4.1. Data Collection and Preprocessing
Ele.me and Meituan are the two largest O2O takeaway platforms in China, and their market share exceeds 90% of China’s takeaway market. Relevant information (“Gender equality and women’s empowerment”
https://baijiahao.baidu.com/s?id=1594344770944521799&wfr=spider&for=pc (accessed on 5 June 2021). “2017–2018 Research Report on China’s Online Food Delivery Market”.
https://www.iimedia.cn/c400/60449.html (accessed on 3 November 2024). “Analysis of the Market Size and Competitive Landscape of China’s Food Delivery Industry in 2021: Two Major Advantages of Douyin Entering the Food Delivery Market”.
https://bg.qianzhan.com/report/detail/300/211223-4098960b.html (accessed on 3 November 2024).) shows that the market share of “Ele.me” in the fourth quarter of 2017 accounted for 55.3%, and “Meituan” 41.3%. By the first quarter of 2021, the market share of “Meituan” reached 67.3%, while “Ele.me” fell to 26.9%, descennding to the second largest in the industry. The sharp decline in Eleme’s market share is closely related to the loss of riders. To explore the causes, this paper looks at RTCL riders’ satisfaction on the “Ele.me” platform by analyzing online reviews.
Different from the data sources of previous online review studies, the data for this study comes from the online review data of the “Rider community” on the “Ele.me” platform, rather than the user reviews on the Appstore. The data collection format is shown in
Table 2. The RTCL riders’ online reviews provided by the official platform are not only standardized and objective, but also reflect the increasing attention paid to the needs of RTCL riders by the “Ele.me” platform.
The online reviews selected for this study include review content, review time, and review likes. After removing the duplicated data and irrelevant data in the original data, we use the gender prediction method proposed in
Section 3 to classify online reviews by gender of RTCL riders. A total of 446 women rider reviews and 927 men rider reviews were obtained. The specific information is shown in
Table 3. From
Table 3, it can be seen that 1373 online reviews correspond to 8438 likes, and a high number of likes indicates that the online reviews are representative. Moreover, on the one hand, according to the 2023 Ele.me Rider Rights Protection Report, the proportion of women riders on the “Ele.me” platform is only 7.5% (“Ele.me Establishes the ‘Blue Rose Women’s Federation”,
https://news.sohu.com/a/808459157_362225 (accessed on 21 October 2024)), but the data in
Table 3 show that the number of reviews made by women riders has reached half of the reviews made by men riders. This reflects the improvement of women riders’ awareness of right safeguarding, which is progress on gender equality trends. On the other hand, the number of men riders are more than 12 times as many as that of women riders, which reflects those men riders, as the leading participants in the RTCL service of the “Ele.me” platform, have a deeper understanding of the RTCL service of the platform, and their needs are more likely to resonate with the rider group.
This study collects in-depth interview data from riders, aimed at providing a more comprehensive understanding of satisfaction factors including platform image (), rider expectations (), service perceived quality (), product perceived quality (), and perceived value (). Open-ended responses in these interviews are expected to reveal detailed perceptions that complement the insights obtained from online reviews. By incorporating interview data, this study aims to enhance the understanding of insights gained from review data, allowing for a richer, more nuanced perspective on satisfaction factors. Preliminary analysis indicates that the results from interviews are largely consistent with those derived from review data, which strengthens the reliability of our findings on rider satisfaction.
To address potential biases in online review data, this study applied several data-cleaning techniques to ensure data reliability. First, removed outliers by filtering comments with extremely high or low ratings. Then used time-balanced sampling to distribute reviews evenly across the study period, minimizing the impact of temporary events. Finally, conducted sentiment analysis to identify and adjust for emotionally extreme reviews. These steps helped to reduce selective reporting bias and ensure a more representative dataset.
4.3. Calculation of Index Weights
Making convenience of calculation, we integrated the satisfaction data of riders’ online reviews on the “Ele.me” platform, and the results are shown in
Table 4 and
Table 5. Firstly, Freq-AHP is used to calculate the weight of the index. The maximum eigenvalue λmax of each index is calculated by the judgment matrix. Secondly, the consistency index CI is calculated by Equation (1), and then Equation (2) is used to test the consistency of the judgment matrix, and a CR value is obtained. The results are shown in
Table 6, so the judgment matrix passed the consistency test. And the cross-validation results confirmed that the model’s accuracy remains stable across different data splits, which reduces the risk of overfitting and enhances the generalizability of our findings. Finally, MATLAB R2024b software is used to calculate the weights of each index, and the weights calculated by Freq-AHP are shown in
Table 7 and
Table 8.
Next, we use the PLTSs to calculate the weight of the RTCL riders’ satisfaction index. Firstly, we list the RTCL rider’s PLTSs and normalize them by Equations (3) and (4), and then we use Equation (14) to calculate the normalized PLTSs weights, and the weight information is displayed in
Table 7 and
Table 8. Because the results of Zhang et al., (2016) show that Freq-AHP and PLTSs are important, we use Equation (19) to get the weights for both [
38].
The comprehensive weight of the indexes objectively reflects the importance of the needs of the RTCL riders of “Ele.me”. By analyzing
Table 9, we obtain the ranking of the need’s importance of women riders and men riders, which is shown in
Table 10 and
Table 11 respectively.
From
Table 10 and
Table 11, it can be seen that there is a significant difference in the importance ranking of perceived value and service perceived quality. women riders pay more attention to perceived value, that is, more attention to delivery unit price and dispatch quantity than to platform’s reply feedback and rider complaints. On the contrary, men riders pay more attention to the platform reply feedback and rider complaints than the delivery unit price and dispatch quantity. Maslow’s of needs says that job security conditions like delivery unit price and dispatch quantity are security needs, belonging to low-level needs. Platform feedback and rider complaints are higher-level respect needs, and the satisfaction of respect needs can have an effect on the morale of the men riders.
In terms of perceived value, women riders pay more attention to the dispatch quantity than men riders. This is related to the speed and skills of taking orders. The high-frequency interaction between groups enables men riders to master more order-taking skills, while women riders’ lack of group interaction makes it more difficult to take orders, thus increasing the attention of women riders to the dispatch quantity. men riders pay more attention to the delivery unit price which is directly related to income. As for service perceived quality, women riders pay more attention to the platform’s reply feedback, indicating that women riders pay more attention to the platform’s service attitude, while men riders pay more attention to rider complaints, because they often need to appeal more in disputes with merchants and consumers. Therefore, they are more concerned with rider complaints.
In terms of platform image, product perceived quality, and rider expectations, the importance ranking of men and women riders is consistent, whereas vary significantly in importance degree. In terms of platform image, the attention of women riders is remarkably lower than that of men riders. As specifically reflected in the weight of the tertiary indexes, women riders pay more attention to the platform popularity while men riders pay more attention to the platform reputation such as timing rules, dispatch distance, reward and punishment standards, and other services closely related to their own interests, which indicates men riders’ pursuit of service practicability. In terms of product perceived quality, the attention of women riders is higher than that of men riders. In the weight of the tertiary indexes, women riders pay more attention to system fluency, while men riders pay more attention to system optimization. As the RTCL industry is a labor-intensive industry, the advantage of physical fitness should have played a key role, but the unreasonable rules for dispatching orders make it difficult for men riders to give full play to their physical advantages. In terms of rider expectations, the attention of women riders is higher than that of men riders. In the weight of the tertiary indexes, their high attention to humanistic care reflects that they pay more attention to spiritual pursuits. Their low attention to consumer attitudes reflects the differences in personality between women and men riders. women riders have a relatively mild attitude, which makes consumers more friendly. Therefore, they have fewer disputes in the communication process.
4.4. RTCL Riders’ Satisfaction Evaluation
According to the comprehensive weights shown in
Table 9, this paper adopts FCE to calculate the RTCL riders’ satisfaction.
Firstly, we set the target layer as RTCL riders’ satisfaction and the criteria layer as the secondary index .
= {Platform image, Rider expectation, Product perceived quality, Service perceived quality, Perceived value}.
The scheme layer is set to the tertiary index .
= {Platform popularity, Platform reputation},
= {Humanistic care, Merchant attitude, Consumer attitude, Rider income},
= {Platform reply feedback, Rider complaints},
= {Software function, System optimization, System fluency},
= {Delivery unit price, Dispatch quantity}.
Secondly, we establish the evaluation set .
= {women rider expectation } = {Very unsatisfied, Unsatisfied, General, Satisfied, Very satisfied}.
Thirdly, we construct the fuzzy evaluation matrix, and the fuzzy evaluation matrix
by taking the as an example.
According to
Table 8, the weight of women rider expectation is w
2, w
2 = (0.337, 0.2077, 0.2058, 0.2496)
Therefore, the comprehensive evaluation result of the women rider expectation is obtained by calculation as
.
Using the same method, the comprehensive evaluation results of other indexes can be obtained and shown in
Table 12.
Then, we use the comprehensive evaluation results of the secondary indexes to construct the fuzzy relation matrix R and R’ of women riders and men riders respectively.
According to the comprehensive weights w = (0.2681, 0.1677, 0.1683, 0.2238, 0.1723) and w’ = (0.315, 0.1484, 0.1672, 0.2078, 0.1617) of the women and men riders’ secondary indexes, we can calculate the comprehensive evaluation results SC and SC’ of the women riders and men riders satisfaction respectively, and the results are shown as follows.
According to the maximum membership principle of FCE [
46], the maximum membership degree of women riders and men riders’ satisfaction is 0.6074 and 0.7617, respectively, and the corresponding evaluation levels are both “very unsatisfied”. Therefore, this paper demonstrates that the overall satisfaction of women riders and men riders under the maximum membership principle is “very unsatisfied”. 0.6074 < 0.7617 shows that men riders are more dissatisfied. In addition, according to the weighted average principle, the satisfaction scores of women riders and men riders are 32.196 and 28.308 respectively, which are between “very unsatisfied” and “unsatisfied”. 28.308 < 32.196 also shows that the overall satisfaction of men riders is lower than that of women riders, which is consistent with the results of rider satisfaction calculated by the maximum membership degree principle. Also, we can get the satisfaction information with all of the different indexes based on the maximum membership principle and the weighted average principle, and the results are shown in
Table 13.