Too Vague, Too Wrong, Just Right: What Three Memos Teach Us About Delivering Hard News
Leadership tends to fail when it comes to communicating hard news in one of two ways: it says too little, or it says something untrue. Amazon, Meta, and Uber each had to deliver difficult news to their workforce in the past two years. Amazon announced a five-day return to office and a push to cut managers. Meta and Uber announced job cuts. Same kind of moment, three very different outcomes.
Amazon: when vague invites a theory
Since 2023, Amazon had spent years managing a version of hybrid work that most employees had come to treat as permanent. Three days a week in the office wasn't universally loved, but it was settled enough that people had rebuilt their lives around it (commutes reorganised, childcare scheduled, side projects fit into days at home). A meaningful chunk of the workforce had joined during that era; for them, hybrid wasn't a temporary accommodation; it was simply how Amazon worked.
Then in September 2024, Andy Jassy sent a memo titled "Strengthening our culture and teams." It opened not with the news but with a long reflection on Amazon's history — the idea that the company should still feel like "the world's largest startup" twenty-seven years later. Buried inside that narrative were two significant decisions: every org would need to raise its ratio of individual contributors to managers by at least 15% by the end of Q1 2025, and starting January 2, 2025, employees would be required back in the office five days a week.
With one reason covering both moves: CULTURE. Faster decisions, closer mentorship, a return to how things used to work.
That is a lot of weight for one abstract word to carry, and it points to one of the problems when it comes to a multi-audience: words mean different things depending on where you sit. The further someone is from a decision, the more abstractly they think about it. The closer they are to doing the work, the more concrete they need it to be.
Psychologists call this construal level theory. “Culture” may motivate at the CEO altitude it was written, but at a Frontliner's desk on Monday morning, it is neither an instruction nor a reason.
Readers fill that space with whatever feels most plausible. A memo that justifies full-time office attendance by pointing to bloat and inefficiency reads, to an anxious employee, like a way to shrink headcount without the cost of severance. A Blind survey found that 73% of employees said they'd considered quitting after the announcement. Within a day, Slack channels and social platforms were full of the same theory: this is a layoff in disguise.
Reporters picked up the thread. At an all-hands soon after, Jassy was asked directly whether the mandate was a backdoor layoff, and he denied it on the record in front of the whole company. That denial only had to happen because the memo left the question open.
One clear sentence, “this is not a cost-cutting measure, and here is why,” would have taken the rumour off the table before it spread. Instead, Amazon spent weeks re-litigating a story it could have preempted.
Meta: when the explanation is wrong
In early 2025, Zuckerberg told employees Meta would “raise the bar on performance management and move out low performers faster.” The cuts were framed, explicitly and publicly, as performance-based.
The mechanics made that label more dangerous than it sounds on paper. Rather than asking managers to identify genuinely underperforming employees, leadership set a quota: every team had to sort 12–15% of its people into a lower tier, whether the team was struggling or thriving. A high-performing team with no one who deserved a low rating still had to produce that percentage from somewhere.
Employees who had strong recent reviews found themselves swept into the low-performer category. Kaila Curry, a former content manager, posted on LinkedIn that she'd received an “exceeds expectations” rating at her midyear review, had never been placed on a performance improvement plan, and had never been given corrective feedback. “I am not a low performer,” she wrote. Others came forward with similar stories.
For these people, the layoff wasn't the worst part. The worst part was that the company's official, public explanation for why they lost their job was untrue, and there was no way to correct the record. A severance check doesn't undo a professional reputation.
Vagueness at least leaves room for people to write a more generous story. A false, specific label closes that door and asserts something untrue about a person's work, in public. Meta's problem wasn't too much detail. It was the wrong detail, handed to thousands of people at once.
Uber: the true reason, early
In September 2026, Uber cut roughly 10% of its workforce under the banner “Building a simpler, faster Uber.” CEO Dara Khosrowshahi's memo did three things most leaders skip:
1. It led with the news instead of wrapping it in preamble. The scale of the cuts and the reasoning (fewer management layers, consolidated micro-teams) came first, with no history lesson and no build-up. Suspense only gives people time to write their own story.
Then came the sentence that did the real work:
2. It answered “was this about me?” before anyone asked. One sentence said the changes were about how Uber is organised, “not about anyone’s contributions.” That’s the question every reader asks first, and the memo got there first.
"It's important to say that these changes are about how we're organized and what we're prioritizing, not about anyone's contributions to Uber, which we will always value."
That single line did two things Amazon and Meta each failed to do make the reader feel valued.
3. It said the hard question out loud.“Why, and why now?” The memo went as far as to acknowledge the business was performing well, which is exactly why people would ask. Then it gave the reason: growth had piled on layers, coordination, and fragmented ownership, making success the explanation, not the contradiction.
Structurally, it hit the target Amazon and Meta missed: specific enough to close the door on speculation, careful not to open a false one in its place, and addressed to its audience in the first few sentences.
Why it's hard to deliver hard news
When something bad happens to us, we look for a cause. Nobody sits comfortably with “I don't know why this happened to me.” The brain wants a reason, fast, and if the message doesn't supply one, we build our own. The reason we build is almost always about us, not the market, not the org chart, not bad luck. This is attribution theory, and it explains all three outcomes. Amazon left the cause open, so readers wrote the worst one available. Meta supplied a cause that contradicted what people knew about their own work, so it landed as an accusation. Uber supplied the true cause, early, in a sentence built to intercept the self-blame reflex.
Same psychological mechanism, three different collisions.
Run it through F.A.S.T. before you send
The F.A.S.T. Framework is a pre-send diagnostic: four checks to run on a draft before it goes out.
F: Frame it
Lead with the news, not the narrative. Amazon put company history first. Put the decision in the opening lines and let context follow. Suspense only breeds speculation.
State what this is NOT, not just what it is. The absence of a denial reads as confirmation. If there's an obvious wrong conclusion people could draw (layoff in disguise, a verdict on performance, a sign of instability), name it and rule it out.
Check that the why is concrete. Ask two people at the same level what changes for them on Monday. If their answers differ, the why wasn't specific enough.
A: Anchor it
Write for the person with the most to lose, not the most context. The exec team already knows the rationale. Draft for the employee three layers down.
Quote the audience back to itself. Uber's memo worked partly because it used language employees had already used. Pull from engagement surveys, skip-levels, or exit interviews before you draft.
Equip managers before the memo goes wide. Give them the memo, the likely questions, and suggested answers an hour or a day ahead, so the people translating it aren't guessing in real time.
S: Simplify
Run a jargon audit. Highlight every internal-only term (s-team, PxT, GMs, “two-way doors”) and cut it or define it in line. Insider language is a tell that the memo was written for the room, not the recipient.
One reason per decision. Amazon stretched a single word, “culture,” across two hard asks, and it stopped being believable for either. Give each decision its own reason, even if the memo gets longer.
Read it out loud, cold. Jargon and hedging are easier to hear than to see. If you stumble on a sentence, so will the reader.
T: Travel
Draft the cynical headline first. Write the worst-faith, one-line reporter's summary of your own memo (“Company uses fake performance excuse to cut costs”). If it's more compelling than the true version, try again.
Watch where it goes after you hit send. Monitor Slack, Blind, town halls, and the first round of manager questions. That is your fastest signal on whether the intended why survived, and whether a quick follow-up (“to clarify one thing…”) is needed before a wrong story hardens.
Uber's sentence works because it fills that gap before the reader can fill it themselves. Ask these two questions before you send
1. If someone reads only the first two sentences, do they know this isn't a verdict on them? If the reassurance sits further down, it's too late. The emotional read happened in sentence one.
2. What would a person with zero context, reading this cold, repeat back to a friend? Find that reader, someone outside the room where the decision was made, and ask them before you send. If their answer is the cynical one, rewrite your first paragraph.
Want a Messaging Autopsy?
After seeing years of good companies struggling with their message, I created the F.A.S.T Framework . It isn't just a lens for reading a memo after it's landed. It's a pre-send diagnostic: run your draft against Frame, Anchor, Simplify, and the Travel test before it goes out, and ask what version a hostile reader — or a hostile reporter — would repeat back. If that version is worse than the truth, you haven't finished writing it yet.
Download the test sheet and see how well your messaging lands.
