promote sustainable clothing early in product design, improve consumer purchase desire, and help companies formulate sales plans, attention should be paid to perceived quality, value, purchasing behavior, and intentions. These four dimensions correspond to four particular market sales strategies.

Firstly, for perceived quality, consumers typically assess the quality of a product based on factors like durability, fabric quality, and overall construction, which directly influence their purchasing decisions and brand loyalty. Kim et al.⁸⁴suggest that high-quality products can showcase their environmental benefits, thus enhancing consumer recognition and loyalty to the brand. As discussed in the interpretation of Theme 1, this study found a close link between perceived quality and the sustainability of fashion products. High-quality fashion products often win broader market recognition by reducing environmental impacts. It may be beneficial to emphasize the durability and fabric advantages in marketing materials, which could help consumers recognize the long-term value of the products. Additionally, when products display excellent quality, consumers see them as worthy investments, expecting them to have a longer lifespan and thus minimize environmental impact⁸⁵. Demonstrating the product’s durability in actual use can enhance consumer perception of its ecological benefits, which could be beneficial for increasing market acceptance.

Secondly, perceived value is one of the core factors consumers consider when evaluating a product involving a cost-benefit analysis. Sustainably, if high-quality products are priced reasonably, consumers are more willing to purchase, reflecting that consumers are not only concerned with price but also value ethical and environmentally friendly production standards. Desch et al.⁸⁶ show that consumer perceived value depends not only on price but also closely on the product’s social, emotional, and environmental values, which are particularly important in sustainable fashion consumption. Our research supports this view, identifying in the explanation of Theme Model 2 that consumer price concerns are based on perceptions of other factors, such as the emotional value of sensory experiences like color, smell, and touch. Storytelling that emphasizes brand effects, emotions, and environmental values might make it easier for consumers to understand the link between high prices and high values. Moreover, the brand’s reputation and consumers’ trust in the brand also affect the construction of perceived value, thereby influencing the final purchasing decision⁸⁷. Thus, consumers may recognize a product that, despite its higher price, can demonstrate its long-term economic and environmental benefits.

Thirdly, purchasing behavior is directly influenced by perceived quality and value. When consumers believe that a brand’s products excel in these aspects, they are more likely to make a purchase decision⁸⁸. Targeted marketing strategies, such as showcasing the quality and value of products to specific consumer groups, may be beneficial in prompting them to make a purchasing decision. Additionally, consumers’ purchasing behaviors can also be influenced by marketing, especially when campaigns emphasize the sustainability and long-term benefits of products⁸⁹. Interactive social media and online marketing activities emphasizing the product’s sustainability and long-term benefits may enhance consumers’ purchasing intentions. However, it’s worth noting that excessive marketing can lead to consumer skepticism about the intrinsic value of products, resulting in the so-called “greenwash” phenomenon.

Fourthly, although perceived quality and value significantly impact purchase intentions, the formation of purchase intentions is a complex process influenced by multiple factors. Research shows that marketing, social influences, and personal environmental beliefs may affect consumer decision-making⁸⁴. To increase the market acceptance of sustainable fashion products, brands must adopt multifaceted strategies to enhance product quality and effectively promote high perceived value and long-term benefits⁸⁷. Integrating multi-channel marketing strategies, combining marketing, social influences, and personal environmental beliefs, may effectively increase market acceptance of sustainable fashion products. Educational advertising and public relations activities that communicate sustainable products’ long-term economic and environmental benefits may influence consumer purchasing decisions.

Finally, the implementation of sustainable clothing design strategies may increase production costs, which may affect the price of the product and, consequently, the price competitiveness of the company in the market. The concept of this set of strategies may seem difficult to realise for the designers of sustainable clothing and the companies involved. We recommend that businesses continue to introduce new designs that are in line with sustainable clothing and use environmentally friendly and innovative materials while maintaining profitability in order to maintain a high level of consumer appreciation of sustainable clothing. In addition, it is important to note that yes cosmetic design can easily lead to the loss of sustainable attributes in garments, i.e. intricate patterns and coloured designs when focusing on aesthetics will not lead to a significant positive change in consumers.

In conclusion, this study adopted a comprehensive approach, utilizing LDA (Latent Dirichlet Allocation) analysis of user comments to guide product design improvements, particularly in developing and validating sustainable clothing design strategies. By employing LDA text mining technology in conjunction with the Delphi method, we successfully developed sustainable clothing design strategies tailored to user needs, demonstrating the rationality and effectiveness of this process. As a result, our research outcomes provide valuable insights into sustainable clothing design, aiding in optimizing its design and production processes.

Theoretical significance

This study holds significant importance for understanding sustainable clothing trends and critical issues.By analyzing user demands to examine the implementation of sustainable design strategies in the fashion industry and their impact on consumer behavior, this research provides valuable insights for expanding the theoretical framework of sustainable clothing. In particular, we explored consumer acceptance of sustainable clothing and specific influencing factors, enriching relevant theories. This study further enhances our understanding of analytical methods in this field by systematically generating topic models using LDA. We firmly believe that human intelligence plays an irreplaceable role in efficiently extracting profound insights from vast amounts of data. While machine learning is practical, it still requires theoretical reasoning to refine and interpret data. LDA aggregates textual data into a thematic structure, providing visualization and modeling, but it does require considerable time and effort from researchers to interpret these findings. Notably, the Chinese language has polysemy and ambiguity issues, which may affect the accuracy of the analysis. Therefore, the results of LDA need to be cross-checked and interpreted using other methods, such as the Delphi analysis. In sustainable clothing design strategies, we also explored how various dimensions play a crucial role in enhancing the attractiveness of sustainable clothing and reducing environmental impact.

Practical significance

This study modeled the topic on user review data from JD.com and developed a sustainable clothing design strategy based on online comments. The strategy primarily focuses on perceptual quality, sensory comfort, perceived value, and sustainability. These topics encompass clothing quality, comfort, consumer perception, and environmental and social impact. Perceptual quality concerns the durability, adaptability, and material quality of clothing. Sensory comfort considers the tactile feel, breathability, and temperature regulation of clothing. Perceived value includes design aesthetics, multifunctionality, and cost-effectiveness, while sustainability focuses on the environmental and social impacts of clothing production and use. In conclusion, the integrated consideration of these strategies helps guide the design and production of sustainable clothing. For businesses and manufacturers, understanding and applying these strategies will contribute to enhancing product competitiveness, meeting user needs, and achieving sustainability goals.

Limitations and the future

This study acknowledges several limitations in using the LDA model to develop sustainable clothing design strategies. It is important to note that this study primarily utilized user review data extracted from JD.com, which, although extensive, represents a specific segment of the e-commerce landscape in China. These findings may only partially generalize across different platforms or offline retail contexts where broader consumer behaviors are evident. Future research could expand on this by integrating data from multiple sources to better understand consumer attitudes toward sustainable fashion. Future research could benefit from including more participants from different countries or continents to consider consumers’ attitudes and behavioral influences toward sustainability across various cultural backgrounds. The sample scope of this study was limited and may not fully represent the viewpoints and preferences of the entire consumer population. Secondly, respondents’ answers may be influenced by subjective biases and social expectations, leading to limitations in the research results that could steer companies toward misguided product improvements.

Therefore, future research could improve by expanding the sample size, employing appropriate research methods, and exploring consumer behaviors and preferences. Further research could also investigate the effectiveness and market performance of sustainable clothing design strategies. Furthermore, exploring new technologies and materials to enhance the design and production of sustainable clothing could also be valuable in promoting the application of sustainability in the fashion industry. In summary, future research could build upon existing studies to delve deeper into refining sustainable clothing design strategies and advancing the sustainable development of the fashion industry.