Affect-as-Information: Customer and Employee Affective Displays as Expeditious Predictors of Customer Satisfaction

This study introduces affect-as-information theory to the service encounter, integrates it with the peak and end model of affect, and thereby shows that these dynamic customer and employee affective displays can be used to estimate post-encounter customer satisfaction. A large-scale dataset of 23,645 real-life text-based (i.e., chat) service encounters with a total of 301,280 genuine messages written by customers and employees were used to test our hypotheses. Automatic sentiment analysis was deployed to assess the affective displays of customers and employees in every individual text message as a service encounter unfolded. Our findings confirm that in addition to customers’ overall (mean) affective display, peak (i.e., highest positive or least negative), and end (final) affective displays explain customer satisfaction. Further, as customer displays may not fully capture their satisfaction process and employees understand the service quality they deliver, we propose and confirm that employee displayed affect explains further variance in customer satisfaction. We also find that the predictive power of affective displays is more pronounced in service failure than non-failure encounters. Together, these findings show that automatic monitoring beyond customer overall affect (i.e., adding customer peak and end, and employee affective displays) can expedite the evaluation of customer satisfaction.