In this work, a real world milk collection problem is introduced. A milk processing company, located in south of Chile, needs to collect milk to elaborate their products. To this work, the company has a heterogeneous truck fleet. The farms are scattered in a large geographical area. In this problem, the company buys only the minimum quantity of milk required to its daily operation. Exceeding this minimum quantity of milk generates a cost of over demand. The objective is to find efficient collection routes, satisfying the minimum required quantity and minimizing the over demand and transportation costs. To solve this version of the prize collection problem we propose an integer programming model to solve simpler instances and a GRASP metaheuristic to solve more complex instances in reduced time. Real world problem instances can consider up to 500 farmers. We tested our approaches using small real world cases and possible expansion scenarios. We concluded about the key components of our approaches and their capabilities to solve the problem at hand.