Nowadays, mobile recommendation has become an important research topic in data science. While many researchers focus on developing new applications and designing algorithms for computational efficiency in different business areas, some significant technical problems in classical algorithms are rarely studied under these applications. Simulated annealing (SA), a key approach in solving global optimization problems, is one of them. To this end, this paper aims to investigate the performance of SA in mobile recommendation problems with a focus on identifying the optimal cooling schedule method. We also discuss the move generation, parameter estimation, and the balance between efficiency and effectiveness in SA. Specifically, our tests are based on two problems: a travelling salesman problem and a mobile route recommendation problem. The results suggest that the exponential-based method performs the best to achieve the optimal final energy, while the greedy method, constant-rated-based method, and logarithm-based method are dominant in terms of computational efficiency. Our studies would serve as a guidance of SA for mobile recommendation algorithm designs, especially for the selection of cooling schedule and related parameter estimation.