I would argue that measuring productivity is only possible if you’re trying to detect the lack of it. Ie it’s definitely possible to see if someone’s slacking b

I would argue that measuring productivity is only possible if you’re trying to detect the lack of it. Ie it’s definitely possible to see if someone’s slacking behind the rest if whatever output they’re expected to create is way less than the average of all colleagues, but at some point you have to accept that the confidence intervals are pretty vague. In a situation where workload is at least sort of randomly distributed, which it normally is given time, there’s no other good explanation for statistical deviancy than simply less “work” being done. But again, this only works for detecting problems. As you point out, and I agree with you, as soon as you attempt to measure “change in productivity” you’re on very thin ice.