Trying to answer industrial challenges regarding production planning and control, Ptak and Smith propose a new method called Demand Driven MRP. Their basis is to cope with variability instead of facing it. Amongst the different key points of the method, buffer location choice is not clearly defined. We aim to investigate its impact on the industrial performance. This study will rely on a design of experiments led with a discrete event simulator. It includes an in vitro case study with 7 different buffers placements and three types of variability (customer, supplier and process) with 3 levels each. The results show that the choice concerning buffer placement can affect the performance up to 15 point of OTD and 100% of Working capital. They also demonstrate that there is no buffer placement that would allow dealing with all types of variability and levels. Buffer placements policy has to be adapted to variability context to reach a high level of performance.