A comparative paradigm of sequential mediation models of influencers’ credibility using PLS-SEM with machine learning

Companies employ social media influencers SMIs due to the compelling evidence of their advertising effectiveness. However, for optimal selection of SMIs, more research is needed to assess and compare the effect of different types of factors as in product-related or audience-related in SMIs advertising. To do so, we investigate the effect of source influence SI of SMIs on the intention to purchase I2P through the sequential mediation of parasocial relationships PSR with audience benign envy BE (audience-related) and PSR with brand–influencer fit BIF (product-related). Two independent samples (N = 411; N = 355) are used to perform: (i) PLS–SEM analysis to obtain the model’s predictive power, and (ii) classification–based machine learning ML to evaluate the model’s accuracy. A comparative paradigm is proposed based on within–study and between–studies comparative analyses. Both samples indicate a high role of trustworthiness and expertise in forming SI. Comparative mediation analysis and predictive accuracy scores show that BE as an audience–related feature, plays a more vital role in affecting/predicting the followers’ I2P compared to the brand-related feature, BIF. The results of comparative analysis contribute to the knowledge of SMIs’ credibility and provide a better understanding of SMIs’ selection strategies for marketing practitioners and researchers.