We present a new mathematical model for a single machine scheduling problem originated from the tile industry. The model takes into account the sequence-dependent setup times, the minimal batch size, heterogeneous orders of customers, and a stock in storage. As the objective function we use the penalty for tardiness of the customers’ orders and the total storage cost for final products. A mixed-integer linear programming model is applied for small test instances. For real-world applications, we design a randomized tabu search algorithm. The computational results for some test instances from a Novorossiysk company are discussed.