Generally it’s determined on the output side. Either longer case life and/or better accuracy down range through more consistent case neck tension (bullet hold).
However, measuring what’s actually happening to the brass during the annealing process is a bit of a black box.
Using these statements from
@Nature Boy to explicate my own frustration, which seems particularly apt for this specific engineering/design project the OP is pursuing:
In process design, especially in continuous improvement, we have two types of “measures” of process performance which we can identify as “key performance indicators.” These are inherently linked to what we also call “QA/QC” - Quality Assurance and Quality Control - which are bifurcated for a reason: Quality Assurance includes process measures which “assure” we will get the quality outcome we desire, while Quality Control includes measures which “control” the quality of product we let out the door. This is the difference between “Lead Measures,” and “Lag Measures.” Lead measures are things we measure to know, up front and in process, we are “making things the right way”. Lag measures are things we measure on the back end to determine if it was “made right”.
In the instance of annealing, we have multiple indirect Lag Measures - sizing force reduction and consistency, seating force reduction and consistency, brass hardness, velocity consistency, group size, brass neck/shoulder life, etc, but we have exceptionally few Lead Measures. In most instances, the only LEAD measure shooters measure is EXPOSURE TIME, which they may have calibrated based on bad advice, but certainly on inexact measurements (brass color change or glow, flame color change, tempilaq, etc). Few others (myself included when I was flame annealing) employ gas regulators to improve (indirectly measured) flame temperature control as well, but it’s exceptionally rare gear in this game. “Power switch On,” “gas knob on,” and “speed control setting” are typically the only Lead Measures most shooters have to control their annealing process.
This is the reason I focused above on adding direct performance measures as well as improving SOME kind of calibrated LEAD measure control(s)on the process.