Recently online advertisers utilize Recommender systems (RSs) for display advertising to improve users’ engagement. The contextual bandit model is a widely used RS to exploit and explore users’ engagement and maximize the long-term rewards. However, the current models aim to optimize a set of ads on only a specific domain and do not share information with other models for multiple domains. In this paper, we propose dynamic collaborative filtering Thompson Sampling (DCTS), the novel yet simple model to transfer rewards among multiple bandit models. DCTS adds similarity-based terms as a prior distribution of Thompson sampling. Similarities are obtained based on the contextual features of both ads and users. Similarities enabled models in a domain that didn’t have much data to converge more quickly by transferring rewards. Moreover, DCTS incorporates temporal dynamics of users and can track the user’s recent change of preference. We first showed transferring rewards and incorporating temporal dynamics im- proved the performance of the base model on a synthetic dataset. Then we conducted empirical analysis on a real-world dataset among two domains and DCTS improved click-through rate by 9.7 % than a state of the art model. We also analyzed hyper-parameters that adjust temporal dynamics and similarities.