Originally Posted by fishboat

in a past life, if i had your setup to measure what you're measuring, i'd have done a gage r&r study. No doubt shimano and campy use gr&r or gr&r-type studies to monitor their measurement systems that guide the manufacturing processes used to make the derailleurs/shifters your measuring.

Gr&r is a very organized, objective way to determine whether your measurement system is capable of reliably measuring, in this case, derailleur movement or cable pull. It can also be used to identify what, i.e. Cable pull, your measurement system can reliably measure. For example, using arbitrary numbers, if a measurement system is determined to be incapable of reliably measuring 3mm of movement, you can use the results of the study to determine the measurement system is capable of reliably measuring 10mm of movement. Another example..if someone needs an accurate determination of, say +/- 0.01mm on a 15mm(target) steel rod diameter, a metric tape measure would be an incapable measurement system while a micrometer would be a capable measurement system.

All that said, a good gr&r study can be a fair amount of work. In a precision production environment, it's a great investment. It would certainly apply to your study, if you want to dive in.

Off the top of my head a couple thoughts.
1. I'd use new cables, if you're not already, just to eliminate any measurement variation noise added by worn, or less than ideal, cables. only one cable and it and the casing are new and dry of lube (as i will be changing out levers and ders lube would get messy)
2. New shifters would (probably) lower the measurement noise, but wouldn't be practical for what you're doing. I would use the same shifters for all measurements within a study and make sure they are in good condition and lubricated as needed. both shifters (and ders) are used but do function well and don't show wear or slop beyond what i feel is normal
3. You seem to be measuring cable pull through the rear derailleur body. no, the cable pull is measured at the cable's exiting the down tube casing stop. For this no der would even be needed as long as some sort of tension could be applied to the cable i'd think the der body would add random noise to your measurements and degrade your results. It may be useful to measure rear cable pull by measuring the cable movement itself. You could also measure the same cable-pull der movement as you have setup now. In an ideal world they'd be the same, or the same within some consistent scalar amount. A comparison of the two sets of data would tell you how much noise measuring through the der adds. A new versus used der movement would be another study to give you an idea how much influence the der is having on your results.
4. Make sure you use averages for each measurement and not single points. i was averaging the measurements and for this series was doing them twice but i think 3x would yield better results. each data point for each shift-movement should be at least an average of 3 measurements. A better estimate of the true movement would be 10 measurements of each shift-movement. The estimate of the true shift-movement by averaging say 5 measurements isn't (statistically) any better than averaging three points. You need 10 determinations to get a better estimate than averaging 3.
5. It would be time well spent to just take a bunch of measurements (i'd do 10 each for the time being) of shift-movements and determine the average and standard deviation. Then divide the standard deviation by the average and see where you're at. Again, random numbers, if you have an average of 10mm with a std dev of 3.2mm then you have a relative std dev of 32%. Ouch.. If you compare the ave, std dev, and relative std dev of two adjacent shifts you may get some indication of how capable your measurement system is in measuring shift movements. There's more statistics that can be applied to this, but start small first. this is some of what i am considering next. I have a good friend who taught college stats and has worked for various industries (he would have to kill me if he told me the companies
6. It's worth the time to sit down and write out exactly the goal of your study(s) and what measurements you'll do to gather your data. It'll help keep you on track and it'll be easier to identify when you stray from goal or process. i know what my final goal is and if i never reach it this is fun anyway. Andy

even "simple" studies can get quite complicated.

I think you're on the (critical) right track at the moment as you're asking "what am i trying to do?". If your not sure where you're going then any path will get you there.