Be aware of product problems and defects before lab results by modeling the differences between good and bad product, then put the models on-line and real-time trend the probability of defects and be alerted when that probability crosses a threshold.