Re-evaluating the Evaluation Process

Annual performance reviews are where good intentions go to die. Managers dread them, employees dread them, and it is honestly hard to think of another process that everyone involved dislikes quite so evenly.

I should say up front that I am not writing this from the safe side of the table. As a CTO, I am one of the people who has to run this process, so if it is broken, I am part of what is broken.

I have seen this pattern before, in board exams. We spend twelve-plus years training kids to crack exams instead of think, and then wonder why workplaces are full of adults who optimize for the review instead of the result. I suspect those two things are connected.

The current circus

Here is how it usually goes. Once a year (twice, if your company is feeling progressive), we sit people down and rate them on a scale, pretending we can put a number on creativity and a decimal on collaboration. The process looks something like this:

• A self-evaluation where you either undersell yourself (imposter syndrome) or oversell yourself (survival instinct)
• A manager evaluation based on whatever they remember from the last two weeks
• A 360 review that's either a popularity contest or a passive-aggressive bloodbath
• A rating on a scale nobody understands but everyone's career depends on

Then we tie it all to compensation, so that someone's EMI depends on whether their manager remembers the bug they fixed in February.

Why it doesn't work

Start with memory. We evaluate a whole year based on the last month, which is like judging a movie by its credits. It is the same thing we did in school, where twelve years were judged by three hours of panic-writing, except that now we call it an annual review.

Then there is the bell curve. "Only 20% can be top performers." In other words, we have designed a system that tells most of the team, every single year, that they are not good enough. It is the same logic as "only the top 1% get into IIT," an artificial scarcity of success. You can call it math if you want, but for the people on the wrong side of the curve, it quietly wears down morale.

We also measure what is easy instead of what matters: lines of code, tickets closed, hours logged. All that does is reward busy work. I wrote about this earlier this year in Systems Need Change Every Few Cycles: the moment a measure becomes a target, people start gaming it. Whether it is students with marks or engineers with tickets, it is the same game with a different scoreboard.

And we save feedback for special occasions, as if it were fine wine. After spending eleven months guessing whether they are doing the right thing, people find out in a single meeting. It is the same as studying all year, getting the marksheet, and only then realizing you were doing it wrong.

School with paychecks

Look closely at most Indian workplaces and you can see the school they came from. Annual exams became annual reviews, percentages became ratings, and class rank became stack ranking. Coaching classes became interview prep, and we graduated from memorizing formulas to memorizing STAR answers.

The kids who stayed up memorizing answers grew into adults memorizing KPIs, and the question we heard at home, "Sharma ji ke bete ko kitne mile?", simply moved to the office as "where are you on the curve?"

I wrote in Choice: Knowledge Vs Grades that somewhere along the way we started valuing the marksheet more than the thing it was supposed to measure. Instead of ever stopping, it turns out we just attached salaries to it.

What I'd try instead

While I do not have a perfect replacement, I am suspicious of anyone who says they do. Still, a few things seem clearly better than what we have now.

Talk more often, about smaller things. A real 1:1 every week or two, about growth and blockers rather than status updates, is far more useful than one big annual verdict, and it should be the meeting you never cancel.

Evaluate work when a piece of it ends, rather than on a fixed date every year. "How did this project go?" is a far more useful question than "how was your year?"

Let people call out each other's wins as they happen. It does not need to be formal, because "hey, Sarah killed it on that migration," said when Sarah actually does, is enough.

Map skills instead of handing out ratings. "You should get comfortable with GraphQL next" is something a person can act on, while "you're a 3.7 out of 5" sounds like a hotel review.

And separate pay conversations from growth conversations. Because money makes everything awkward, it is hard to talk honestly about what you are bad at when your raise is sitting on the table.

None of this needs a company-wide rollout. Pick one team, drop the annual review, try it for a couple of quarters and see what breaks. Different roles probably need different versions anyway, since judging a researcher and a salesperson on the same form is like grading fish on how well they climb trees.

Why we keep it anyway

Honestly, because it is easier. Saving difficult conversations for one scheduled day is easier than having them every week, and hiding behind a form feels safer than looking someone in the eye and saying "this isn't working, let's fix it."

It is the same reason we cling to board exams and JEE. Trusting teachers to judge students as whole people feels risky, so we trust a three-hour paper instead. We have been trained since childhood to believe that evaluation is supposed to be painful, periodic and a little bit punitive.

What is really hard is helping someone grow week after week without a rating to hide behind. That is the job, though, whether they are 16 and studying for boards or 36 and hoping for a promotion.

I do not think the annual review can be fixed with small tweaks. It would be better to retire it altogether and build something that helps people get better, instead of just making them nervous.

I know this is hardly a revolutionary idea. But in a country that perfected teaching to the test, actually teaching might be the most radical thing we can do.

P.S. If your own review is next week, good luck. Remember, they probably only remember the last two weeks.