Aiming at multi-product two-stage serial lines with batch machines, the impact of different scheduling policies on system performance is studied, performance indicators such as productivity and work-in-process level are analyzed analytically. First, the system state space is described for four different scheduling policies, namely highest upstream machine efficiency policy, highest downstream machine efficiency policy, longest queue policy, and cyclic policy. The analytical model is established based on Markov process to quantify the performance indicators such as productivity. Then, considering the relationship between buffer capacity and batch capacity, the impact of four scheduling policies on system performance is compared through numerical experiments. The results show that when buffer capacity and batch capacity are independent, the cyclic policy has better performance than the other three policies. However, when there is a positive correlation between buffer capacity and batch capacity, the longest queue priority policy has the best performance.