The goal of this thesis is to investigate what compensation level is effective to retain customers’ repurchase likelihood in different service failure conditions. Specifically, this thesis examines how magnitude of failure and two causes attributions: stability and locus of responsibility influence the effects of compensation. Additionally, we employed Discrete Choice Experiment (DCE) to quantify compensation amount to figure out the optimal compensation level for different service failure conditions. The results provide support for the interaction effects of magnitude of failure, compensation and stability/locus of responsibility in both airline and hotel contexts, such that for a high (low) magnitude service failure, high compensation will lead to higher customers’ repurchase likelihood when the failure is unstable (when the company is not responsible for the failure) than when the failure is stable (when the company is responsible for the failure). These findings provide implications for companies to effectively recover service failures using compensation.