This is the most important sentence: > Performance reviews are riddled with bias. Let me describe the apocryphal big tech company performance review process.

This is the most important sentence: > Performance reviews are riddled with bias. Let me describe the apocryphal big tech company performance review process. Once or twice a year you ask colleagues for feedback. Hopefully they do. They can also choose not to. You generally need 2-8 people to provide feedback. Less for a "standard" cycle. More for a promotion cycle. The specifics of this process vary from company to company (eg some reviewers might see your self-review, some processes just give textual feedback while others require reviewers to rate you in addition to supporting text). Your manager may then distill all this feedback into a single page. Some processes have the full packet. Others do not. Your manager will, based on this total packet, come up with a provisional rating and then go into calibration to argue your case. This provisional rating tends to be reasonably accurate. It may get argued down. It less often gets argued up (but it can happen). In this process you're essentially being stack ranked by similar employees (ie same job ladder and level). To avoid ratings inflation across orgs (typically 100-150+ people), you are graded on a curve. Only a few people can get Greatly Exceeds Expectations. Some will require a percentage to get subpar ratings (eg Meets Most Expectations). This is the tech company equivalent of cutting the bottom 5-10% every year. These people may end up on PIPs or may be effectively forced out of the company because people with subpar ratings can't transfer or new teams just don't want to gamble on them. Promotion decisions may be made by your org. They be made by committees of higher level individuals on your ladder (eg L5->L6 promotions decided by committees of L7/L8s). Promotion candidates are also effectively or actually stack ranked. As a cost-saving measure there are promotion targets. Only the top X% will get promoted. So how does this go wrong? 1. The number 1 most important thing is for your management team to like you. If they do, your failures will be viewed in a positive light (eg "we learned a lot from this failed project that we will take into future projects") as opposed to a negative light (eg same facts leading to "failed to ship project"). 2. If the influential people in your org don't like you, it's going to be an uphill battle to do well; 3. Managers are subject to the same forces so may curry favor by supporting certain candidates for pomotion and higher ratings over others. There is an awful lot of horse trading that goes on behind the scenes. You will see none of this; 4. To curve fit the org, after all calibration is done people will have their ratings changed (generally down). This is more likely to happen if the org doesn't like you or simply likes you less than someone else; 5. Promotion target percentages are a key cost control measure. Reducing it saves significant costs (eg Ruth Porat did this in a leaked memo when she became CFO at Google). This creates a backlog for pomotion. Standards go up. You have to "wait your turn" and this may take years. This limbo will often keep people around who don't want to leave to get more opportunities. 6. Meeting goals, a key issue in defining "impact", is highly subjective too. You may get an impossible task dumped on you, fail and get dinged for it while someone else gets some way easier goal. The whole process is honestly the worst part of working in big tech.