I don't know how many variables there are, but the most critical are on the front end. then through the process, there is another set. keep in mind, that no matter where the variables lie, there are downstream consequences for all of them.
variables must be adjusted for, and all adjustments have down steam consequences. all controls must therefor be predictive in nature. therein lies the bug for training. the operator must be able to use predictive judgment at all times and it takes a long time to become proficient.
software eases the burden, but it comes from experience too. it takes a long time to experience enough variations to improve software to the point that it truly aids the operator. even then the operator must have the skills to predict and handle the software shortfalls.
I had people working with machines for 20 years that never attained a proficiency equal to mine. at one point early on I was concerned that I was surrounding myself with too much incompetence but after some analysis of my own capabilities, I found that I was exceptionally gifted in this area and that I had normal people and had to adjust my approach to training and expectations.
I would expect that jbi will never find an operator with john's capability at predictive adjustment, thus requiring a permanent ongoing software improvement program. improvement never ends. it is never finished.