A Dynamic Hidden Markov Model Of The Online Consumer Repurchase Behavior

To increase the repurchase rate in online shopping,merchants often choose to release two signals: the frequent product promotion and the loose return policy.Therefore,it is very important for merchants to understand how these signals impact the motivating process of the repurchase behavior.In the prior literature,consumermerchant relationship is considered to be a key psychological factor affecting online repurchase intention,and a lot of research has focused on this.However,there are few studies based on second-hand data to explore the impact mechanism of merchants' actual actions on the repurchase behavior.Moreover,as far as the frequent promotion and the loose return policies are concerned,previous studies have also differed on their effectiveness.In order to clarify the specific mechanism in which returns or promotions may have different effects,this work proposes a multi-stage hidden Markov model(MSHMM)from the perspective of the signal theory.In the hidden Markov model,the consumer-merchant relationship can be represented as the latent state.The state transition is based on the state transition probability matrix,and the corresponding observation behavior(i.e.,repurchasing behavior)is ultimately generated by the latent state.This work assumes that the impact of returns and promotions on the repurchase is time-phased: on the one hand,past returns or promotions will affect the consumermerchant relationship,and thus have a subsequent impact on the current repurchase behavior;on the other hand,the perceived return or promotion signals released by the merchant will directly impact the current purchase behavior.In addition,this work introduces the concept of the retailer's market expansion,assuming that in different market expansion stages,returns and promotions will have different effects.That is,the dimension of stage heterogeneity is introduced to discuss the impact of returns and promotions in real-world business.Finally,the work carried out extensive experiments based on real-world data,and verified that the proposed MS-HMM is able to capture the impact of returns and promotions at different stages of different customer groups.Through the performance comparison,the effectiveness of the stage-heterogeneity was verified.As for the theoretical contribution,this work is based on the signal theory and provides new insights into how signals can differentially influence the formation of the consumermerchant relationship and repurchase choices.In addition,by incorporating the concept of market growth into the original signal theory,this paper extends the applicability of the differences in the signal influence.Moreover,the research results of this paper can provide practical support for the resource allocation of e-commerce market retailers.