Better.com, Google, Meta got the layoff math right, the delivery wrong

By Paul Walker
4 min read
layoffsAIworkforce restructuringcrisis communicationsOracleMetaPayPalGoogleexecutive decision-making

Better.com fired 900 people on a three-minute Zoom. Google cut 12,000 by shutting off their badges overnight. Meta cut 8,000 in a year it earned $201 billion. Same action, three deliveries, and every time the how became the story, not the why.

I ran the 2026 layoff wave through CrisisCommand® to see which announcements created legal and reputational exposure and which contained it. The pattern was clean: the companies that lost control lost it at the moment of delivery, not the moment of decision.

Better.com decided to say it fast. On December 1, 2021, CEO Vishal Garg told about 900 employees, roughly 9% of staff, that if they were on the call they were being let go. Three minutes, one Zoom, days before the holidays and just after a fresh cash infusion. He later accused departed staff of stealing time. The cut was defensible. The delivery made Garg the permanent symbol of how not to do it, and the reputational damage outlived the restructuring.

Google decided to cut the access first and explain later. On January 20, 2023, Alphabet ended about 12,000 roles, its largest layoff, and many long-tenured employees found out when their access died overnight. Sundar Pichai said he took full responsibility. It did not matter. The overnight cutoff, at a company that sold itself as people-first, drew protests and organized pushback from the Alphabet Workers Union, and it spent down the exact culture that recruited the talent.

Meta decided the timing did not matter. In May 2026 it cut about 8,000 people, roughly 10%, and moved another 7,000 into AI teams, in a year it reported $201 billion in revenue and $22.8 billion in a single quarter's profit. It called the cuts structural, not performance. Cutting into record numbers while spending north of $115 billion on AI infrastructure told employees the reduction was a choice about priorities, and it turned every future "we had no option" into a line they no longer believe.

The decision to reduce headcount is not the wound. The wound is the gap between the decision and its delivery: the channel, the timing, the words, and the story told to justify it. A layoff is a governance event that produces a legal record, a press cycle, and a workforce that stays. Better.com, Google, and Meta each got the math roughly right and lost on the delivery.

The 2026 wave is running the same play at larger scale. Oracle shed about 21,000 roles, roughly 13% of its people, and booked $1.8 billion in restructuring costs, much of it to redirect cash toward AI data centers. PayPal moved to cut a fifth of its workforce. This lesson has been paid for in full, in every layoff on the list above. It still never reaches the room where the next number is set. When the email lands at 4 a.m., or names AI as the reason while the quarter prints a record, the people who remain learn how they will be treated the day they become inconvenient, and that lesson sets retention, organizing, and the next hiring cycle long after the severance clears.

The failure is sequencing. Finance sets the number, an internal date leaks or an access system flips, communications writes the note last, and legal reviews the tone after it is already set. By the time anyone asks how this will read to the people staying, the badges are already off. The fix is pre-commitment: decide the delivery, the channel, and the justification before the number is locked, and test the announcement against the record it will create.

What works

  • Deliver the news like it will be screenshotted. It will be. Write the announcement for the employee who stays and the reporter who quotes it, not for the board deck.
  • Separate the decision from the excuse. If the cut is a bet on AI, say that plainly. Do not launder a priority choice as an unavoidable correction.
  • Sequence the delivery before finance locks the number. Channel, timing, and words are part of the decision, not a downstream task.

What does not work

  • Speed standing in for clarity. Better.com's three minutes were fast and cost years.
  • Letting a system break the news. An overnight access cutoff is a message, and it says the company could not be bothered to tell you itself.
  • Cutting into record results and blaming forces outside your control. The numbers are public. The contradiction becomes the headline.

Better.com got the number right and made its CEO a punchline. Google got the number right and taught its own workforce to organize. Meta got the number right and spent the trust it will need the next time it says there was no choice. The layoff was never the crisis. The crisis was everything the company decided after the number, and by the time each of them noticed, the decision had already been made for them.

Commentary on public reporting. CrisisCommand does not represent the companies named. CrisisCommand is decision-support software, not legal advice.

Paul Walker headshot

Paul Walker

Founder

Veteran strategist with a career spanning PulsePoint Group, Accenture, Y&R/Burson-Marsteller, Cohn & Wolfe, and The University of Texas. Paul has built and led businesses across the U.S., Asia, and Europe — from startups to major universities to Global 1000 companies.

Ready to Get Crisis-Ready?

See how CrisisCommand can help your organization prepare for and manage crisis communications with AI-powered intelligence.