Performance reviews

Anton V Goldberg
3 min readDec 9, 2020


It’s the time to be jolly and write performance reviews. Every year we get told that we’ll have to write no more than some really small number of reviews. I swear, sometimes I feel like some people have never seen an org chart, much less a social graph.

If there is one thing I am definitely not missing from Amazon it’s performance reviews and stack ranking. Stack ranking is where organizations have a percentage quote on how many people can get the highest rating and how many have to be put on a “performance improvement plan” (i.e. managed out). Imagine a good organization of 100 people. It’s a good organization but it still has to find the worst 10% and fire them. As with any organizations, it has a few really outstanding folks but the rest of the talent is fairly uniform within the levels. It’s a good organization, so the attrition is low. At the beginning of year X-1 there were between 93 and 100 people in the organization. Some of them went away, perhaps the org expanded a little but 7 new people were hired. The organization now numbers 100. All of them new hires and all the old timers became eligible for the review in the year X. 1 out of 7 was a bad hire and so can be safely fired. However the leadership needs to fire 10 people (10% of 100). Perhaps one person within the org decided that their future is in baking and just works out their last days before the next vesting day when they finally have enough to open their own bakery. The leadership has a cause to get rid of the bad hire (1) + perspective baker (1). The leadership needs to find 8 (10–2) more. Good leadership at that point pushes back against the HR and protects the org. However If the push back attempt wasn’t successful the leadership needs to find 8 more people to put on the plan. The process can’t be anything but random and capricious. Nothing destroys the culture within an organization like random and capricious firing. People begin looking for an opportunity to leave. People who stay begin picking up such tasks as looking good on the review instead of looking for real work. And a year later the org lost its morale.

Going deeper into reviews in general, I don’t understand the need for the whole review process as it exists now. Clearly, management quality has to be assessed but every good manager knows performance and character of every direct report of theirs. No amount of peer-to-peer reviewing can significantly improve their knowledge. Every good manager also keeps employees aware of the manager’s view of employee performance and any intra-team issues that need to be addressed. In fact, if an employee doesn’t know what the manager thinks of their performance outside of the review period, that employee should seriously think of asking or moving to a better managed place.

In terms of assessing a manager’s performance, there are many ways to unobtrusively do so. When Amazon started “developers surveys” among questions about various technologies, issues slowing down development and so on, there were clever questions addressing manager’s performance — directly and indirectly. From what I remember, about 50 managers were fired shortly after. Given that Amazon’s tech community was about 1000–2000 people at the time, that’s a sizable clean up. All that was required was 5–10 well-defined questions instead of hours spent writing reviews, asking for feedback, managing load and so on.

I believe we need a mechanism that lets employees anonymously review their managers while managers should continuously provide feedback and performance improvement advice to the employees. The simpler and less intrusive these mechanisms are the more natural they will seem and the better they will work for healthy workplace dynamics.