"Only one drive manufacturer had enough failed drives in the dataset to produce a large enough subset, which we’ll refer to as Manufacturer A. Because this data

"Only one drive manufacturer had enough failed drives in the dataset to produce a large enough subset, which we’ll refer to as Manufacturer A. Because this data treatment was so effective, and because most of the drives that had been in our fleet long enough to begin failing were already from Manufacturer A, we decided to reduce the scope of smarterCTL to only their drives." I'd have dropped Manufacturer A from the supplier list and used the A only model for the remainder of their drives. Then I'd have another go with SmarterCTL for "not A".