http://www.businessinsider.com/google-engineer-20-time-is-de...
It can be an effective way to work out a transfer. Google's transfer process is broken-- you are fucked without a high-level (preferably Director or VP) sponsor-- and that requires that you kill it on an audition project. In practice, though, this usually ends up requiring more than 8 hours per week, so you're either working long hours or putting less than 80% into your assigned project, which is a gamble.
- There's no such thing as a "Perf Room" as a special thing. calibration happens in regular meeting rooms.
- There quite possibly are some bad managers, but they are few and far between. There's plenty to gripe about at Google, but immediate managers are very rarely the topic.
- The transfer process is not broken, though it is not a revolving door either. You don't require high level sponsorship, though high level folk will be required to sign off on transfers for obvious reasons (what if 50 people all suddenly changed to teams in particular office?)
- 20% time is a great way to "audition" with a potential team you might like to transfer to, but I have never heard anyone but michaelochurch describe taking appropriate 20% time as "a gamble". There are naturally times when your manager might ask you to defer 20% work (occasional crunch time, or similar), but that's still a rarity.
Really, when a Google manager threatens to "take you into the Perf Room", it's purely metaphor. Calibration applies to all employees (not just the disfavored ones) and no one is actually in that room as they are "calibrated", so the phrase has no meaning other than as a general-purpose threat.
I've heard plenty of stories about bad managers at Google: people given low calibration scores to keep them captive on undesirable projects, people threatened with PIPs for failing to fall in line on RNCH, and one case of a manager who used phony performance problems to find things out about peoples' health and personal issues, then exploit the knowledge in sadistic ways.
The transfer process is pretty damn broken at Google. A company with that much in the way of resources has absolutely no excuse not to use open allocation. It's actually an ethical problem. If you're poor and fighting for survival, you may not be able to afford open allocation. If you're a rich as Google, it's an ethical imperative.
20% time is a huge gamble. It might be one worth making, sure, but it's dishonest to say there's no risk. If your manager doesn't get a lot of calibration points and you have a 20% time project, you'll often be thrown under the bus so more "loyal" reports can get a decent score. That's just the way stack ranking works. If one person on each team is to be shot in the head, the person perceived as least loyal (flight risk/20%T) will be the first one gone.