Reputation mechanism is critical for a distributed service environment. A customer can select the optimal services to compose a value-added composite service based on services' reputation values. After executing (either successfully or unsuccessfully) a composite service, a customer gives a feedback to the composite service. Then the composite service needs to propagate perception of customer's feedback to each of the component services. However, how to fairly distribute the customer's perception to the component services is a challenging issue, as each of its component services may perform differently in contributing to the success or failure. In this paper, we present a QoS-aware approach for reputation propagation to address this problem. The proposed techniques enable a composite service to propagate customer's perception to its component services based on their actual performance in the latest execution and their services reputation trends over a period of time. Experiment results show that our methodologies can effectively and fairly distribute customer's perception from a composite service to its component services.