I have spent a lot of time on Blind where this is discussed often and I think I have an idea what is going on.
At a high level, Amazon orgs have Unregretted Attrition (URA) targets/quotas for orgs. Not every manager needs to hit the exact target but under a certain level of middle management they do. Because of this built-in churn they need to have a pipeline of low performers - because once the current set of URA targets get managed out or leave you will need to still hit that quota next quarter/half/year/whatever. So maybe your quota is 5% but you have another 5-10% extra in earlier stages of your pipeline so you can sacrifice them later.
In the later stages when you’re actively getting managed out they put you on a formal PIP. In theory it’s possible to “get out” of it but because of the quotas it’s pretty much a lost cause. So on Blind the most common suggestion for people on PIP is to do the absolute bare minimum and start interviewing elsewhere, expecting to fail the PIP.
I imagine the reason Amazon doesn’t want people to know about Focus/Dev plans is that once you know you’re in the earlier stages of the pipeline, you may come to a similar conclusion. Basically if a target is painted on your back and you know there’s a high probability you’re going to get managed in the next 6 months, it’s better to do the bare minimum at work and focus as much as possible on getting a new job. But from Amazon’s perspective that’s a lot of lost work/time/money, and they’re not even sure yet that they want to get rid of those people. And further, if employees don’t know about the performance plan there is less chance they will deliberately sandbag and may actually improve their ranking enough to move off the early stage pipeline.
In my opinion the root of all this disfunction is the quota system. If managers were not forced to manage out so many people and being on a “low performance plan” at any stage were seen as an actual opportunity to identify issues and improve performance, there would be much less incentive to sandbag by doing the bare minimum, and also much less need for secrecy about the process.
But this is Amazon, I am sure they have thought deeply about this. My guess is that they have done the math on many alternatives and this is the least-worst. For example by managing out so many people they can be less selective in hiring which may be better overall (eg an overall acceptance rate of 20% with 10% of those being false positives fills more roles than an overall acceptance rate of 10% with a 0% false positive rate). Of course managing people out also has a cost in creating an unpleasant/toxic work environment that itself increases churn, which they may or may not account for. Keep in mind due to their vesting models, a good performer burning out and quitting after 2 years is actually good for their bottom line. As long as they continue to fill roles I doubt they will be incentivized to change this model.