Development and Validation of a Framework on Consumer Satisfaction in Fresh Food E-Shopping: The Integration of Theory and Data

Abstract

Consumer satisfaction critically determines the operational sustainability of fresh food e-commerce platforms, yet integrated investigations combining multi-source data remain scarce. This study develops a theory–data fusion framework to identify key satisfaction drivers in China’s fresh e-commerce sector. Utilizing Python-based crawlers, we extracted 1252 online reviews of Aksu apples from a certain fresh produce e-commerce platform alongside 509 validated questionnaires. Through systematic literature synthesis, three core dimensions—perceived value (price–performance balance), platform experience (interface usability), and perceived quality (freshness assurance)—were operationalized into measurable indicators. The final structural equation model reveals that perceived value, platform experience, and perceived quality all have significant positive impacts on consumer satisfaction. This study pioneers a methodological paradigm integrating computational text mining (Octopus Collector + SPSS Pro) with traditional psychometric scales, achieving superior model fit (RMSEA = 0.023, CFI = 0.981). These findings empower platforms to implement a precision strategy. The validated framework provides a theoretical basis for omnichannel consumer research while addressing the data-source bias prevalent in prior studies.

1. Introduction

Fresh food e-commerce refers to online platforms specializing in the sale of perishable agricultural products through cold chain logistics and instant delivery systems [1,2]. These platforms are integral to contemporary society. Data from Whale Staff [3] indicate that the combined online sales of major fresh food e-commerce platforms (e.g., Jindong, Tmall, and Taobao) reached approximately 64 million units and CNY 3.9 billion in February 2024, marking a 30% increase from 2023. This growth rate significantly exceeds that of the broader e-commerce sector, underscoring the substantial potential and dynamism of the fresh food e-commerce market [4,5]. Concurrently, the evolution of these platforms has not only enhanced consumer convenience and efficiency in shopping [6,7] but also spurred the modernization and transformation of the agricultural sector [8,9]. However, fresh food e-commerce platforms face challenges such as market competition [10], supply chain management, and the diversification of consumer demands [2,6]. As consumer satisfaction directly determines repurchase intention and platform loyalty in this high-frequency purchase sector [11], understanding its multidimensional drivers becomes critical for sustaining competitive advantages. As technology progresses and living standards rise [12], consumer demands have evolved from basic sustenance to encompass safety, social interaction, esteem, and self-actualization [13,14,15]. FRESHIPPO (FRESHIPPO is Alibaba’s integrated fresh food supply chain brand, combining online retail, offline supermarkets, and logistics to ensure rapid delivery of perishables.) [16] has garnered a user base with its high-frequency, low-cost, direct-from-orchard model, while Qian Da Ma (Qian Da Ma, a community-based fresh food chain, is renowned for its “no overnight meat” policy and daily inventory clearance strategy. It participated in drafting China’s national standard Agricultural Socialization Service—Community Fresh Store Service Specifications (GB/T 44591—2024), reflecting its industry influence). [17] has established an industry standard with its “no overnight meat” policy. Thus, the ability of fresh food e-commerce platforms to accurately discern and address consumer needs is crucial for achieving a competitive edge.

Consumer satisfaction is a critical construct in e-commerce platform research. Previous studies have provided valuable insights into consumer preferences and satisfaction in the fresh food e-commerce sector. Chinese consumers prefer fresh produce, particularly meat, eggs, vegetables, fish, and seafood [18]. Key product attributes—including nutritional content, certified natural ecological value, origin [19,20], food safety compliance [21], eco-certification [22], and sensory properties [23]—substantially influence consumer satisfaction [24]. Platform factors, including information and system service quality, also play a significant role [25]. Furthermore, product quality [26], pricing, ease of use, reliability [27], responsiveness, service convenience, and personal norms contribute to satisfaction levels [2]. During the COVID-19 pandemic, food safety became a paramount concern for consumers [28]. Additionally, logistical reliability, humanized delivery [29], health commitments, consistent service efforts, and empathic engagement strategies emerged as crucial factors affecting satisfaction in fresh food e-commerce [30]. These findings underscore the complexity and multidimensional nature of factors influencing consumer satisfaction in this domain. However, most existing studies rely on a single data source, with few integrating offline questionnaires and online review data. Existing studies exhibit two key limitations in investigating consumer satisfaction: First, heavy reliance on single data sources—either survey self-reports or online reviews—fails to reconcile stated preferences with actual behavioral data, risking common method bias [31]. Second, the theory–data dichotomy persists, where theoretical models inadequately explain emerging phenomena like pandemic-induced safety concerns [28], while data-driven approaches lack theoretical anchoring. This paper pioneers a novel mixed-methods approach that combines offline questionnaires (N = 509) with big data analytics of 1252 online reviews. This integration enables cross-validation of findings while capturing both stated preferences from surveys and revealed preferences from behavioral data [32,33]. The methodology improves research findings by investigating consumer satisfaction on e-commerce platforms through merging these two data sources, thereby addressing the single-source bias prevalent in existing studies.

This study contributes to current literature by proposing a roadmap to develop an integrated framework with theory and data. Prior studies on customer behavior have struggled between theory-first and empirical-first approaches for a long time. On one hand, the theory-first (TF) methodology is initiated with a literature review, seeking to identify moderators, mediators, extensions, and applications of established effects, often focusing on determining the direction of causal effects [32]. On the other hand, the empirical-first (EF) approach is stimulated by real-world data, enabling the discovery of novel research questions unconstrained by existing theory [32]. To systematically investigate factors influencing customer behavior in the context of online shopping, the dichotomy between theory and data must be reconciled through an integrated framework. This approach ensures that theory provides interpretability for the findings, while data bolsters the relevance and immediacy of the discussion and its implications.

The objective of this study is to construct and validate a consumer satisfaction framework specifically tailored for fresh food e-commerce platforms. The structure of the subsequent sections of this paper is as follows: Section 2 presents a literature review, precedes the preprocessing of data obtained from online crawlers, and culminates in model development based on the aforementioned findings; Section 3 delineates the questionnaire design and the formulation of hypotheses; Section 4 carries out the empirical analysis; Section 5 discusses the results and implications; Section 6 summarizes this study’s conclusions and limitations and suggests directions for future research.

2. Research Background and Model Development

Our roadmap towards an integrated framework is depicted in Figure 1. This study distinguishes itself from prior research. Initially, the satisfaction model and its theoretical underpinnings are elucidated through a literature review, during which the determinants of customer satisfaction are identified. Concurrently, word frequency analysis and statistical processing are conducted on user comment data extracted from online sources. Subsequently, by aligning the factors gleaned from theoretical literature with offline user comment data, a questionnaire is formulated, and reliability and validity analyses are performed to establish the hypotheses of this study. Ultimately, conclusions are drawn from the empirical analysis of the proposed roadmap.

The concept of “integration” in our study title serves as an analogical framework to characterize the interactive relationship between theoretical foundations and empirical evidence. Our analytical framework integrates two dimensions: (1) theoretical constructs grounded in academic literature and (2) empirical patterns identified through semantic analysis of consumer narratives. This two-way interaction mechanism overcomes the constraints of purely deductive or inductive methodologies, providing a comprehensive perspective for understanding consumer satisfaction. In rapidly evolving sectors like fresh food e-commerce, where rigid predefined models often fail to capture market dynamics, the continuous dialogue between theoretical postulation and data calibration significantly enhances the explanatory power of research frameworks.

2.1. Literature Review

2.1.1. Theoretical Foundations of Consumer Satisfaction and Loyalty

Philip Kotler, a pioneer in the field of marketing [34], posited that customer satisfaction reflects the mental state of consumers who feel either pleased or dissatisfied with the actual outcomes of a product when compared to their expectations [35]. Customer satisfaction typically denotes consumers’ post-purchase evaluation grounded in cognitive-affective appraisal processes [36]. Customer satisfaction is inherently dynamic [37]. Although consumers are influenced by both objective and subjective factors, their attitudes toward the same products and services may fluctuate in response to environmental changes and shifts in personal perceptions. Consumer satisfaction is not static; it can vary even when the characteristics of a product brand remain unchanged [38]. Therefore, it is crucial for companies to understand customers’ ongoing needs and tendencies [39] while continuously adapting to market fluctuations and evolving consumer psychology in order to enhance customer satisfaction and maintain a competitive advantage. In addition to focusing on product quality and service levels [40], organizations must also stay attuned to consumer feedback and expectations. This vigilance allows for appropriate adjustments and improvements that ensure customer satisfaction remains consistently high [41].

While this study primarily focuses on satisfaction drivers, the satisfaction–loyalty nexus remains theoretically critical. The expectation–disconfirmation paradigm [36] establishes satisfaction as a prerequisite for loyalty, where cumulative positive experiences foster both attitudinal commitment (preference) and behavioral loyalty (repurchase). Customer loyalty constitutes a dual-dimensional construct encompassing both attitudinal commitment (psychological preference) and behavioral persistence (repeat purchase) [42]. Within the fresh food e-commerce context, loyalty manifests through platform stickiness (frequency of repurchases) and advocacy behaviors (word-of-mouth recommendations). Although loyalty is regarded as a natural outcome of satisfaction, this study focuses more on identifying the antecedent drivers to guide the pre-loyalty strategies for intervention measures.

2.1.2. Comparative Analysis of Classic Satisfaction Models

The American Customer Satisfaction Index (ACSI) establishes satisfaction as a core driver of loyalty through perceived quality, expectations, and value [43]. The European Customer Satisfaction Index (ECSI) extends this framework by emphasizing corporate image as a mediator between satisfaction and loyalty [44]. The Chinese Customer Satisfaction Index (CCSI), tailored to China’s consumer-centric market, prioritizes perceived quality and post-purchase behavior [45]. While these models share a hierarchical structure linking antecedents (e.g., quality, value) to outcomes (loyalty), their cultural and contextual adaptations highlight the need for a hybrid approach in fresh food e-commerce research. Our model synthesizes ACSI’s causal-chain logic, ECSI’s mediation emphasis, and CCSI’s quality centricity, while integrating emergent dimensions from user-generated data (e.g., platform experience) to reflect digital consumption dynamics.

2.1.3. Perceived Value

Perceived value is defined as “the consumer’s overall assessment of the utility of a product based on perceptions of what is received and what is given” [46]. It is achieved by weighing the benefits that customers can identify against the costs of obtaining the product or service [47]. High perceived value typically indicates that customers evaluate the product or service favorably, which subsequently enhances customer satisfaction [48]. Customers are more inclined to experience satisfaction with a product or service when their perceived benefits outweigh the costs incurred [49]. Research has demonstrated that service quality positively influences perceived value, which in turn further impacts customer satisfaction [50]. For instance, the service quality of intellectual property (IP) intermediaries, which are defined as platforms facilitating brand-certified agricultural product transactions [51], manifests through four dimensions: (1) interaction quality, reflecting responsiveness and empathy in customer service [52]; (2) output quality, involving accuracy in order fulfillment such as freshness consistency [25,53]; (3) IT quality, covering system reliability and interface usability [54]; and (4) integration quality, requiring seamless coordination between suppliers and logistics [2,55]. These dimensions collectively enhance customers’ perceived functional value (utilitarian benefits like cost-effectiveness) and affective value (emotional benefits like trust), thereby increasing overall satisfaction [56].

Perceived value serves as a dominant mediator in the relationship between service quality and customer satisfaction, where service quality primarily enhances satisfaction by elevating customers’ perceived value [57,58]. This unidirectional causal chain is empirically supported in fresh food e-commerce contexts: functional value (e.g., cost-effectiveness) mediates the effects of output quality and interaction quality on satisfaction, while affective value (e.g., trust) mediates IT quality and integration quality’s impacts [56]. A secondary feedback loop exists post-purchase: satisfaction retrospectively amplifies perceived value through cognitive reappraisal [59]. For instance, satisfied customers who repurchase or recommend products [60] reinforce their perceived value via social validation [61]. However, this reverse effect (satisfaction → value) is context-dependent and subordinate to the primary mediation.

2.1.4. Platform Experience

Platform experience encompasses the overall perception users have when engaging with a specific platform. This includes various elements such as interface design, interaction fluency, functional utility, content quality, responsiveness, personalization levels, and safety and security measures, among other factors [62]. A positive platform experience can render the usage process enjoyable, convenient, and efficient [63]. An aesthetically pleasing website interface that is clear, user-friendly, and fast significantly enhances customer access and engagement [64]. Providing objective, accurate, comprehensive, and timely information about products or services helps to prevent misleading or deceptive practices while fostering customer trust. Ensuring transaction security along with offering multiple payment options and effective order management features can mitigate perceived risks for customers and enhance their satisfaction during transactions.

A favorable platform experience contributes to increased user satisfaction [65], which in turn encourages more frequent use of the platform, creating a positive feedback loop. Satisfied users are more inclined to become loyal patrons of the platform; they are also likely to engage in word-of-mouth promotion that attracts new users. Conversely, a subpar platform experience diminishes user satisfaction, leading to churn rates increasing while potentially inciting negative word-of-mouth publicity detrimental to the platform’s reputation.

2.1.5. Perceived Quality

Product quality serves as a crucial criterion for consumers in assessing the value of a product, directly influencing their purchasing decisions and overall satisfaction [66]. High-quality products are capable of fulfilling customer needs and delivering positive experiences, thereby enhancing customer satisfaction [67]. Such products offer stable and reliable performance, minimize failure rates and maintenance costs, extend product lifespan, and alleviate customers’ perceived risks. The recognition of high-quality products by customers often translates into loyalty towards the company [68], which subsequently bolsters the company’s brand image and competitiveness through word-of-mouth communication [69].

2.2. Data Crawling and Preprocessing

Data crawler technology is widely used in data collection on major public websites due to its accuracy and efficiency [70]. This technology automatically captures publicly available information on web pages and stores and processes it as needed [71]. During this process, we pay special attention to the legal origin of the data and the protection of user privacy [72].

Specifically, user reviews of Aksu apples were crawled from the public comment section of Jing Fresh (a leading Chinese fresh food e-commerce platform, [Jingxiansheng Akso Apples 5-pound Pack] Jingxiansheng Apples from Xinjiang Akso, net weight 5 pounds, fruit diameter 85–90 mm. Fresh fruits [Market Prices, Quotes, Prices, Reviews]—JD.com (jd.com)) using Octopus Collector, a web scraping tool compliant with the platform’s robots.txt protocol. During the crawling process, we made sure that the collected data were obtained entirely based on public channels and did not involve any personal sensitive information, such as names, contact information, addresses, and other private content [73,74]. At the same time, we strictly de-identified the crawled comment data to ensure that they could not be directly traced back to specific users, thus fully protecting user privacy.

We used Octopus Collector to collect review data and further implement word segmentation processing, word frequency analysis, and word cloud visualization to visually display the service indicators of fresh e-commerce platforms that consumers are more concerned about [75,76].

In this study, Octopus Collector was used to obtain the data of user reviews of Jing xian Sheng Aksu apples on 12 March 2024 from a large online shopping platform. Initially, 1252 data were crawled, and after filtering and eliminating invalid data, 1046 valid reviews were obtained. The data include user comment ID, product name, evaluation star rating, evaluation content, number of likes, number of comments, time of follow-up comments, follow-up comment content, and product attributes.

2.2.1. Sentiment Analysis of Comments

Sentiment analysis of captured user reviews was conducted using Octopus Collector, with sentiment scores ranging from −50 to 50. The analysis showed 84% positive, 11% neutral, and 5% negative. By looking at the sentiment analysis graph, it can be seen that the sentiment scores range from −50 to 50. A score closer to 50 indicates a higher level of user satisfaction and vice versa. According to Figure 2, the review sentiment scores are mainly centered on 0 and above, which indicates that most of the users are satisfied with the Aksu apples.

2.2.2. Analyzing the Text of Good Reviews

Consumers’ concern for product quality is reflected in the reviews, showing satisfaction with taste, freshness, and cost-effectiveness. Consumers are willing to share their love for apples and their willingness to recommend them. To show the positive consumer reviews of Aksu apples, this study screens high-frequency words and generates histograms.

Figure 3 shows that the words that consumers use more frequently, such as “very good”, “pretty good”, “satisfactory”, “worthwhile”, etc., reflecting that the product meets consumers’ needs and is attractive and worth buying. Furthermore, lexical descriptors such as “fresh” (core attribute), intensifier-modified terms (e.g., “very sweet”), and evaluative adjectives (“cheap”, “tasty”) epistemologically manifest consumers’ multidimensional perception patterns, as evidenced through our sentiment-embedded text-mining framework.

The words “sound” and “sufficient” show customers’ expectations and anticipation of the product. Taken together, consumers’ positive ratings show a high level of satisfaction with Aksu apples and indicate that they are willing to continue to buy them and actively recommend them to others, highlighting the level of customer satisfaction with the product.

2.3. Model Development

The linkage between satisfaction and loyalty has been extensively theorized in classic models such as the American Customer Satisfaction Index (ACSI), European Customer Satisfaction Index (ECSI), and Chinese Customer Satisfaction Index (CCSI). ACSI posits that satisfaction directly drives loyalty through reduced complaints and enhanced perceived value [43], while ECSI emphasizes the mediating role of corporate image [44]. CCSI further integrates perceived quality as a core antecedent, reflecting China’s consumer-centric market dynamics [45].

As depicted in Figure 1, our roadmap synthesizes theoretical constructs (CCSI dimensions) with data-driven insights from text mining. For instance, literature-derived variables like perceived quality were operationalized using review keywords (e.g., “fresh” and “tasty” in Figure 3), while platform experience emerged from sentiment analysis of interface-related terms. This dual approach ensures theoretical rigor while capturing real-world consumer concerns.

Based on the statistical analysis of consumer reviews and literature materials, the concerns of consumers regarding Aksu apples can be classified into the following ten parts. From the perspective of logistics service quality, consumers pay attention to delivery speed and handling of emergencies. In terms of product quality, consumers are concerned about safety, freshness, brand, and taste. Platform service quality includes customers’ evaluations of problem solving. Platform design includes interface clarity. By analyzing switching costs, discounts and promotions, perceived value, customer trust, satisfaction, and loyalty, we deeply explore consumers’ demands and evaluations of products and services, providing guidance for enhancing customer experience.

Through the analysis, sorting, and summarization of the results in this chapter, it provides a basis for determining satisfaction-related questions in the design of the survey data questionnaire and the selection of models. The model constructed in this paper includes 10 primary dimensions, illustrated in Figure 4. These dimensions comprehensively reflect consumers’ multi-faceted demands and attitudes towards products and platforms. By comprehensively drawing on the ACSI, ECSI, and CCSI models and conducting word frequency analysis on customer reviews through data crawling, we were able to construct the theoretical model we hypothesize, as shown in Figure 4.

3. Questionnaire Analysis and Hypotheses Development

Traditional satisfaction studies often predefine hypotheses based on theoretical models (e.g., ACSI). However, our integrated framework necessitates a phased hypothesis development process. Initial hypotheses were implicitly guided by the literature (Section 2.1) and data-mined dimensions (Section 2.2). The final hypotheses (Section 3.2) were explicitly formulated after principal component analysis (PCA) reclassified variables into four consolidated dimensions (Figure 5). This approach aligns with exploratory sequential mixed methods [77], where quantitative data refine theoretically informed constructs. By adapting our model through data-driven insights, we reduce confirmation bias and strengthen practical relevance, particularly in emerging markets like fresh food e-commerce where empirical benchmarks remain scarce. This iterative refinement process ensures methodological alignment with the dynamic nature of agricultural e-commerce systems.

3.1. Questionnaire Analysis

This part will focus on the design and implementation of the survey. It will include descriptive statistics, reliability and validity analysis, and principal component analysis.

3.1.1. Descriptive Statistics

To ensure the timeliness and validity of the survey results, we specifically selected consumers who have recently purchased fruit on fresh e-commerce platforms [78]. The survey was conducted in Tianjin, China, from 15 April to 20 April 2024. A random sampling method was employed to recruit participants who had purchased fruits on fresh e-commerce platforms within the past three months. In compliance with China’s Personal Information Protection Law (PIPL) [79], the survey did not collect sensitive demographic data (e.g., residential address or settlement type) to protect respondent privacy.

Of the total 509 survey respondents, a slightly higher percentage (50.5%) were male, the primary age group was 26–35 years old (22.2%), the most common education level was a bachelor’s degree (34.0%), the most common monthly income was RMB 5000–8000 (47.7%), and the most common occupation was that of a business employee or manager (36.3%).

Table 1. Social demographic description (N = 509).

Item	Option	Frequency	Percentage (%)
Gender	Male	257	50.5
	Female	252	49.5
Age	18 years and under	54	10.6
	18–25 years	88	17.3
	26–35 years	113	22.2
	36–45 years	92	18.1
	46–55 years	59	11.6
	56–65 years	89	17.5
	66 and over	14	2.8
Educational attainment	High school and below	102	20
	Tertiary	170	33.4
	Undergraduate	173	34
	Master’s degree and above	64	12.6
Income level	USD 3000 and below	89	17.5
	USD 3000–5000	68	13.4
	USD 5000–8000	243	47.7
	USD 8000–10,000	79	15.5
	USD 10,000 and above	30	5.9
Occupation	Academic researcher	31	6.1
	Institutional/professional staff	48	9.4
	Business employees/managers	185	36.3
	Professionals and technicians (e.g., doctors/teachers/lawyers/sports/journalists, etc.)	143	28.1
	Freelancers	43	8.4
	Retirees	45	8.8
	Other	14	2.8

shows the demographic characteristics of the respondents.

3.1.2. Statistical Characteristics of Purchasing Behavior

Of the 509 survey respondents, the vast majority (507, or 99.6%) have had the experience of purchasing fruits from fresh produce e-commerce platforms. In terms of purchase frequency, 159 (31.2%) of them purchased one to three times per month. Purchase concerns showed that a slightly higher proportion of respondents (59.1%) had product quality as their main concern.

Table 2. Statistics of sample purchasing behavior.

Item	Option	Frequency	Percentage (%)
Any experience in purchasing fruits from fresh food e-commerce platforms	Yes	507	99.6
	None	2	0.4
Purchase frequency	Less than once a month	129	25.3
	1–3 times per month	159	31.2
	4–5 times per month	152	29.9
	More than 5 times per month	69	13.6
Purchase concerns	Product quality	301	59.1
	Product prices	208	40.9

presents the characteristics of respondents’ purchasing behavior.

3.1.3. Reliability Analysis

In this paper, the online tool SPSS Pro (https://www.spsspro.com) was used to conduct a reliability analysis of 10 measurement variables. According to

Table 3. Questionnaire reliability analysis.

Variable	Number of Questions	Coefficients
Quality of logistics services	2	0.486
Product quality	4	0.570
Platform quality of service	4	0.618
Platform design and features	2	0.465
Switching costs	3	0.536
Promotions	2	0.426
Perceived value	3	0.569
Customer trust	3	0.554
Customer satisfaction	3	0.540
Customer loyalty	5	0.681
Total	31	0.927

, the initial Cronbach’s alpha values for sub-dimensions ranged from 0.426 to 0.681, suggesting item refinement needs, while the overall scale (α = 0.927) confirmed high internal consistency. Thus, exploratory factor analysis (EFA) was adopted in the next step.

3.1.4. Validity Analysis

According to Kaiser’s criteria, KMO values were used to assess the structural validity of the questionnaire. KMO values above 0.8 indicate high validity, above 0.7 is better reliability, above 0.6 is acceptable, and below 0.6 is not recommended for low validity [80]. The validity of the questionnaire can be assessed by applying SPSS for factor analysis and combining the KMO value and Bartlett’s spherical test. Please refer to the specific results in

Table 4. Questionnaire validity analysis.

Variable	Number of Questions	KMO	Sig
Quality of logistics services	2	0.500	<0.01
Product quality	4	0.668	<0.01
Platform quality of service	4	0.699	<0.01
Platform design and features	2	0.500	<0.01
Switching costs	3	0.618	<0.01
Promotions	2	0.500	<0.01
Perceived value	3	0.617	<0.01
Customer trust	3	0.616	<0.01
Customer satisfaction	3	0.617	<0.01
Customer loyalty	5	0.774	<0.01
Total	31	0.956	<0.01

.

Comprehensive validity analysis showed that the KMO values of the 10 variables exceeded 0.5, with an overall KMO value of 0.956, and the significance probability of Bartlett’s test of sphericity was less than 0.05, indicating that the questionnaire had good structural validity. However, the results of the combined reliability and validity analyses showed that the questionnaire questions did not adequately reflect the actual situation, and further optimization of the designed model was recommended.

3.1.5. Exploratory Factor Analysis

As can be seen in

Table 5. Rotating load sum of squares.

	Rotational Load Sum of Squares
Component	Total	Percentage of Variance	Cumulative %
1	3.384	10.917	10.917
2	3.371	10.874	21.791
3	3.279	10.576	32.367
4	2.943	9.495	41.863
Extraction method: principal component analysis

, components 1–4 explain 41.863 percent of the data.

By looking at the rotated component matrix, the model was reclassified into four categories: perceived value relates to customer feelings; platform experience relates to customer satisfaction with the platform; perceived quality relates to the customer’s experience of the products and services upon arrival; and customer satisfaction relates to the maintenance of the relationship between the platform and the customer. The reclassified model is clearer and more accurate, which helps to further understand and optimize the research results.

The items were logically reclassified into four groups based on principal component analysis (PCA), options with loadings below 0.4 were removed, and the model was reclassified into four first-level dimensions [80].

3.2. Hypotheses Development

Our roadmap (Figure 1) operationalizes a mixed-methods framework that triangulates theoretical constructs with empirical data. Phase 1 (theoretical foundation) derives satisfaction determinants from the literature (ACSI/ECSI/CCSI) and identifies gaps through comparative analysis. Phase 2 (data-driven insights) employs web scraping and sentiment analysis to extract user concerns from online reviews, ensuring real-world relevance. Phase 3 (model synthesis) aligns theoretical variables (e.g., perceived quality) with data-mined dimensions (e.g., platform experience) to design a hybrid questionnaire. Phase 4 (validation) utilizes principal component analysis (PCA) to refine the model structure based on survey responses (as shown in Figure 6), enabling hypothesis formulation grounded in both theory and empirical patterns. This iterative process ensures that our hypotheses (Section 3.2) are neither purely deductive nor inductive but emerge from a dialectical integration of top-down and bottom-up approaches. According to the new theoretical model, this paper makes the following assumptions:

H1: 

Perceived value has a significant positive impact on customer satisfaction.

H2: 

Platform experience has a significant positive impact on customer satisfaction.

H3: 

Perceived quality has a significant positive impact on customer satisfaction.

The new model is shown in Figure 5.

4. Results

This part mainly includes the reliability and validity analysis of the new model, CFA model fit test, descriptive statistics and normality test, correlation analysis, SEM model fit test, multiple regression analysis, and finally draws a conclusion.

4.1. Reliability and Validity Analysis

4.1.1. Reliability Analysis

According to the data in

Table 6. Model reliability analysis.

Variable	Number of Questions	Coefficients
Perceived value	7	0.765
Platform experience	7	0.765
Perceived quality	7	0.767
Customer satisfaction	5	0.717
Total	26	0.915

, the Cronbach’s alpha values of each scale are all around 0.7, indicating that they have good reliability. In conclusion, the data from the questionnaires used in this paper demonstrated good reliability, are reliable and stable, and are suitable for further testing [80].

4.1.2. Validity Analysis

Factor analysis was conducted using the online tool SPSS Pro with KMO values and Bartlett’s sphericity test (please refer to the following

Table 7. Model validity analysis.

Variable	Number of Questions	KMO	Sig
Perceived value	7	0.856	<0.010
Platform experience	7	0.855	<0.010
Perceived quality	7	0.851	<0.010
Customer satisfaction	5	0.794	<0.010
Total	26	0.954	0.000

). Based on the results of the validity analysis, the KMO values of these 10 variables all exceeded 0.7. This indicates that the scale (compared with Table 4) has relatively high acceptability, and the questions can better reflect the actual situation. The validity test was passed; therefore, the data collected through this questionnaire are valid.

4.2. Structural Equation Model Analysis

4.2.1. CFA Model Fit Test

From the results of the model fitness test in

Table 8. Model fit test.

Indicator	Reference Standard	Actual Results
CMIN/DF a	1–3 excellent, 3–5 good	1.269
RMESA b	<0.050 is excellent, <0.080 is good	0.023
IFI c	>0.900 is excellent, >0.800 is good	0.981
TLI d	>0.900 is excellent, >0.800 is good	0.979
CFI e	>0.900 is excellent, >0.800 is good	0.981

Note: ^a: chi-squared degrees of freedom ratio; ^b: root mean square of error; ^c: incremental fit index; ^d: Tucker–Lewis index; ^e: comparative fit index.

, the CMIN/DF (chi-squared degrees of freedom ratio) is 1.269, which is in the ideal range, and the RMESA (root mean square of error) is 0.023, which is excellent. The IFI, TLI, and CFI indicators are all more than 0.9, which is at an excellent level. Therefore, the comprehensive analysis results show that the CFA model of customer satisfaction has a good fit.

After confirming the goodness of fit of the CFA model for the customer satisfaction scale, the convergent validity (AVE) and construct reliability (CR) of the dimensions were further assessed. First, the standardized factor loadings of the measurement items on the corresponding dimensions were calculated based on the established CFA model. Then, the AVE and CR values for each dimension were calculated. According to the criteria, AVE needs to reach 0.5 and CR needs to reach 0.7 or more to indicate good convergent validity and construct reliability.

Analysis of

Table 9. Convergent validity and combined reliability tests for each dimension.

Pathway Relationship			Estimate	AVE a	CR b
A1	<---	Perceived value	0.580	0.320	0.767
A2	<---	Perceived value	0.562
A3	<---	Perceived value	0.620
A4	<---	Perceived value	0.511
A5	<---	Perceived value	0.564
A6	<---	Perceived value	0.567
A7	<---	Perceived value	0.552
B1	<---	Platform experience	0.591	0.320	0.766
B2	<---	Platform experience	0.568
B3	<---	Platform experience	0.634
B4	<---	Platform experience	0.564
B5	<---	Platform experience	0.533
B6	<---	Platform experience	0.517
B7	<---	Platform experience	0.544
C1	<---	Perceived quality	0.569	0.318	0.765
C2	<---	Perceived quality	0.549
C3	<---	Perceived quality	0.526
C4	<---	Perceived quality	0.552
C5	<---	Perceived quality	0.543
C6	<---	Perceived quality	0.604
C7	<---	Perceived quality	0.598

Note: ^a: average variance extracted; ^b: composite reliability.

shows that after the validity test of the customer satisfaction scale, the AVE value of each dimension did not reach 0.5 and the CR value was more than 0.7, while Fornell and Larcker [80] argued that 0.5 is not an absolute criterion, and that at an AVE of slightly less than 0.5 and a combined reliability (CR) of more than 0.6, the scale can still be considered to have good convergent validity. In summary, all dimensions demonstrate strong combinatorial validity and convergent reliability. Therefore, the comprehensive analysis results indicate that the CFA model of customer satisfaction has a good fit, as shown in Figure 6.

4.2.2. Descriptive Statistics and Normality Tests

Table 10. Descriptive statistics of each dimension and normality test results of measurement items.

Dimension	Measurement Question Item	M a	SD b	Skewness	Kurtosis	Overall M	Overall SD
Perceived value	A1	3.420	1.207	−0.581	−0.620	3.397	0.794
	A2	3.380	1.215	−0.546	−0.709
	A3	3.360	1.257	−0.506	−0.802
	A4	3.440	1.232	−0.532	−0.681
	A5	3.410	1.245	−0.493	−0.760
	A6	3.390	1.232	−0.556	−0.723
	A7	3.390	1.223	−0.539	−0.692
Platform experience	B1	3.470	1.176	−0.658	−0.414	3.451	0.769
	B2	3.470	1.219	−0.529	−0.695
	B3	3.40	1.228	−0.489	−0.774
	B4	3.420	1.197	−0.551	−0.649
	B5	3.460	1.209	−0.590	−0.564
	B6	3.450	1.164	−0.589	−0.502
	B7	3.490	1.160	−0.511	−0.569
Perceived quality	C1	3.430	1.196	−0.598	−0.547	3.398	0.777
	C2	3.470	1.211	−0.602	−0.580
	C3	3.490	1.158	−0.616	−0.404
	C4	3.450	1.210	−0.557	−0.635
	C5	3.330	1.222	−0.476	−0.752
	C6	3.330	1.205	−0.454	−0.723
	C7	3.290	1.235	−0.454	−0.789

Note: ^a: mean; ^b: standard deviation.

demonstrates the results of the descriptive statistical analyses and normality tests for the factors used in this study. The mean values ranged between 3 and 4 using a positive scale of 1 to 5. The results show that the study participants’ cognitive and behavioral levels in terms of customer satisfaction are above the medium level.

According to Kline’s [80] criterion, the skewness coefficient within an absolute value of 3 and the kurtosis coefficient within an absolute value of 8 indicate that the data are approximately normally distributed. Analysis of Table 10 shows that the skewness and kurtosis coefficients of the question items in this study are within the criteria, indicating that the data are approximately normally distributed.

4.2.3. Correlation Analysis

The correlation among the constructs was examined using Pearson correlation analysis. The results, presented in

Table 11. Pearson’s correlation analysis between dimensions.

Dimension	Perceived Value	Platform Experience	Perceived Quality
Perceived value	1
Platform experience	0.676 **	1
Perceived quality	0.685 **	0.681 **	1

Note: ** significant correlation at the 0.01 level (two-tailed).

, demonstrate a significant correlation between all constructs, with each achieving significance at the 99% level. The correlation coefficients, r, were all greater than 0, indicating that the constructs showed a positive association. This analysis assumes that the data are approximately normally distributed, which was confirmed by the skewness and kurtosis values being within acceptable ranges (see Table 10). The significant positive correlations suggest that higher scores in one construct (e.g., perceived value) tend to coincide with higher scores in others (e.g., platform experience), providing preliminary support for their interrelated roles in influencing customer satisfaction. Notably, the correlation coefficients (r < 0.7) do not indicate severe multicollinearity concerns, ensuring the robustness of subsequent structural equation modeling (SEM) analyses.

4.2.4. SEM Model Fitness Tests

Without the inclusion of covariates, the results of hypothesis testing concerning the path relationships in the SEM model for factors influencing customer satisfaction are presented in

Table 12. SEM model path relationship test results.

			Estimate	S.E. a	C.R. b	p-Value
Perceived value	--->	Customer satisfaction	0.414	0.630	6.589	***
Platform experience	--->	Customer satisfaction	0.263	0.510	5.145	***
Perceived quality	--->	Customer satisfaction	0.493	0.700	7.068	***

Note: ^a: standard error; ^b: critical ratio. *** significant correlation at the 0.001.

. According to the analysis results displayed in Table 12, perceived value exhibits a significant positive effect on customer satisfaction (β = 0.414, p < 0.01), thereby supporting hypothesis H1. Additionally, platform experience also demonstrates a significant positive impact on customer satisfaction (β = 0.263, p < 0.01), which supports hypothesis H2. Furthermore, perceived quality has a substantial positive influence on customer satisfaction (β = 0.493, p < 0.01), thus corroborating hypothesis H3. The structural relationships among these variables were presented in a visual way in the analysis of structural equation model, as shown in Figure 7.

According to

Table 13. Regression analysis of customer satisfaction.

		Non-Standard Regression Coefficients	Standardized Regression Coefficients	p-Value	VIF	R-Square	Adjusted R-Square
Dimension	Constant term					0.561	0.542
Perceived value	A4	0.073	0.106	0.002	1.334
	A5	0.050	0.074	0.039	1.422
	A6	0.093	0.135	0.000	1.444
	A7	0.059	0.085	0.018	1.426
Platform experience	B2	0.059	0.085	0.018	1.416
	B5	0.061	0.088	0.013	1.378
	B6	0.076	0.104	0.003	1.328
	B7	0.060	0.082	0.021	1.387
Perceived quality	C2	0.062	0.088	0.012	1.357
	C5	0.100	0.145	0.000	1.399
	C6	0.087	0.124	0.001	1.549
	C7	0.046	0.067	0.070	1.506

, the adjusted R-square is 0.542, which indicates that the explained variables account for 54.2 percent of the total variables. This means that three factors such as perceived value, platform experience, and perceived quality explain 54.2% of the variation in customer satisfaction. This result further supports the regression effect. In addition, the VIF values in the multiple regression analysis ranged from 1–5, indicating that there was no significant covariance between the variables.

As shown in Table 13, the standardized regression coefficients for the subscale items measuring perceived value, platform experience, and perceived quality demonstrate statistically significant positive effects (β > 0, p < 0.05). This suggests that perceived value, platform experience, and perceived quality have a significant positive effect on customer satisfaction, supporting the validity of hypotheses H1, H2, and H3.

5. Discussion and Implications

5.1. Main Effect

The findings of this study reveal that perceived value, platform experience, and perceived quality exert positive influences on customer satisfaction. This research offers empirical evidence that consumers conduct a comprehensive assessment of the value-for-money proposition associated with a product or service throughout their purchasing journey. In the context of the fresh food e-commerce sector, consumers exhibit concern not merely for the product price [81] but also for factors such as quality [2], freshness, nutritional value, and ancillary services like expedited delivery and superior packaging [82].

H1 posits that perceived value exerts a positive and significant influence on consumer satisfaction. It is observed that consumer satisfaction escalates when e-commerce platforms provide fresh produce that is both high-quality and cost-effective [83,84]. To remain competitive in the market, fresh food e-commerce platforms need to engage and retain consumers through promotional activities, membership programs, and enhanced brand development, thereby augmenting customer-perceived value [2]. Concurrently, enhancing the transparency of product information, including details on direct sourcing and organic certifications, is essential to ensure that customers perceive value for money.

Secondly, H2 verified that platform experience is an important factor influencing consumer satisfaction. This variable is determined by external variables such as system design characteristics, user characteristics, and task characteristics [85]. In the digital era, consumers are also increasingly demanding convenience [86], interactivity, and personalization in the shopping process [23,87]. Therefore, e-commerce platforms can improve consumers’ sense of ease of use and reduce barriers and inconveniences in the shopping process by providing customized service interfaces, fast order processes, and convenient shopping experiences. At the same time, personalized recommendation services are provided to push relevant products based on users’ purchase history and preferences to improve shopping efficiency and satisfaction [29,88,89]. In addition, companies should strengthen after-sales service [90], such as responding quickly to customer inquiries and providing convenient return and exchange services.

Finally, H3 confirms that strengthening perceived quality management is a key factor in improving consumer satisfaction. Surveys show that product quality and freshness are the two aspects consumers are most concerned about in the fresh food e-commerce space [91]. Consumers are more inclined to buy products with a strict quality control system [92], including whether the goods sold comply with relevant standards [93]. Strictly controlling the quality of products to ensure the freshness [94], safety, and taste of the fresh produce sold can better meet consumer expectations [95,96], and product quality can be ensured from the source by establishing its own supply chain system and cooperating with high-quality suppliers [97]. In addition, logistics and distribution capabilities should be strengthened to ensure that products are delivered to customers in a timely manner and in good condition [98]. At the same time, consumer satisfaction should be enhanced through continuous improvement and innovation.

5.2. Theoretical Implications

This study advances consumer satisfaction research by integrating classic satisfaction models (ACSI/ECSI/CCSI) with data-driven text mining, a novel synthesis that addresses the limitations of purely deductive or inductive approaches. Prior studies often rely on predefined theoretical constructs (e.g., ACSI’s perceived quality) or isolated text analytics of user reviews. This research innovatively integrates offline questionnaire data, where consumers highlight price sensitivity, with online review data, which reveals that logistics speed significantly influences satisfaction. Such integration not only augments this study’s comprehensiveness but also bolsters the reliability of the findings, offering a more nuanced direction for optimization in e-commerce platforms and assisting them in enhancing user experience to gain a competitive edge.

This study provides valuable insights and empirical evidence for future research in related domains. Firstly, the methodology of integrating multi-source data deserves further exploration and application, particularly in the realm of consumer behavior studies. By amalgamating data from various sources, researchers can attain a more holistic and intricate comprehension of consumer requirements and behavioral tendencies. Secondly, online user review analysis, as an emerging technique in data mining, holds substantial promise for application. This study utilizes big data analytics to scrutinize user reviews, underscoring the considerable potential of such data in understanding consumer behavior, identifying market trends, and refining operational strategies. Fresh food e-commerce platforms should capitalize on big data and artificial intelligence technologies to thoroughly extract user information, accurately identify customer needs, and offer personalized services along with targeted marketing. The investigation of its application scenarios and efficacy across different sectors remains a vital area for further research. Lastly, this study emphasizes the significance of bridging theory with practice. It is only through a robust integration of theoretical insights with empirical evidence that more scientifically sound and rational conclusions, coupled with actionable recommendations, can be formulated.

5.3. Managerial Implications

This study offers several insights into pertinent management practices. Firstly, it has been determined that perceived value is a pivotal variable in enhancing consumer satisfaction. Consequently, merchants should endeavor to augment consumers’ perceived value. By tailoring actions to product characteristics and platform features, merchants could implement targeted and persuasive measures, such as enhancing product quality and refining platform design, to fulfill consumers’ functional shopping needs and thereby boost their utility value. Additionally, merchants should bolster pre-sale and post-sale services, promptly address consumer issues, dispel doubts about the platform or products, and render shopping more comfortable and convenient, allowing consumers to derive psychological satisfaction from the shopping experience. Concurrently, efforts should be made to cultivate emotional value. For instance, by optimizing the response speed of pre-sales consultation (such as through intelligent customer service systems) and the after-sales service system (such as the worry-free return and exchange policy), the risk perception of consumers can be reduced, thereby enhancing their psychological utility.

Secondly, our analysis reveals that product quality and platform attributes are significant determinants of elevated consumer satisfaction. Consumers place the highest priority on the quality of fresh products, particularly their freshness and nutritional value. Consequently, companies must strengthen the supervision and management of the fresh product production process to ensure health and nutritional integrity, eschew artificial additives, and embrace environmentally sustainable production and packaging methods. In terms of the platform, information quality is paramount, succeeded by system and service quality. Enterprises should therefore refine platform page design, ensure the accuracy and timeliness of platform information updates, simplify the consumer search process, and introduce personalized features catering to various consumer groups. These measures are designed to augment customer satisfaction and loyalty, thereby conferring a sustainable competitive edge to the platform. Furthermore, it is necessary to meet the logistics of the emergency-handling demands of consumers. This requires the platform to establish an elastic supply chain mechanism. For instance, through regional warehouse layout and real-time logistics monitoring, the delivery time can be shortened. In addition, solutions should be proactively pushed out during abnormal events (such as weather delays) to maintain trust.

6. Conclusions

This study contributes significantly to our understanding of consumer satisfaction in the field of fresh food e-commerce by examining the role of perceived value, platform experience, and perceived quality. Our findings reveal a nuanced interplay between these variables, demonstrating that all three dimensions exert statistically significant positive effects on customer satisfaction (β = 0.414, 0.263, and 0.493, respectively; p < 0.01). By integrating multi-source data—including 509 validated questionnaires and 1252 online reviews—our framework bridges the theory–data dichotomy prevalent in prior studies, offering a replicable methodology for hybrid analysis in emerging markets.

By integrating data from multiple sources and sophisticated data-mining techniques (e.g., online user review analysis), we emphasize the importance of a holistic analysis of consumer behavior. Theoretically, this study emphasizes the need for future research to apply multi-source data methods and big data analytics to enhance the understanding of consumer needs and preferences and identify market trends. In terms of management, our study provides actionable insights: merchants should improve product quality, especially freshness and nutritional value, improve platform design, and enhance pre-sale and after-sale services in order to increase perceived value and enhance customer satisfaction and loyalty. In summary, our comprehensive framework integrates theoretical insights and empirical evidence to provide scientifically sound conclusions and feasible recommendations for researchers and practitioners in the rapidly evolving field of fresh food e-commerce. Future research should build on these foundations to further refine our understanding and develop innovative strategies.

The limitations of this study are manifested in three aspects: Firstly, the questionnaire design primarily focused on measuring “Product quality” and “Product prices”; therefore, other potential consumer concerns (e.g., data privacy, delivery flexibility) were not systematically explored. Secondly, the research conclusions are based on single-time-point data from the Chinese market (509 questionnaires and 1252 online comments), and the limited volume of online reviews may constrain the robustness of text-mining results, as larger datasets are typically recommended for such analyses. Additionally, their cross-cultural universality needs to be verified through comparisons of samples from multiple countries. Finally, the long-term transmission path of enterprise innovation mechanisms on satisfaction has not been fully tracked. Future research should (1) construct dynamic panel models to identify time-varying effects; (2) expand to emerging markets (e.g., Southeast Asia) and mature markets (e.g., the European Union) for cross-regional comparisons; and (3) explore collaborative evolution mechanisms between organizational innovation and consumer behavior changes.