The worst all-hands I ever sat through ran ninety minutes and left me with exactly one piece of information: our new office would have a nap pod. Revenue was down that quarter, two big customers were wobbling, and half the engineering org was quietly updating their CVs. Nobody mentioned any of it. We got slides about the nap pod, a video montage set to licensed indie rock, and a raffle. I remember thinking that if this was the best signal leadership could send about where we stood, the silence about everything else was the actual message.
I've run all-hands at three companies since, in payments and lending, where a bad quarter isn't abstract and regulators are reading over your shoulder. I've learned that the meeting is almost never the problem. The problem is that people treat it as a broadcast when it should be a contract. Here's how I think about making the hour worth the salaries in the room.
Decide what the hour is actually for
Most all-hands fail because nobody can say, in one sentence, why they exist. So they become a dumping ground: a bit of strategy, a bit of celebration, a product demo, a values reminder, a birthday shout-out. The result is a meeting that does six things badly and nothing well. If you can't name the single job of your all-hands, you're going to fill the vacuum with theatre.
My answer is that an all-hands exists to align people on reality and reduce the number of anxious hallway conversations happening after it. That's it. Everything else is secondary. When I plan one, I ask whether a given segment reduces uncertainty or increases it. A confident, vague statement about "exciting momentum" increases uncertainty, because people know it's hiding something. A specific number, even a bad one, reduces it. That test kills about a third of the slides before we start.
Show the numbers before you spin them
I open with the same three or four metrics every single time, in the same order, in the same format. For us that's monthly recurring revenue, net revenue retention, gross transaction volume, and a rolling 30-day incident count. Boring on purpose. The point of consistency is that people learn to read the dashboard themselves. When the same chart appears month after month, a dip is legible to everyone, not just the finance team.
The temptation is always to lead with the story and let the numbers support it. I do the opposite. Numbers first, on screen, unspun, and then I tell you what I think they mean. If MRR grew four percent but two-thirds of that came from one enterprise deal we're not sure will renew, I say that out loud. People are not stupid. They can smell a curated number from across the room, and the moment they catch you presenting a flattering slice, they stop believing the flattering slices that are actually true.
Deliver the bad news yourself, early, in plain words
The single highest-leverage thing a leader does at an all-hands is name the hard thing before someone forces them to. When we lost a top-five merchant to a competitor last year, I put it on slide two, said the number it cost us, and said what we were changing because of it. It was uncomfortable for about forty seconds. What it bought was a room full of people who then believed me for the rest of the hour.
Compare that to the alternative, where the news leaks through a Slack channel or a departing colleague, and by the time you address it three weeks later you're not delivering information, you're doing damage control. The story has already been written without you, and it's worse than the truth.
If your team finds out about the important thing from someone other than you, you have not just lost an information race. You have taught them that the all-hands is where the company performs confidence, not where it tells the truth.
Give every segment one owner and one message
I've seen all-hands where six people present because six people asked to present. That's how you end up at ninety minutes. Now I'm ruthless about it. Each segment has exactly one owner and I make them tell me the one thing they want people to remember before I let them near a slide. If they can't compress it to a sentence, the segment isn't ready and it doesn't go in.
This does two things. It forces clarity on the presenter, and it protects the audience from the death-by-committee effect where three VPs each say a slightly different version of the same update and nobody knows which one is real. A real example: our security lead used to present a nine-slide compliance update. Now she gets ninety seconds and one message, usually something like "our SOC 2 audit is on track and here is the one control that's behind." People retain that. They retained nothing from the nine slides.
Enjoying this article?
Get more like it in your inbox — practical engineering leadership, fintech, and AI. No spam, unsubscribe anytime.
Treat the Q&A as the meeting, not the afterthought
The presentation is the warm-up. The questions are the event. I protect at least a third of the runtime for them, and I've stopped pretending the polite silence at the end means everyone's satisfied. Silence usually means people don't trust that asking is safe, or they've learned the answers are useless.
So I seed it. I collect questions anonymously beforehand, and I deliberately answer the sharpest, most uncomfortable one first, verbatim, before anyone has to raise a hand. "Someone asked whether the hiring freeze means layoffs are coming. Here's the honest answer." Once people see that the pointed questions get real answers, the live questions start flowing. Here's what I've found actually generates good questions:
- An anonymous submission form open all week, not just a link dropped the morning of.
- Answering the hardest submitted question first, not saving it for when time has run out.
- Saying "I don't know, I'll find out by Friday" when it's true, and then actually following up on Friday.
- Never punishing a question with a defensive answer, because the whole room is watching what happens to the person who asked.
Close the loop on last time
Nothing corrodes trust faster than an all-hands that never references the previous one. If you promised in April to fix the on-call rotation and you never mention it again, everyone learns that promises made in this room evaporate. So I keep a short running list of commitments made from the front of the room, and I spend the first five minutes of each meeting on what happened to them.
Some of those updates are admissions. "Last month I said we'd have the new expense policy out by now. We don't. It slipped because finance and legal disagreed about approval limits, and here's when it'll actually land." That's not a fun slide to present. But a leader who reliably reports on their own missed commitments is worth ten who only report wins, because you can plan around the first kind and you can't trust a word from the second.
Respect the clock like it's real money
Do the arithmetic once and it changes how you plan. A two-hundred-person all-hands at a loaded cost of roughly a hundred dollars an hour per head is twenty thousand dollars a session. If you run it weekly, that's a million dollars a year of attention. I would never approve a vendor contract that size without knowing precisely what it delivered, and yet companies run seven-figure meetings on autopilot because "we've always done it monthly."
My default is monthly, forty-five minutes, and I'd rather end ten minutes early than pad to fill the slot. If there isn't forty-five minutes of genuinely useful content, I cut it to twenty and give people the time back. The instinct to fill the calendar slot because it's booked is the same instinct that produced the nap-pod meeting. Length is not a proxy for importance, and treating the audience's time as free is the fastest way to teach them the meeting is optional.
Design for the person watching the recording
Half my org is remote, and a meaningful fraction never attends live because of time zones. For a long time we designed the all-hands for the people in the room and treated everyone else as an audience of ghosts squinting at a laptop camera pointed at a stage. That's backwards. The median attendee now watches a recording at 1.5x speed three days later, and the meeting should be built for them.
Concretely, that means no "you had to be there" energy, no jokes that only land for the people who can see the room, and slides that make sense without the voiceover. It means posting the recording, the deck, and a three-bullet summary within a day, because the person catching up on Thursday deserves the same information as the person who was there Tuesday. If your all-hands only works live and in person, you've built it for a shrinking minority of your company.

The meeting is the message
The uncomfortable truth is that your all-hands is already telling people exactly how much you respect them, whether you mean it to or not. A meeting that hides the bad numbers, dodges the real questions, and runs long out of habit says one thing. A meeting that opens with the ugly metric, answers the question nobody wanted to ask, and gives people ten minutes back when there's nothing to say says something else entirely. People believe the second signal far more than any values slide. Run the meeting you'd actually want to sit through as a skeptical engineer in row four, and most of the hard decisions make themselves.
Get new posts in your inbox
Occasional, practical notes on engineering leadership, fintech, and building with AI. No spam, unsubscribe anytime.
Comments (0)
Leave a Comment
No comments yet. Be the first to comment!
