The previous results included time to complain as a feature; again, this is the time lag between the last QC reading by the system and the time when a customer placed a complaint (Figure 1). This is unrealistic, as in a real application, we would not know when a customer is going to complain. As a result, we assessed the impact of removing the time to complain feature from our classifiers. Both CART and adaptive boosting were affected by this removal, even if all five assays still had median error rates < 50% and as low as 20% with adaptive boosting (Multimedia Appendix 5). This increase in error rate after removal of this feature shows that time to complain is an important determinant of a complaint, which in turn suggests that customers are quick to complain after detecting a QC shift.Note, however, that this removal of the most important feature did not affect the relative importance of the other features: those involved in the timing of maintenance events and those describing the variability of concentrations (SDs) were still the most important predictors (Multimedia Appendix 7).