A long time ago I was tasked to trace technicians' work, especially those who where called during the night because of some urgency on French telecom networks.
Astonishingly before that time, it was not known who worked, how many times and what was the outcome of the intervention.
It turns out that only 5 people were working on the 15 in the team.
I remember a union member pleading with me with tears in his eyes to recruit more people because there was too much work. Indeed he was one of the five people who worked hard.
Curiously those five never complained about the others' behavior, on the contrary, the team reacted strongly against attempts at putting some light on what was going on.
Is there another source for this information that isn't twitter with infographic style posts?
They've got a different paper at https://arxiv.org/abs/2409.15152 ... is there an equivalent for the ghost engineers data?
"We have data on the performance of >50K engineers from 100s of companies."
Sounds to me like "we have some very common performance metrics on a huge number of people...the great majority of whom are old hands at gaming metrics".
If you are looking to downsize - the first places to look are both ends of the bell curves for well-known metrics, both ends of the pay curve, and upstream from the worst of the bad decisions, drama, and tensions.
I think those numbers are probably based on source code contributions. I can believe those statistics, what is missing is the why.
Some people just aren’t good programmers. I knew some of those, one turned out to be great at testing instead. Junior programmers, new programmers (to the project/company or language or field) are slow and they might get better. Then you have burnout. There was a story here about a guy that didn’t contribute much because he mentored and pair programmed and did other tasks to make the life of other programmers easy.
There are a lot of good reasons so before you go optimizing on numbers, check the why.