The Google career path, Part 3: Performance reviews and prom

# The Google career path, Part 3: Performance reviews and promotions

This is the third part of a series of posts on what the career trajectory looks like at Google for a software engineer coming from a research background. The first post talked about [how to get started](http://matt-welsh.blogspot.com/2014/05/the-google-career-path-part-1-getting.html); the second was about [how to initiate new projects](http://matt-welsh.blogspot.com/2014/05/the-google-career-path-part-2-starting.html). In this post, I want to talk about the performance review and promotion process.

Standard disclaimer -- This is my personal blog, nothing I'm saying here is approved or endorsed by Google, and any opinions I express are mine alone. Also, keep in mind that I am only describing the process from my own perspective as an engineering manager; there are many different ways in which performance reviews work at Google, and I'm not the final authority on any of this stuff. Other Googlers are welcome to correct me in the comments if you disagree with any of the below.

**Preliminaries: Job levels**

Like most companies, Google has a "level" system for engineers. These were originally numbers from 1 to 10 (where most engineers are levels 4 through 6), but at one point they decided that they needed to promote Jeff Dean and Sanjay Ghemawat again, so they added a level above 10 -- like the amps in _This is Spinal Tap_, our leveling system quite literally goes to 11. Each job level corresponds to a title; for example, L5 is "senior software engineer" and L6 is "staff software engineer". There are corresponding titles for people on each level in management, sales, etc. ladders as well. I'm going to stay focused on software engineers since that's what I know best.

Someone joining straight out of their bachelor's degree would typically be an L3, and a freshly-minted PhD would generally come in at L4. A newly tenured professor (as I was when I joined) would be L5 or L6. Industry experience counts for more than academic experience -- for example, five years' post-Bachelors at another company would also equate to an L4 or L5, so it's true what they say that doing a PhD doesn't usually pay off.

Unlike many other companies, your job level level generally has more to with your compensation than what kind of work you do. That is, nobody is going to say, "You're only an L4, therefore you can't work on this project." Indeed, nobody seems to care what level you are. I work with a wide range of people from across many different levels, but what matters is not how much they are paid, but how they are contributing.

It is true that higher levels correlate with greater degree of responsibility, but (and this is important) job levels reflect the impact you're having, not the other way around. In other words, you get promoted at Google based on what you have accomplished -- you don't get promoted and _then_ take on some big task. What I always tell people: If you want to get promoted, just start acting like someone at the next level up. Eventually they'll realize you're not being paid enough and will promote you.

**Performance reviews**

Engineers at Google go through a performance review every six months, once in the spring and once in the fall. (Actually, the spring cycle is only required if you're going up for promotion.)

Unlike at many companies, at Google the performance review is **peer-driven**: Everyone solicits reviews from their peers, and in turn, everyone writes up short blurbs about the people they have worked with, how they're doing, how they could improve, and so forth. You get to see these reviews when the process is over, and critically, you know who wrote them -- it's not anonymous! However, there is a way to provide private feedback that can only be seen by your manager; in my experience this backchannel is rarely used.

Managers take all of your reviews as input and determine a performance rating, which is a score in one of five buckets to indicate how well you're doing overall. If your scores do well over several review cycles, you have a strong case to go up for promotion. These scores are also **calibrated** against other managers across the organization, to prevent managers from biasing their scores -- an individual manager can't game the system.

The peer review process is not meant to be tedious, but if you have a lot of reviews to write (or you're a manager), it can be very time consuming. People take it pretty seriously though. It's a fantastic way to get direct feedback on how you're doing and learn how you can improve. People tend to be nice, but they don't hold back if there's something you could be doing better.

**Promotions**

I've served on a number of promotions committees so I have some idea of how the sausage is made. Every promotion case gets reviewed by a committee of several Googlers who are above the level at which the promotion decision is being made. (So if you're going up for promotion to L5, the committee would be people L6 or above.) Different levels have different expectations of what constitutes a strong promotion case. **Launching something** is one of the best ways to get promoted; but of course, launches vary in terms of scope, complexity, and impact. Launching a tweak to the Slovenian-to-Finnish translation backend might get you promoted to L4, but not to L6.

In general it's expected you've been at your present level for a year or more before getting promoted, but this is not a hard and fast rule. A strong promo case is one where you have sustained high performance at your current level and are already having impact that would be expected for the next level.

Everyone on the committee stack ranks the ~25 or so cases that the committee is looking at, and a whole day is spent discussing all of the cases and making a promote/no-promote decision. In some cases, promotions are passed through a secondary committee as a cross-check; _both_ committees must approve the promotion for it to go through. And finally, a manager is able to appeal an unfavorable decision for one of their reports if they feel that the committee overlooked some important piece of information (this often happens when the committee misunderstands some aspect of the packet).

Having seen the promotions process in action, I think it works pretty well. We don't always get it right, and the process is somewhat noisy by its very nature. Unlike some other companies, people don't get promoted just because their manager says they should be; it's a pretty rigorous review process and there are plenty of checks and balances.