A midrange phone company I'd applied to twice put out a statement in the spring that said two things in the same breath: yes, we are cutting staff, and no, we are not exiting your market, those reports are false. In the same paragraph, they mentioned they were selling more handsets than they ever had in their price bracket.
I read it three times looking for the lie. There wasn't one. Both halves were probably true. That was the part that took me most of a year to understand, and it changed how I read tech company layoffs — as someone who could get caught in one, and as someone deciding where to send the next application.
What I logged, and what it showed
I started keeping a spreadsheet after I got burned. I'd taken a contract at a company six weeks after they'd published a post about their strongest quarter of customer growth. The contract lasted nine weeks of a planned twenty-six.
So I logged company statements. Only companies I had actual skin in — applied to, interviewed with, got ghosted by, or worked for. Blog posts, press statements, LinkedIn posts from a VP, the occasional all-hands note somebody forwarded me. Forty-one of them over about eleven months. Then I put a reminder on my calendar ninety days out from each one and wrote down what happened.
Nine of the forty-one were followed by job cuts inside ninety days. That number is not a prediction engine; my sample is small and biased toward companies I was interested in, which skewed toward consumer hardware, fintech, and mid-size B2B SaaS. But seven of those nine had done the same specific thing in the original statement, and once I saw it I could not stop seeing it.
They answered a question about money with a number about volume.
Why both halves can be true at once
Go back to the phone company. The rumor they were rebutting was a solvency rumor — this company is running out of road and pulling out of markets. The number they produced in response was a unit number. Record sales in their price segment.
Units are not margin. A phone in the two-hundred-to-four-hundred-dollar band has almost no room in it. When memory and display costs run up across the whole sector — and component pricing has been genuinely brutal for everyone building at that tier, not a story any single company invented — the flagship makers pass it to the customer, because someone paying twelve hundred dollars will absorb another eighty. The midrange maker cannot. Their customer is buying on price. That is the entire proposition.
So the record volume and the layoffs are not contradictory. They are the same fact seen from two sides. You can sell more units than you ever have and make less money doing it, and if your cost base was built for the margin you had two years ago, you cut the cost base. The statement was accurate. It was also an answer to a different question than the one being asked.
This is not villainy. Nobody in comms sat down to deceive me. A company under pressure reaches for the truest number it has, and the truest number a hardware company has is usually units. The distortion happens in the gap between the metric they offered and the metric the question was about.
How can you tell if a company is about to have layoffs?
You can't. Nobody outside the room can, and people inside the room often find out the same morning you do. What you can do is narrower and more useful: notice when a company answers a cost question with a growth number, and treat that gap as information about pressure rather than proof of collapse.
The tell is a mismatch of category. Someone asks about runway, cash burn, or whether a market is being abandoned — a money question. The response is a volume, a rank, a share of segment, a headcount of engineers hired, a number of markets served. All real. None of them denominated in dollars-per-unit or dollars-remaining. When the categories don't line up, the company is telling you which numbers it has that look good.
The other reliable signal is what I started calling the tense shift. Statements about the past are checkable. Statements about the future are not. Read a company post and mark every sentence as past or future. A healthy announcement is mostly past tense with a short future clause. The ones that preceded cuts in my log were the reverse: one past-tense sentence for cover, then four paragraphs of what the company is positioning itself to become. The word "focus" showed up in six of the nine.
Reading the statement in four columns
Here's the version I actually use. When a company I care about publishes something, I split the load-bearing sentences into four columns:
| The sentence | What it measures | What it leaves out | What I check next |
|---|---|---|---|
| "Record sales in our category" | Volume | Margin per unit | Whether input costs rose sector-wide |
| "We're profitable" | One quarter, one definition | Which definition, and cash on hand | Whether "adjusted" appears anywhere nearby |
| "Realigning our structure to our strategy" | Intent | Net headcount after the realignment | Open req count in 30 days vs today |
| "Doubling down on AI" | A budget moving | Which budget it moved out of | Which teams the new one is staffed from |
The fourth column is the whole exercise. Every one of those checks is something you can do from outside with a free afternoon. Req counts are public. "Adjusted" is a word you can search for. Component pricing gets covered by trade press because it affects everyone in the sector, which is exactly why a company citing sector-wide cost pressure is usually being straight with you even when the rest of the statement is positioning.
What this changes about where you send applications
Here's why I care, and why you might. You have limited applications in you. Not literally — you can send five hundred — but you have limited good ones, the kind where you read the job description twice and write something specific. Maybe eight a week before the quality falls off a cliff. I know because mine did.
A company six weeks from a restructure still has requisitions open. The posting is live. A recruiter may screen you. Some of those reqs are already dead and nobody has taken them down, and some are alive but about to be frozen mid-loop, which is how you end up doing four rounds for a role that evaporates between the onsite and the offer. That happened to me at a payments company in the fall. Nobody lied to me there either. The req was real on the Tuesday I interviewed and gone by the following Thursday.
So I stopped reading company statements to decide whether a company was good or doomed, which was never a question I could answer, and started reading them to decide where a specific application ranked in a week where I had eight good ones. Pressure signals didn't move companies to a blacklist. They moved them down the list.
The question I ask now
In the last ten minutes, when they ask if I have questions, I ask the hiring manager one thing: what changed about this team's headcount plan in the last six months?
Not "is the company doing well." That question has one answer. The headcount question is narrow enough that a manager can answer it honestly without saying anything they shouldn't, and the answers sort themselves into three piles fast. Some say nothing changed and give a number. Some say the plan got cut from four heads to one and this is the one, which is useful and often fine. And some go vague — the plan is being finalized, they're working through it with leadership — and vague on a question that specific is its own answer.
I ask it every time now. It has never cost me an offer, and twice it moved a company down my list in a week when I had somewhere better to put the effort.
Last month my own employer published a launch post that used the word "focus" four times and had exactly one past-tense sentence in it. I didn't panic and I didn't tell anyone. I spent that Sunday updating my résumé, because I've learned that the useful thing to know is not what a company means — it's which question they answered instead of the one you asked.
