How to decide whether to kill your startup

Kill a venture when the evidence says continuing only buys information you already have. The procedure that works is to write kill criteria before you need them, review them on a fixed schedule against numbers rather than narrative, and hold a pre-committed decision meeting where the default is stop unless a specific threshold was met. The reason it has to be written in advance is that on the day, you will have a dozen reasons why this month was unusual — and every one of them will feel true.

01

Write the criteria before you need them

The right time to write kill criteria is the week you start, when you have no emotional investment in the answer. The second-best time is today, before you look at this month's numbers.

A usable criterion has three parts: a number, a date, and a consequence.

  • "If we don't have 20 paying customers by 31 March, we stop."
  • "If monthly retention is below 40% after three cohorts, we stop."
  • "If CAC exceeds LTV for two consecutive quarters, we stop."

Bad criteria are the ones that can be argued with on the day: "if it's not working by spring" or "if we're not seeing traction". Traction is a word you can redefine in the moment, and you will.

Write them where a co-founder or a friend can see them, and tell that person their job is to read the criteria back to you on the date. External accountability is the whole mechanism — a private note is one you can quietly edit.

02

Separate the evidence from the narrative

Before any decision meeting, write two lists in this order.

List one: numbers only. Revenue this month and the three before. New customers. Churned customers. Cash. Burn. Retention by cohort. Traffic, and where from. No sentences.

List two: the story you've been telling. "The market's still early." "Enterprise takes time." "We just need one more feature." "Q4 was seasonal."

Now go through list two line by line and find which number in list one supports each claim. Most won't have one. Those are the stories you've been running on, and they're not lies — they're the way a founder maintains the energy to keep going, which is a real and necessary function. They're just not evidence, and this is the one meeting where the distinction has to hold.

The six-question quiz below is a fast version of this exercise. Its usefulness is entirely in what it refuses to ask: nowhere does it ask what you've already spent.

Kill-or-Continue Quiz — free, right here

About this tool
  1. 1.Has anyone paid you actual money for this?

  2. 2.When you stop pushing for a week, what happens?

  3. 3.Of the people who tried it, how many still use it a month later?

  4. 4.Do you have one channel that reliably brings strangers?

  5. 5.Honestly — do you still want to work on this?

  6. 6.In the last month, did you learn something that changed the plan?

Answer all six for a verdict. None of them ask what you've already spent — that's the point.

03

Hold the decision meeting with a pre-set default

Put a date in the calendar. Two hours. Co-founder, advisor, or a friend who runs a business — someone who will say the uncomfortable thing.

The default is stop. Continuing requires the criteria to have been met. This inversion is the entire mechanism: when the default is continue, inertia decides, and inertia has kept more dead companies alive than optimism ever has.

Run it in this order:

  1. Read the criteria out loud, from the file, before showing any results.
  2. Present list one. Numbers only, no commentary.
  3. The other person asks: "If someone showed you these numbers for a company you had no history with, what would you tell them?"
  4. Decide. Write down the decision and the reason, immediately, in the same file.

If the decision is continue, it comes with new criteria and a new date. "Continue and see" is not a decision — it's the absence of one, wearing a decision's clothes.

04

If you continue: change exactly one thing

A continue decision that changes nothing is a deferral. If the criteria weren't met but you're continuing, something specific has to be different, and it has to be one thing.

Pick from the real levers, in rough order of how often they're the actual problem:

  • The person. Same product, different buyer. Most often correct, most often skipped.
  • The price. Doubling it changes who shows up, and frequently improves everything downstream.
  • The wedge. Same market, narrower entry point.
  • The channel. The product may be fine and unreachable.

One change, a date, and a number that has to move. Then the same meeting again. Changing four things at once means that when something moves you won't know what did it, and you'll have burned the runway learning nothing repeatable.

05

If you kill it: do it in one week, cleanly

Drawing it out is worse than doing it. One week:

Day 1 — Tell customers. Personally, with a date, and a refund if they've paid for time they won't get. This is the part people avoid and the part that determines whether they'll ever buy from you again. A quiet disappearance is remembered.

Day 2 — Export everything. Code, customer list, content, analytics. Cancel subscriptions in order of cost. Note the renewal dates you're escaping; the number is usually larger than expected and it's a small honest consolation.

Day 3 — Write the post-mortem. What you believed, what turned out to be true, what you'd test first next time, what you'd never do again. Write it while it's uncomfortable — the version written six months later is tidy and false.

Day 4 — Keep the assets. Domain, audience, relationships, the reusable code. Most second companies are built from the wreckage of the first.

Then stop for two weeks. Ideas generated in the week after a kill are reactions, not opportunities, and starting one immediately is how founders end up killing two things instead of one.

What automates this

kitty.build carries explicit kill criteria on every venture it runs, checks them continuously against real metrics, and says the word when they're met — the part a founder reliably can't do for their own company.

Feed her an idea

Questions

How do I know when to quit my startup?
When criteria you set in advance haven't been met and continuing only buys evidence you already have. Setting those criteria beforehand is the whole technique — deciding in the moment means deciding as someone with a large emotional position in one of the answers.
What is sunk cost fallacy in startups?
Counting time and money already spent as a reason to continue. Those costs are identical under both decisions, so they carry no information about which is better — but loss aversion makes an abandoned investment feel like a realised loss and a continued one feel recoverable. It isn't; it's just unrealised.
Should I pivot instead of killing it?
Pivot when the underlying observation about the world still holds and the business you wrapped around it doesn't. Kill when the observation itself turned out to be false. A pivot that keeps the same false premise is just a slower version of the same failure.
How do I tell my customers I'm shutting down?
Directly, early, with a specific end date and a refund for time they've paid for and won't receive. Offer an export of their data and, if you can, a recommendation of where to go. It costs a day and it's the difference between a graceful exit and a reputation you carry into the next company.
What should I do right after killing a startup?
Write the post-mortem within a week while it's still accurate, keep the assets — domain, audience, code, relationships — and then take two weeks off before starting anything. Ideas generated immediately after a kill are reactions to the kill rather than opportunities.

Next procedures