Step 1 Select between 20 to 30 test samples that represent the full range of variation encountered in actual production runs. Practically speaking, if “clearly good” parts and “clearly bad” parts are ...