This paper considers the problem of production scheduling in a real-life manufacturing plant belonging to the apparel industry. The problem is modeled as a two-stage flexible flow shop, with minimization of both makespan and number of tardy jobs. For this new scheduling problem, we propose a bi-objective evolutionary algorithm. A set of experiments is performed using real data from the company’s database. Experimental results show the relevance of the proposed procedure in terms of performance metrics, as opposed to the current manual scheduling procedure as well as against available commercial schedulers. The procedure was actually implemented as the scheduler procedure at the company. Such results also illustrate the improvement in key performance metrics of the production line.