Sunk Cost: Why You Cannot Kill the Thing That Is Not Working

What it actually looks like

There is a project you would not start today, and you are still funding it.

The feature took seven months. Four customers use it. You are currently scoping the improvements that will finally make it land, and you have scoped them before.

A channel has never produced a customer at an acceptable cost. The budget is still in it, because the pipeline is finally about to work and stopping now would waste everything spent so far.

You have a hire who was wrong from the second month. It is month fourteen. The reason you have not acted is partly kindness and mostly the amount of time you have already spent trying to make it work.

When you defend the thing, notice what your arguments are made of. Almost all of them point backwards — how much has gone in, how far you have come, how close you are after all this effort. Very few of them are about what the next six months would return.

You have caught yourself hoping for a reason to stop that comes from outside. A customer leaving, a budget cut, someone else calling it. That would let it end without you having to be the one who ended it.

And the decision keeps getting deferred with more information. One more test, one more month of data, one more version. The information never settles it, because information was never what was missing.

Who this happens to

This is common among people who follow through on things — which is most of the reason they have got anywhere.

It shows up in founders with a large, visible investment behind them: a year of building, other people's money, a public commitment, a story told to a board or an audience. The more people watched you commit, the more expensive stopping feels.

It shows up in anyone who has learned that quitting is a character flaw. If persistence has been your differentiator, ending something reads as becoming the kind of person you have spent your life not being.

It shows up hardest when the original decision was yours alone. Then killing the project is not just an expensive write-off; it is a public verdict on your judgment, delivered by you.

It travels closely with pivot paralysis, which is the same knot with the exit blocked from both sides — unable to double down, unable to stop — and with analysis paralysis, which supplies the endless data-gathering that makes deferral look responsible.

What sets it off

A bet that is clearly not working, but not dramatically enough to force the issue. Slow failures are far stickier than fast ones.

Any moment you have to say the number out loud — a board meeting, a budget review, a co-founder asking how it is going.

The decision to cut something with a name attached: a feature, a channel, a partnership, a person.

Pressure to justify what has already been spent. The instant the past spend needs defending, continuing becomes the cheapest defence available.

A small, late, ambiguous win. Nothing extends a doomed project like one good week arriving in month nine.

Why it keeps happening

Because stopping feels like destroying the investment, and continuing feels like protecting it. Both of those feelings are wrong, and they are wrong in the same way.

The money and the months are gone. They are gone whether you continue or stop — that is what makes them sunk. Nothing you do next recovers them. The only live question is what the next six months of your attention and capital are worth in this project compared with anywhere else, and the past spend has literally no bearing on that answer. It feels like it should. It is the most expensive feeling in business.

The reason it persists is that we are built to hate waste far more than we like gain, and stopping converts a vague ongoing situation into a definite, countable loss. Continuing keeps the loss theoretical. As long as the project is alive, the eighteen months are still an investment; the moment you kill it, they become a number you spent and did not get back. So the decision is not really about the project. It is about which day you agree to feel it.

Then there is the identity cost, which is usually the bigger one. Ending it means saying out loud that a decision you made, defended, and asked other people to believe in was wrong. Persisting lets you stay the person who was right but early. That trade — real money for a story about yourself — is made constantly, and almost never consciously.

And here is what it actually costs. Every month the failing thing stays alive, it consumes the resource that would have funded the thing that works. The loss is not the eighteen months. That is spent. The loss is the next eighteen, and unlike the first, they are still yours to direct.

What actually helps

Ask the only question that is actually load-bearing. Knowing everything I know today, with the resources I have today, would I start this now? Answer it honestly and fast — the first answer is the true one, and everything after it is negotiation. If the answer is no, you have already made the decision. What remains is scheduling.

Write the tuition receipt. Take one page and put down exactly what this project taught you: what you now know about the market, the channel, your own estimates, that you did not know before. This is not consolation. It is accurate accounting. You bought information at a high price and you own it permanently — that is what the eighteen months actually produced, and it is worth considerably more once you stop paying for the same lesson twice.

Then give it a date and a number. Not "we'll wind it down soon" — that is how a dead project survives another quarter. A specific date, a specific final spend, and one named condition that would change your mind, defined in advance so it cannot be invented later to justify continuing.

Separate the decision from the verdict. You are not deciding whether you were foolish. You are deciding where next quarter's money goes. Those are different questions, and merging them is what has kept this open for a year. The best operators kill things regularly, and it reads as judgment, not as failure, because they do it before the evidence becomes humiliating.

Say the release sentence out loud, once. "That was tuition. The lesson is paid for." It sounds slight. It does a specific job: it closes the account in your head, so the next decision is made forward-facing rather than in defence of a spend that no longer exists.