01OKRs set direction, and are kept apart from ratings
Objectives and key results are usually set each quarter at company and team level. Most guidance, and most experienced practitioners, advise against scoring individuals on OKR attainment. When a rating depends on hitting a key result, people set safe targets and the method stops working. OKRs explain what the team was trying to achieve. The review then asks what the individual contributed to it.
02Career ladders define what good looks like
Engineering, product and design teams publish levels with expectations for scope, technical skill, collaboration and leadership. Many offer parallel routes for managers and senior individual contributors. A review compares someone’s work with their level, and a promotion case shows sustained performance at the next one. Where ladders are vague, reviews drift towards visibility and confidence, which disadvantages quieter engineers and people working remotely.
03Peer feedback and calibration do the heavy lifting
Because a manager sees only part of the work, reviews lean on written feedback from peers, product partners and tech leads. Calibration sessions then compare ratings across teams to keep standards consistent. Both can become expensive. Companies that keep the cost down ask for fewer, more specific peer reviews, gather feedback through the year and limit calibration to the cases where managers disagree.
04Output metrics mislead
Lines of code, tickets closed and commits are easy to count and easy to game. Delivery measures such as deployment frequency and lead time are designed for teams and systems, and lose meaning when applied to a person. Good reviews describe impact in words: the incident that was handled well, the design that unblocked two teams, the junior engineer who was mentored to independence. After a re-organisation, the record needs to follow the person to their new manager.