Real-world routing problems are often represented by large and complex models, and instances of realistic size are very hard to solve. In most cases one cannot find optimal solutions, but has to use heuristic solution methods to find good solutions. It is, however, also important to be able to solve smaller sized instances to optimality for comparisons with heuristic methods. In this paper, we show that small instances of a real-world routing and inventory problem, the so-called Livestock Collection Problem, can be solved to optimality by using a column generation approach.