Typical testing of solder joints, (e.g. joints at printed circuit board pads) has not proven totally predictive of the ultimate performance of such joints. It has been found that this lack of reliability is, at least in part, due to the tendency during testing for these pads to lose adhesion to, or delaminate from, the underlying substrate. In contrast, such occurrence is not typical of phenomena induced during typical device usage. To remove this source of unreliability, a test structure is made together with the manufacturing device lot. The same pad processing is used and the pad size is substantially enlarged in the test structure. The test structure is employed to predict performance of devices in the lot and then the lot is processed accordingly.