Nowadays, two different methodologies are used to address the CMP issues. On one hand, we find basic design oriented methods consisting of reaching a minimal density of geometries in the design. On the other hand we find model based approaches in which complex process related parameters are used. This makes these techniques, either not accurate or not usable by designers. In both cases there is no efficient monitoring of the CMP effect through Process Control Modules. This paper presents a new methodology to improve CMP process yield from the designer side. A prediction function of metal thickness variations due to CMP is established thanks to specific test structures. A method to monitor the CMP process evolution at no cost is presented and finally a technique for using the prediction function to drive metal filling procedure is described.