Why Meta Paid Bonuses to People It Fired

Last week, something odd surfaced from the latest round of layoffs at Meta.

The company’s evaluation system uses ratings like “Met Some” to automatically flag employees for termination. This rating is theoretically reserved for those who are struggling. Yet some of the engineers on that list had just received performance bonuses of 120%.

On one hand, the system paid them a premium for excellence. On the other, it fired them for failure.

It looks like a glitch. But it makes more sense when you separate two things: the actual work people do, and the systems companies build to measure that work. At small scale, these stay close together. At large scale, they drift apart. One system saw these engineers as top performers. Another saw them as underperformers. Both were looking at the same people.

In the book I’m writing, I call this crossing a threshold. It’s the moment when an organization gets too big to see the work directly, and has to start relying on systems to represent it.

When you manage a team of five, or even fifty, you can see the work directly. You notice the late-night commit that saved the launch. You hear the mentorship happening in the hallway. This is what I mean by the Territory: the actual reality of what’s happening.

But somewhere past 150 people (the famous “Dunbar’s Number“), this breaks down. You can no longer keep the reality of everyone’s work in your head.

To keep things running, you build a Map. That’s the ratings, levels, and KPIs we talked about. You compress complex people into data points that fit in a spreadsheet.

This isn’t a mistake; it’s a survival mechanism for the organization. You cannot run a 60,000-person company without a Map.

The danger isn’t the Map itself. The danger is what happens to the culture once the Map takes over.

This is the drift I’ve been observing. Once an organization crosses that threshold, the Map becomes easier to see, easier to use, and more “real” to leadership than the work itself.

At Meta, the bonus system was likely tied to the Territory (specific project outcomes, code shipped). But the layoff process was tied to the Map (the “Met Some” label). When the two conflicted, the system defaulted to the Map.

This is what happens next. The system begins to optimize for what can be measured, rather than what matters. Employees learn that it is safer to protect their rating (the Map) than to take risks for the product (the Territory).

This leaves us with an uncomfortable question. If we need the Map to scale, but the Map blinds us to reality, what is the role of the manager?

It suggests that the job of a leader at scale isn’t just to read the dashboard. It’s to constantly test it. To recognize that you’re no longer seeing reality directly, and to find ways to check whether the map is lying.

The error at Meta wasn’t using a spreadsheet. The error was trusting it completely.