During a product's design phase, some decisions may be taken based on previous experiences and empirical assumptions. This study's target is to bring light to this grey area. For that, a quick and fully automated DOE technique is applied through a FEA simulation to a medium size car radiator. Objective is to give, not only answers to our main queries, but also a whole map of how design variables impact the global structural performance of the component. The most important durability test (cycled pressure) and its typical mode of failures (header's opening and groove crack) were approached. To avoid these premature failures, different solutions have been proposed worldwide throughout time, but which are really efficient is something much more difficult to define. For that reason, 7 well known design features were selected and based on them, over 2000 different combinations were tested in a full factorial experiment that provided some quite amazing conclusions. By choosing wisely the input parameters, stress could be significantly reduced. Surprisingly, some of the largely implemented solutions would actually work in the reverse direction, by jeopardizing the results, while the most effective is actually rarely used. At the end, a validation case is also set in order to confirm the DOE achievements. In global terms this study shows how naïve we may get in our everyday extinctive decisions and how powerful a Design of Experiment may be in order to lead us in the good direction by providing a rich map of the problem.