.::NeO::.: you are forgetting possibility that only same small part/batch, eg. 5-10% may fail even with rumored bad plating. So you have to test big pile of block samples, preferably from different batches/times. It rises testing load/work/time substantially, even with parallelizing process.