I think where Google screwed up is that it instituted a bicameral performance review system without understanding how to implement one. You can have a peer-review system or a manager-driven system. If you're doing things right, you have both and an OR-gate between them. People whose peers think they're doing great but whose managers don't recommend them can still advance, and vice versa. Conversely, to be demoted or fired, people have to fail both. The OR-gate system removes the career SPOFs that lead to degenerate behavior: managers can't exploit their position because people with strong peer support can still succeed at the company, and people don't avoid or shirk low-visibility but important projects

The problem with Google's system is that it has an AND-gate for promotions and an OR-game for demotions and firings, and that's a disaster. At Google, you need strong peer reviews (which largely comes down to visibility) to get promoted, but you also need to do well on a "calibration score" stack-ranking system (i.e. Jack Welch vomited, someone put a candle in it and called it a birthday cake) that is strictly manager-driven, extremely opaque, and often deeply unfair. (Google has historically had a problem of managers who abuse the calibration score opacity, giving low ratings to make it impossible for their reports to transfer.)

The downside of a manager-driven system is that managers can exploit the power, being career SPOFs. The downside of a peer review system is that people jockey for visibility and work that is important but not visible or "sexy" gets neglected. Google's AND-gate system delivers the negatives of both, but none of the advantages.

Google has some really great people, and in many ways, it's a really good company, but their abortion of a "Perf" (look, even the name is imbecilic) system is not one of their strong points.