Caller abandonment could depend on their past waiting experiences. Using Cox regressions we show that callers who abandoned or waited for a shorter time in the past abandon more in the future. However, Cox regression approach does not shed light on callers’ prior belief about the duration of their delays. Moreover, Cox regressions cannot separate the impact of callers’ parameters such as their waiting costs on their abandonment behavior from the impact of their beliefs about their delay durations, which are affected by their past waiting experiences. To tease out the impact of callers’ waiting experiences on their abandonment behavior, we use a structural estimation approach in a Bayesian learning framework. We estimate the parameters of this model from a call center data set with multiple priority classes. We show that in this call center new callers who do not have any experience with the call center are optimistic about their delay in the system and underestimate its length irrespective of their priority class. We also show that our bayesian learning model not only has a better fit to the data set compared to the rational expectation model in Aksin et al. (2013), Aksin et al. (2017) and Yu et al. (2017) but also outperforms the rational expectation model in out-of-sample tests. In addition, our bayesian framework does not lead to biased estimates, which would happen under the rational expectation assumption if callers’ belief about their waiting durations does not match their actual waiting time distribution. Our bayesian framework has managerial implications at both tactical and operational levels such as managing customer expectation about their delays in the system, and implementation of patience-based priority policies such as Least-Patience-First and Most-Patience-First scheduling.