Fire or keep

Strip the euphemisms. The call is: end this person's employment now, at a known cost in severance, disruption, and discomfort, or keep them, at an unknown cost that accrues weekly and is mostly invisible to you.

That asymmetry is why the call breaks the way it breaks. The cost of firing is concentrated, visible, and lands on you: one terrible conversation, a gap on the org chart, the admission that you got a hire wrong. The cost of keeping is distributed, invisible, and lands on everyone else: the A players managing around the problem, the standard quietly resetting to whatever the tolerated person delivers, the best people updating their resumes because a visibly tolerated underperformer is a statement about what the company actually requires. You will always feel the first cost more than the second. The second is almost always bigger.

Michael Seibel's version of the timing rule is weeks, not quarters, and his reasoning is the important part: the team already knows. They knew before you did. What they're watching now is not the underperformer; it's you, to find out whether the bar is real. Every week of delay teaches everyone that it's negotiable.

The record's unanimity creates its own trap: a founder who reads enough of it gets trigger-happy, and starts running exits on people who were never actually the problem. Three questions first, in order, and they have teeth:

1. Is the problem the person, the role, or you? A great person in a role that outgrew them is a move, not a firing. The VP who was perfect at 10 people and is drowning at 40 didn't get worse; the job got bigger. And a person who has never been told the standard isn't underperforming; they're uninformed, and that's your failure, not theirs.

2. Has the bar been stated, once, clearly, with a date? Not implied through sighs and reassigned projects. Said: "this is what good looks like, and I need to see it by March 1." If that sentence has never been spoken, speak it now. Once. That's the only probe this call gets, and it's a real one: a definite standard with a definite date, not a rolling hope. Coaching that never names the bar is hope with meetings.

3. Did they hit it? If yes, close this file and mean it. Closed means closed: no residual campaign, no waiting for the next slip. If no, you have your answer, and you got it in weeks instead of quarters.

If the case survives all three questions, it's real, and everything below applies.

The second failure mode isn't the delay; it's the exit done cheap or cruel. The instinct, especially when money is tight, is to treat severance as a cost to minimize and the conversation as a legal risk to script into nothing. Both instincts read the audience wrong. The person leaving will be gone in two weeks. The people watching how they leave will be running your company for years, and they are the actual audience for everything you do in that final week.

Ben Horowitz's rule for the executive case: an executive firing means you failed at hiring. So own that, out loud, in the room and to the board: "I hired for logos and didn't check for stage fit" costs you one uncomfortable sentence and buys back the credibility of every standard you set afterward. The alternative, where the departed exec is quietly narrated into the villain, fools nobody and teaches your team that failure here gets laundered instead of owned.

The documented standard for generous is Brian Chesky's May 2020 letter. Airbnb cut 1,900 people, 25% of the company, in its worst quarter ever, and the package was: 14 weeks of severance plus a week per year of tenure, 12 months of healthcare, the one-year equity cliff dropped so that even recent hires left as owners, an alumni directory the company built and staffed to get people placed. That letter circulated as a model instead of a scandal. Seven months later the company IPO'd with the trust intact. The lesson isn't that you need Airbnb's balance sheet; it's that generosity is priced as a percentage of your resources and judged as a signal of your character, and the survivors do the judging.

1. Sort the case. Employee: keep-side check, then act. Executive: same, plus you own the hiring failure explicitly. Cofounder: stop. Do not have any conversation yet. Governance first; step four exists because of you.

2. Put the conversation on the calendar, in weeks. Pick the date now, while you're reading this. Not after the next sprint, the next board meeting, the holidays. Every one of those is a reason next month will generate again. Delay has a per-week price paid in other people's morale; if you're going to pay it anyway, at least name what it's buying.

3. Design a generous exit. Severance you're slightly uncomfortable with. Healthcare bridged. References honest and right-sized: what they're genuinely good at, for the stage they're genuinely suited to. Equity treated respectfully. This call runs primarily in the people currency, and the executive case bills identity too, because "I missed this hire" said out loud is the price of keeping the standard credible. Partly decency, mostly signal.

4. The cofounder case: count votes before words. Equity distributes decision rights, not credit, and a cofounder confrontation is a governance move whether you intend it as one or not. Before any conversation, answer on paper: who can remove whom? Board seats, vesting status, protective provisions, the shareholder math. Coalitions form within days of the first conversation, and they form around whoever moved first with a plan. If you can't win the vote, the honest talk you're planning is a resignation letter with your name on it. Lawyer before the conversation, never after. This is the one branch of the call where speed loses to sequencing.

5. Run the ruin check. One question here touches the ruin line: can this person kill the company on the way out? Root access, the key customer relationships, the bank account, controlling equity. Map the single points of failure and close them first, quietly and legitimately. Not because people are villains, but because the ruin line doesn't price intentions, it prices exposure. Then act.

• Hoping the next quarter fixes what the last three didn't. It won't, and you know it won't, and the quarter you're about to spend confirming that is the most expensive way to buy information you already own.
• Loyalty accounting. Year-one heroics aren't collateral against year-three absence. Honor the past with the exit package, not with the org chart.
• The irreplaceable myth. No role survives contact with its holder's absence as badly as you fear. The AE who's been quietly carrying the missed quota is usually already doing the job.
• Cheap severance as punishment. You're not saving money; you're spending reputation in front of everyone who stays.
• Talking before the governance math is done, in the cofounder case. Days. That's how long you have before the coalition forms without you.

A firing decision that's actually been made fits on one page: the call, the date of the conversation, the keep-side check answered in writing, the package, the ruin check, and one tripwire that matters more than people expect: the review date where you grade the decision itself. Founders who write the memo discover something at the review that memory would have rewritten: the team's reaction was relief, the replacement search taught them what the role actually was, and the ten weeks of dread before the decision were the most expensive part of the whole thing.

Decided stays decided when it's written. That's what the memo is for.