Discover more from Software Design: Tidy First?
Performance Review Incentives
Gergely Orosz’s excellent The Pragmatic Engineer substack recently published a couple of pieces on the performance review process at large tech companies. If you are going to be subject to performance reviews, it makes sense to have a plan to navigate the process. Consequential money may be on the line, along with self-esteem & future career prospects.
Gergely hasn’t (yet?) covered the performance review process as an incentive system. Here, at last, we have an incentive system consciously created as & designed as an incentive system (unlike many of our other examples to date). The business goal of performance review is to change employees’ behavior, to align their incentives with those of investors.
As with so many other incentive systems, though, the performance review incentive system creates nasty second-order effects. The effects-as-experienced can contradict the effects-as-intended. In this piece I look beyond the salient features of performance reviews to the incentives those features create.
Every <review period> (typically annually or twice-annually), all employees are evaluated. Good performers are rewarded. Poor performers are counseled or fired.
Managers draft evaluations. Those evaluations are adjusted by being compared to the evaluations drafted by other managers. Eventually the process converges on a grade for the period for each employee.
Because being rewarded is good & punished is bad, employees are incentivized to get good evaluations. Note I didn’t say “employees are incentivized to do good work”. The gap between the forces exerted by the review process & the forces exerted by some mythical “do good work” incentive can be substantial, even contradictory.
Periodicity of reviews makes time lumpy for ICs. Early in the period they can afford to take greater risks in hopes of getting a gold star. Towards the end of the period, if nothing has worked out they have to do something, anything, to get a silver star & it has to be something that will turn into evaluation fodder in 3 months. Then 2. Then 1.
ICs are incentivized to do work that shows up in reviews. ICs are disincentivized from doing work that doesn’t show up in reviews, no matter how important that work actually is. The question I grew to hate is, “Is that OKR work?”
Still, you have to get good reviews to stay in the game. Take notes. Measure befores & afters so it’s easy for your manager to tell your story. Just remember your rating says some about you & a lot about the other folks in your organization.
Managers are incentivized to give good employees good reviews. But what is meant by “good employees”? Easy to manage (with all the bias this introduces). Make the manager look good. Simple to review (reviews take substantial effort for managers).
A particularly nasty form of manager incentive is formal (or informal) stack ranking. What if everyone on your team is doing great? If you take consistent positive reviews into the review process, you’ll definitely be punished for being too soft (never mind that statistically in a large organization some teams are bound to be all high-performers).
When my manager at Facebook gave me a lower-than-expected rating, if they said, “As you know, expectations at your level are very high,” that meant that I had gotten screwed. My manager wasn’t willing to take the hit for grading too many people too well & I’d drawn the short straw. (This was different from when I actually didn’t perform well.)
Whoever manages the review process also experience incentives through the performance review process. If the process runs smoothly, that’s good. If it’s economical in terms of their time (never mind the ICs’ or manager’s time), that’s good. If it results in defensible decisions, that’s good (you hate being charged with bias without a strong shield). If it results is “non-regrettable attrition” (i.e. firing the people you wanted to fire anyway), that’s good.
I wish that everyone involved in performance review would frequently remind themself that this is a big machine they are operating. We’re not trying to produce truth. We’re producing decisions that are pretty good for the business & most of the people involved.
I frequently coached engineers who had gotten less-than-stellar reviews. My first suggestion is that they reframe the review. The bronze star is a judgement of your manager’s storytelling ability. Now, maybe you didn’t give them much to tell a story about. But even if you did, they could muff the story & you’d still get a bronze star regardless of your performance.
Performance reviews introduce the Pie Problem. If you want more pie, you can either make the pie bigger so your slice is bigger or you can just make your slice of the same size pie bigger at the expense of other slices. The relative ease of these two strategies, bigger pie or bigger slice, is strongly affected by the performance review process.
Before going into the performance review process, remind yourself what’s actually at stake. If almost nobody ever gets fired & the difference between a bronze star & a gold star is a 2% raise versus a 5% raise, then who cares?
My gut tells me there are better ways to solve the problems performance review purports to solve. I’ve only been working on this problem for 7 years, though, so it’s too soon to expect any progress.
For a process intended to align incentives, performance reviews sets folk at each others’ throats to a disturbing degree. If what’s good for my review is bad for customers or investors or executives or some other team, guess what I’m going to do? It takes courage & vision to do “what’s right”. Given the power differentials in the workplace, such courage is understandably rare.
That said, one of my favorite ways of breaking out of performance-review-induced paralysis is to ask, “What do I as an investor wish I as an engineer would do here?” Just don’t expect to get rewarded for it.