When one applies SPC to the manufacture of something, it has to start somewhere. SPC starts when we begin to collect data. So begin to collect data on the set up process parameters that are meaningful in understanding the variation and effectiveness of the set up.
"Set ups" are a process. SPC and statistical methods can be applied to any process. Just because the "job" hasn't started yet in terms of traditional manufacture, a process is running. This process is called "set up" and the standard is requiring users to use statistics, where applicable, to understand the effectiveness and variation of the set ups. The intent here is that if the variation and the effectiveness of the set ups is understood and controlled, then the "job" will produce product that meets requirements more often.
This is the Process Approach as applied to the process of Set Ups.
One could look at the variation in set up time, first pass success, variation in any variable data that is collected at set up, etc. Do you want to know how effective the set up is the first time? What if set ups were not approved on the first try and in general it takes 3 tries to get a set up approved? Does the initial output of the set up, which is used to approve the production job, meet requirements? How often? How much variation from the mean or target value? How does the variable data that is collected vary within the control limits? Is the data always above the target value? Below? Is there drift over time in the data collected? Can you prevent nonconforming material by understanding the variation over time of the data collected at start up? Sure!
If the data is an attribute and not variable, you could use a n, np, u or c chart to understand the variation in the attribute.
Side note: I never understood why using any statistical method like a bar chart or a pie chart to control a process is not "SPC." Some say that these statistical techniques are not necessarily "SPC." Maybe it is because one can use a statistical technique like a bar chart for things other than trying to control a process. And maybe when one applies any statistical technique or method to controlling a process it is "SPC." Anyway, 7.5.1.3 requires the use of statistical methods which may or may not be traditional SPC.