Measurement based software process improvement needs a non-intrusive approach to determine what and where improvement is needed without knowing anything about the methods and techniques used during project execution. Beside, it is necessary for obtaining successful business management, an accurate process behavior prediction. In order to obtain these results we proposed to use Statistical Process Control (SPC) tailored to the software process point of view. The paper proposes an appropriate SPC-Framework and presents two industrial experiences in order to validate the framework in two different software contexts: recalibration of effort estimation models; monitoring of the primary processes through the supporting ones. These experiences validate the framework and show how it can be successfully used as a decision support tool in software process improvement.