管理工程学报2026，Vol.40Issue(1)：145-160,16.DOI:10.13587/j.cnki.jieem.2026.01.010
电商同理心能补救消费者流失的满意度吗?
Can e-commerce empathy remedy lost consumer satisfaction?
摘要
Abstract
The impact of the COVID-19 pandemic on fresh food e-commerce has led to the loss of consumers,and e-commerce empathy is regarded as an important means to service remedy.Some"robot"retailers that have no customer care and deal with problems indifferently will inevitably lose consumers in this process,while the humanized support can help to close the psychological distance between consumers and them.In order to explore the connotation and effect of e-commerce empathy and break through the possible bottleneck of"comfort",the paper takes the online reviews of users on Jingdong fresh e-commerce platform as research data,and uses the LDA topic model to explore the characteristics of e-commerce empathy.By constructing a quasi-natural experiment and a sentiment analysis model based on Word2vec,the effectiveness of e-commerce empathy on consumer satisfaction during the emergency period and the post-pandemic stage is evaluated.Meanwhile,the Social Network Analysis(SNA)is used to find specific causes of bottlenecks. Specifically,in terms of data collecting,the paper takes Jingdong Fresh,one of the top five fresh e-commerce platforms in China,as the data acquisition source.In order to reflect the phase characteristics of the epidemic and epidemic prevention and control,web crawler technology is used to collect four data sets of fresh product reviews.The first is the exploration of e-commerce empathy characteristics(data1 from January 2020 to January 2023),the second and third are the test of comfort effects in the emergency period(data2 from September 2017 to March 2022)and post-pandemic stage(data3 from March 2022 to January 2023),as well as data4 for training the sentiment model.Data preprocessing methods such as word segmentation,stop words removal,and feature selection are used to get clean text. For characteristic insights of e-commerce empathy,this paper identifies emotion words based on the service quality model(e.g.,SERVQUAL)and the interpretation of empathy in the past literature and extends them by using word2vec to filter the e-commerce empathy emotion lexicon.Building on this,the paper uses the coherence index to find the best number of topics,and the LDA topic model is employed to perform cluster analysis on online reviews,so as to obtain the characteristics that can reflect e-commerce empathy during the pandemic period(the emergency period and post-pandemic stage),which are taken as the important basis for experimental grouping.The sample of e-retailers with e-commerce empathy is the experimental group,otherwise,it is the control group. Regarding the comfort effect test,the paper compares classification algorithms based on traditional machine learning(SVM)with those based on Word2vec-SVM and selects a distributed sentiment analysis model with better predictive accuracy(Precision,Recall,and F1 score)for vectorization of online reviews,and obtains the satisfaction degree of consumers.The point in time of the outbreak and the characteristics of e-commerce empathy are used as the basis for experimental grouping.Meanwhile,the net effect and the possible bottleneck of e-commerce empathy are tested by using differences in differences(DID)method and quantile regression.If there is a non-significant point in the quantile regression,then it indicates that there is a bottleneck in e-commerce empathy and identifies the point as a threshold. Finally,given the possible comfort bottleneck of e-commerce empathy,the non-significant point in the quantile regression is used as the threshold to filter out the corresponding reviews.The centrality tests(degree centrality,betweenness centrality,and closeness centrality)are carried out by using the SNA method based on the cooccurrence matrix to explore the ability of correlation,control,and independence among nodes.Degree centrality reflects the direct causes of consumer dissatisfaction,betweenness centrality reflects the hidden causes of the bottleneck and the corresponding service remediation techniques,and closeness centrality reveals the details of the consumer dissatisfaction.Based on these,it is possible to know how consumer demands merge and resonate in the network,which can find the specific reasons for their complaint and the key to breaking through the bottleneck and retaining consumers during the emergency period and post-pandemic stage. The results of this study conclude that 1)E-commerce empathy is an e-retailer's ability to put themselves in the shoes of consumers and understand their emotions and feelings.It can be manifested in five aspects:logistics fulfillment,humanized delivery,health commitment,adherence during the pandemic,and customer care.2)E-commerce empathy can effectively alleviate the loss of consumer satisfaction during the COVID-19 period,which is more effective for consumers with low satisfaction and during post pandemic period.However,there is a"comfort bottleneck"during emergency period.3)Further social network analysis shows that the"bottleneck"mainly stems from the experience gap of consumers when they compared their bad experiences in the pandemic times with the normal e-commerce services,which can be repaired through economic compensation(e.g.,offering complimentary gifts and improving the product's cost-effectiveness)and humanized solace(e.g.,pre-cleaning of fresh food). The paper provides a new exploration based on different-period event scenarios and a mixed research paradigm(econometrics and text mining)outside the laboratory in the context of service remedy.关键词
电商同理心/服务补救/多阶段场景/准自然实验/LDA/在线评论Key words
E-commerce empathy/Service remedy/Multiple stage/Quasi-natural experiment/LDA/Online review分类
管理科学引用本文复制引用
杨益兴,吴刚,陈兰芳,郭茜..电商同理心能补救消费者流失的满意度吗?[J].管理工程学报,2026,40(1):145-160,16.基金项目
教育部人文社会科学研究基金(15YJA790064) （15YJA790064）
成都市软科学基金项目(2020-RK00-00071-ZF) The Humanities and Social Science Fund of Ministry of Education of China(15YJA790064) （2020-RK00-00071-ZF）
The Chengdu Soft Science Research Project(2020-RK00-00071-ZF) （2020-RK00-00071-ZF）
