Amazon PIP culture inspired from evolutionary algorithms?
The PIP culture at Amazon seems to have been inspired from Evolutionary Algorithms. From what I know, each year, they have a decided quota to pip out employees - roughly 5%. This is done by stack ranking employees and rating them "Top Tier", "High Value", "Least Effective". There is a fixed quota of 10% to be marked as LE. This means even if those employees are good enough, they are deemed to be relatively least effective by some metric which is supposed to measure how much the company is benefiting from you. Of the 10% LE, about half of them get pip'd out. They hire fresh new faces and the cycle repeats each year. Deliberately discarding x% of candidates and bringing in new ones makes it similar to evolutionarily algos. At an individual level, this is really cutthroat and pressurizing. But at the organization/company level this means the company is always going to keep getting better by the metrics used. Knowing this, and knowing that employment is at-will, is it really bad for some company to be this way? #pip #tech
It assumes you get a steady input of top talent and ramp up is non trivial.
Yes it is bad, for several reasons, but i think the main point is a lot of evolutionary algorithms don’t accurately model this problem nor does it account for enough. An example: If you work with someone who you really respect and think is a great eng, and that person has some issue from life or otherwise that has made them perform poorly gets fired, you as a top performer may get pissed, leave, and bring others away with you. In evolutionary algos, it’s only optimizing for reproduction (getting rid of the weak), but there are second and third order consequences of doing this. Not to mention, those people can go work at a competitor that puts more pressure on your society thriving (less food in algo terms). They may only be low performers due to other circumstances (like poor management or tools), that may not be the case elsewhere. Another thing, how do you know those people are truly low performers? What makes a low and high performer is quite subjective. And the organization is not always culling the weak, maybe just the disliked-by-managers?
Right, the fitness function is short-sighted, poorly implemented, not precise, and just overall can't be guaranteed to be correct.
Also. It treats humans like robots, and doesn't take into account the incredible stress it creates for even folks who are in the top 90 percentile.
It is bad unless you believe that humans are robots. It is not about company. It is about people who work there. While for profit free market is great, ultimately we are all humans. So yes it is terrible to have such system. If someone is working fine, why cause issues for them.
To entertain the premise, GAs have some limitations of their own. One that comes to mind is being prone to finding local maxima and having a hard time getting out of them (can't sacrifice a bit of short-term fitness reduction for a long-term increase).
Amazon has refined and systemized it but managing lower performers within a team is nothing new. As long as there have been performance reviews there have been relative assessments and pruning of the lower performance. It's just that good managers did pragmatically (we need to talk about performance), bad managers did it aggressively (exited the guy who didnt work out), or poor managers didn't do it at all.
Managing out lower performer is one thing and managing out a decent performer because he is least efficient among the group are two different things.
Probably just recycled Jack Welch stuff from his GE days. Forced attrition was all the rage in the 80s and 90s when Bezos began his professional career.
you have been dealing with some cruel evolutionary computation. Care about your solutions, as in no solution left behind.
This topic has been beaten to death on blind and other forums. Can you run a simple search? Some advice for you to ask smarter questions next time: http://www.catb.org/~esr/faqs/smart-questions.html
HR got really mad 😶
Fuck. Started.