One of the key dilemmas faced by all service providers is how to trade off between the high quality of services and timely responses. When either feature is lacking, customersrsquo retrial occurs ndash a calling back behavior for addressing same requests in the previous call. We classify retrial into the congestion retrial where customers abandoned the previous call due to a slow system and the fitness retrial where customers received dissatisfactory services in the previous call due to poor quality. In this paper, we study retrial by connecting customers behavior with their preferences for service aspects: the speed in service access and the quality of service delivery. Based on a call-by-call dataset from a multi-skill service system, we use a random-coefficient dynamic structural model to quantify customersrsquo preferences for speed and quality across different customers segments. Interestingly, we found that business customers have a stronger preference for service speed and are less sensitive to quality compared to private customers. Given the distinct preferences across customers segments, we suggest two economically viable strategies to reduce retrial. First, we suggest by wisely allocating the current service groups along the timeline, the surplus can be increased by 37.9% for business customers and 18.2% for private customers. The second approach suggests customersrsquo surplus can also be improved by expanding the service team with cheap labor source, the general service agents. However, service providers should be aware that adding such agents are not always beneficial to the customers since it means to trade good quality off for a quick response.