@soatok Many years ago, I read a productivity study that said ‘knowledge worker’ (a phrase I hate) productivity increased to 20 hours a week, plateaued until 40 and then decreased. Over 60 it tended to be net negative. A few years later, I read a study that attempted to reproduce the result and came up with numbers that were well within experimental error margins.

This seemed obvious for programmers. It takes ten seconds to introduce a bug that takes a week to debug. If you sleep for ten hours instead of writing buggy code while tired, you will achieve more simply by not creating more work.

I was quite surprised recently talking to someone who studies this kind of thing as her research topic. She was just finishing up a study of construction workers that came up with almost the same numbers. I guess it shouldn’t be surprising that fixing an error in a building is very expensive, so even if they may be easier to find than programming defects they’re much harder to fix and it works out roughly even.

I worked 20 hour weeks for a while when I was contracting (always bid as fixed-price, so the customer didn’t need to know how long I was working). Several people expressed surprise at how quickly I worked. I did the same during my PhD. People in the lab were surprised that I would roll up at 10 or 11, have coffee for an hour, work an hour, go for lunch for a bit, then work a couple more hours and still got a lot done when they were working much longer hours. I tried to explain that I wasn’t more productive in spite of working shorter hours, I was more productive because I was working shorter hours. I’d turn up after a good long night’s sleep, relaxed, and then work for four really focused hours. I got far more done in that time than I would if I had been working a ‘full’ work day.

When people join my team, I tell them this and tell them I want 20 productive hours of work from them. How they spread that out over the week is up to them.