What the Layoffs Left Behind

The morning after a layoff, an engineering organization looks fine. The people who were cut have lost their income, their health coverage in some cases, and a visa timeline in others, and that is the serious harm in this whole subject; nothing further down this page competes with it. But the systems are still up, the sprint board still has cards on it, and the people who remain are at their desks by nine, working. If you were measuring the organization on output that week you would find very little wrong.

What has changed is not visible for a long time. Every engineer in that building has just received a piece of information about how their employment works, and they will act on it, not dramatically, not by slacking off, but in the small technical judgments that make up a working day. Which of two approaches to take. Whether to raise the concern. Whether to say the estimate is slipping. Whether to touch the module nobody understands.

Those judgments do not scatter randomly. They all move the same way, toward the visible, the safe, the defensible and the individually attributable, and the sum of them is an organization that looks busy and agreeable and is quietly making worse decisions than it made a month earlier. That is the part I want to describe, because it is the part that gets missed, and because the bill for it arrives about two years later in the form of systems nobody was willing to touch.

The executives said it themselves, and unusually plainly. Meta cut more than eleven thousand people, about thirteen percent of its staff, on the ninth of November 2022, and Mark Zuckerberg's message said the surge in online commerce had been widely predicted to be a permanent acceleration, that he had believed it too, and that he had significantly increased investment on that basis. Sundar Pichai, announcing about twelve thousand roles on the twentieth of January this year, wrote that the company had hired for a different economic reality than the one it now faced.

Both of those are the same statement: we staffed against a demand curve that did not continue. That is a forecasting error made at the top of an organization, and it is worth saying without malice, because forecasting revenue two years out is genuinely hard and everybody in the industry made a version of the same call. It is also worth saying without softening, because the error was made by people who were not the ones who lost their jobs over it.

The second half is money. In March 2022 the federal funds target range was 0.25 to 0.50 percent. This month it is 4.75 to 5 percent. When capital was close to free, a plan that would pay off in three years was cheap to hold; at five percent it competes with simply not spending, and every internal project gets re-evaluated against a hurdle that moved underneath it. That is arithmetic about the cost of money, and it arrived at every company in the same eighteen months, which is why the announcements clustered.

Neither of those causes is a statement about the people who were cut, and this matters more than it sounds, because the story an organization tells itself about the cause determines what the survivors conclude. If the story is that low performers were removed, everyone remaining infers that the measurement was real, and starts optimizing for what they guess it measured. If the story is that the company hired against a forecast that reversed, the inference is different and considerably closer to the truth.

Technical risk aversion is the one nobody decides. The refactor that would make the next year easier also might break something, and breaking something now has your name on it in a period whose rules you do not know. So it slips. The migration that pays back in eighteen months loses to the feature that demos in two weeks, because one is visible before the next review cycle and the other is not. No meeting produces this. It is the sum of individually rational choices by people who suspect they are being measured and cannot tell on what.

Visibility becomes the target instead of the byproduct. Work that is legible to a manager and work that matters are correlated, not identical, and the gap between them is where all the damage sits. The engineer who quietly kept the build green, unpicked the flaky test, and answered everybody's questions has nothing to point at in a review. Everyone knows this already. In a normal year they do the work anyway. In this year they notice the difference, and some of them stop.

Candor goes first, and it is the most expensive loss. Saying an estimate has slipped, or that an approach is not working, is a small professional cost in a stable year and feels like a large one now. So estimates get padded quietly, bad news moves upward slowly if at all, and the approach that is not working gets another two weeks. Amy Edmondson's 1999 paper in Administrative Science Quarterly named the condition - whether people believe the team is safe for interpersonal risk - though her study looked at teams inside one manufacturing company, a narrower base than the idea's popularity implies.

Knowledge hoarding is the one nobody says out loud. If being the only person who understands the payments system feels like protection, people will behave accordingly, and they will not do it consciously or admit it if asked. It shows up as a documentation task that stays unstarted, a pairing session that keeps getting moved, a runbook that describes the happy path only. Every good practice for spreading knowledge is now running directly against a survival instinct, and the instinct is not stupid. It is just wrong about what protects you.

The scope did not get cut with the team. Teams are reduced and their systems are not. The same services, integrations, on-call rotations and compliance obligations are now carried by fewer people, and unless somebody explicitly retires something, that is a capacity reduction being recorded as a productivity expectation. The maintenance floor is the first thing to be silently skipped, because it is the only part of the load with nobody asking about it this week.

Meta's message on the fourteenth of March, twelve days ago, said the company would cut around ten thousand more roles, and it named a schedule: recruiting immediately, the technology groups in late April, the business groups in late May, and different timelines internationally. Zuckerberg said the advance notice was what employees had asked for after November, and I think it was the right call. It also means that for two months a great many engineers come to work knowing a decision about them exists and has not been shown to them.

That is the sharpest version of a general problem, and there is no clean answer to it. The alternative to telling people early is telling them late, which is worse in every respect that matters, so the honest framing is that a company in this position is choosing between two costly things rather than avoiding a cost. What a manager should not do is pretend the interval is neutral. It is not. It is the period in which every behavior in the previous section is at its strongest.

The thing that fills an information vacuum is not calm. People construct an explanation from whatever is available - a canceled meeting, a hiring freeze on one team and not another, a director who has stopped answering as quickly - and the explanation they construct is reliably worse than the truth, because the worst available reading is the one that feels prudent to prepare for. Silence is not the absence of a message. It is a message that everyone reads the same way.

Engineers at organizations that have never had a layoff read the same news. They see the announcements, they see friends from a previous job posting that they are looking, they see a market where the same role gets many more applicants than it did a year ago, and they draw a conclusion about how safe employment is in general. Then they take that conclusion to work and it changes exactly the same judgments, at a company that is doing fine and has no idea why its teams have become slightly more conservative.

That is a strange thing to sit with as a manager. You have made no cuts, said nothing alarming, and have no bad news to deliver, and your team is still operating with the ambient assumption that the ground might move. The instinct in that position is to reassure, which is the least effective available response, because it is unfalsifiable and everyone knows you would say it either way.

What does work is smaller and more specific: being concrete about what you know of the company's position and where that knowledge ends, naming the invisible work in a review rather than assuming it speaks for itself, and being visibly unbothered by the kind of honest mistake that people are currently afraid to report. That last one is the only real lever anyone has on candor. It is not a speech. It is what happens the first time somebody brings you bad news in a bad quarter.

Say what you know, on a schedule, including that you know nothing. An update that says there is no news is a real update and it should still happen on the day it was promised. The value is not the content, it is that the channel is working and the absence of information means the absence of information rather than something being withheld. The moment a promised update quietly does not happen, everybody notices, and their conclusion is not that you were busy.

Never promise what you do not control. A manager who says the team is safe and turns out to be wrong has spent the only asset the role actually has, and cannot get it back for the people who heard it. Say what you have been told, say who told you, and say what you have not been told. That is less comforting in the moment and it is the only version that survives contact with whatever actually happens.

Make the criteria explicit, because vagueness is what frightens people. Fear is worst when people cannot tell what is being measured, and it drives them to guess, and their guesses are usually about visibility rather than value. Being concrete about what good work looks like on your team, what you will actually be looking at, and what you will not, is calming in a way that reassurance is not, because it can be acted on and it can be checked against your behavior.

Protect the invisible work by name. If the build maintenance, the on-call improvements and the answering of everybody else's questions are not named as work in the same conversation where features are named, the incentive will resolve the question for you and it will resolve it wrong. This costs nothing and it is one of the few places where a manager's stated priorities and the team's actual behavior are directly connected.

A manager cannot make people feel safe when they are not safe. Employment at a company that has just cut thirteen percent of its staff, or that might, is genuinely less secure than it was, and everyone can see that. Any attempt to produce the feeling without the fact is both dishonest and completely transparent to the people it is aimed at, most of whom have been through this before somewhere else and can recognize the vocabulary.

What is available is narrower. Being straight about what is known. Being specific about what is measured. Absorbing as much of the ambiguity as the role allows rather than passing it downward in the form of vague warnings. Noticing when someone has stopped saying difficult things and going to find out why. None of that is control and none of it changes whether a decision two levels up lands on your team.

It is a great deal less than anyone wants, and it is not nothing. The distance between a team whose manager did those four things and a team whose manager did not is, in my experience, most of the difference in how the organization comes out the other side, and it is entirely made of small conversations that nobody records.

I want to be careful with the scope of all this. I have described a direction rather than a magnitude, and I cannot tell you how large the effect is, because I have not measured it and do not believe the numbers people cite. Some organizations go through a reduction and are recognizably themselves within a quarter, usually the ones that told people the truth early and cut once rather than three times. The mechanism is real; how much of it lands depends on things a single manager does not set.

What I am certain of is where the bill shows up. Not in the quarter after, when the charts look better and the payroll line is smaller and somebody concludes that the organization was carrying slack. Two years later, in a system everyone has been routing around, that nobody proposed to fix, because in the year it needed fixing the person who would have proposed it had just watched what happened to people whose work was hard to see.