Sunk Cost Fallacy: What the Research Shows, What the Stoics Knew, and the Operator's Kill-or-Continue Protocol

The money, months and effort you have already spent cannot be recovered, so they have no vote on what you do next.

Every operator carries a few projects that would never get approved if they were pitched fresh today. The course nobody buys. The hire who stopped growing a year ago. The software stack you migrated to and now quietly work around. They survive for one reason: you have already paid for them.

That reason is the sunk cost fallacy. It is not a character flaw and it is not rare. It is a predictable error in how people account for the past, and like most predictable errors, it can be met with a system rather than with willpower.

This piece defines the fallacy precisely, reports what the research does and does not support, shows where the Stoics drew the same line two thousand years earlier, and gives you a four-step kill-or-continue protocol to run on any project that has started to feel like an obligation.

What the sunk cost fallacy actually is

A sunk cost is a cost already incurred that cannot be recovered. The Wikipedia entry on sunk cost states the economic rule plainly: "a sunk cost is a sum paid in the past that should no longer be relevant to decisions about the future."

The fallacy is the gap between that rule and how people behave. The same entry describes it as "a greater tendency to continue an endeavor once an investment in money, effort, or time has been made." The investment becomes the argument. The more you have put in, the harder it feels to stop.

Economists call the correct rule the bygones principle. Only future costs and future consequences matter. The entry gives a clean factory example: a project planned to cost $100 million and return $120 million, where $30 million has been spent and the value forecast drops to $65 million. Finishing costs another $70 million to get $65 million back. The right call is to stop, no matter how painful the $30 million feels.

The three currencies you sink

• Money. Retainers, software contracts, inventory, ad spend on a funnel that never converted.
• Time. The eighteen months spent building a product line, the years invested in one channel.
• Identity. The public commitment, the announcement, the version of yourself that said this would work.

Identity is usually the most expensive of the three, because it is the one you never write on a spreadsheet.

What the research shows, and where it is weaker than the headlines

The effect is real, but the evidence is more uneven than pop-psychology lists suggest. An operator should know both halves.

It shows up across species

A 2018 study in Science from a University of Minnesota team built parallel foraging tasks for mice, rats and humans. The full text on PubMed Central opens with the textbook position: "Sunk costs are irrecoverable investments that should not influence decisions, because decisions should be made on the basis of expected future consequences." Then the finding: all three species showed similar sensitivity to time already invested.

The detail that matters for operators is this one: "sensitivity to time invested accrued only after an initial decision had been made." The bias did not appear while the subjects were still weighing options. It appeared once they had committed. Deliberation was not where they went wrong. Staying was.

Replications are mixed

Several recent preregistered replications tested classic sunk cost studies on larger samples. A 2025 Registered Report revisiting a 1996 study on wastefulness avoidance, with 659 participants, reported in its abstract: "we concluded a mixed replication, with a successful replication of two of the three tested studies." People did avoid decisions that looked wasteful, but not in every scenario, and the effect sizes were smaller than the originals.

A 2023 replication of an influential study claiming the effect is weaker for time than for money, run with 821 participants, found support for that difference in one study but not in another, where "we found sunk cost effects for both money and time." The abstract is worth reading in full. For an operator the practical lesson is simple: do not assume sunk hours are safer than sunk dollars.

The closely related idea of escalation of commitment did no better. A 2025 replication of three studies found only partial support for the classic 1976 study on personal responsibility and escalation, and failed to replicate a 2007 study that linked personal responsibility and regret to escalation.

The measuring tools are noisy

Most sunk cost research uses short written scenarios. A 2025 paper tested whether those vignettes measure the same thing at all. Its abstract reports: "Internal consistency was consistently poor" across scenarios, with weak correlations between them.

Read this correctly. It does not mean the fallacy is a myth. It means that how strongly any one person falls for it, and in which situations, is not well established. You cannot assume you are immune because you passed a quiz. You audit your decisions instead.

Why operators are especially exposed

A salaried employee who backs a failing project risks a performance review. An owner-operator risks their own capital, their own calendar and their own story. Every layer of personal investment makes the past harder to write off.

• You approved it yourself. There is no committee to share the blame, so stopping reads as a personal verdict.
• You announced it. Customers, a team or an audience heard the plan. Reversing feels like losing face.
• You did the work. The hours you spent building it are vivid. The hours you would spend finishing it are abstract.
• Nobody forces a review. Without a board or a budget cycle, a weak project can run indefinitely on inertia.

The Wikipedia entry notes a fair exception: persisting can be rational when you hold private information that outsiders lack, or when abandoning has real costs of its own. The test is honesty. Is the reason to continue a fact about the future, or a feeling about the past?

The Stoics drew this line first

Stoicism treats the past as outside your control and the present as the only thing you can act on. That is the bygones principle stated as a discipline rather than an equation.

Marcus Aurelius writes in Meditations, Book II, section 14: "For a man cannot lose either the past or the future: for what a man has not, how can any one take this from him?" The past is not yours to protect. Only the present decision is.

Epictetus gives the operator a reframe for every write-off. From the Enchiridion, chapter 11:

Never say of anything, "I have lost it"; but, "I have returned it."

Applied to a project, this changes the question. Money spent on a failed launch is not lost and waiting to be won back. It has been returned. It bought you information. What remains is the dichotomy of control: the money spent is outside your control, the next dollar is not.

Seneca supplies the warning about the currency people guard least. In Letter 1 to Lucilius he writes: "Nothing, Lucilius, is ours, except time." And earlier in the same letter: "The most disgraceful kind of loss, however, is that due to carelessness." The sunk cost fallacy turns past time into a reason to spend future time carelessly. That is the trade the Stoics would refuse.

The kill-or-continue protocol

Run this on any project, hire, tool, channel or commitment that you suspect is surviving on past investment. It takes about 45 minutes per project. Run it in your weekly review on one project at a time.

1. Zero-base the decision (10 minutes). Write one sentence answering this question: if you were offered this project today, starting from where it is now, with no history, would you take it? Answer yes or no before you write anything else.
2. Price only the future (15 minutes). List what it will cost from today to finish or sustain it: money, hours per week, attention. Then list the realistic value it will return. Leave the past column blank. If you catch yourself writing "but we already spent", cross it out.
3. Name the opportunity cost (10 minutes). Write what those same future hours and dollars would buy if redeployed to your strongest existing offer, channel or client. The real comparison is never this project against nothing. It is this project against the best alternative use of the same resources.
4. Set a kill criterion and a date (10 minutes). If you continue, write the one measurable result that must be true by a specific date, and what you will do if it is not. Put the date in your calendar. If you stop, write the shutdown steps and the first one you will execute within 48 hours.

Pre-commit before the next project starts

The Minnesota finding suggests the bias attaches after commitment, not during deliberation. So set your exit rules while you are still deliberating. Before any new launch, hire or tool migration, write three lines:

1. The result that would make this worth continuing at 90 days.
2. The budget ceiling in money and hours.
3. The person or review that will hold you to it.

This is a Stoic pre-mortem with a kill switch attached. You decide how to stop before the sunk cost exists to argue with you.

Where it hides in your business, body and stack

The fallacy rarely announces itself. It shows up as loyalty, consistency or patience. Look for it in these places.

In the business

• A client who costs more in management time than they pay, kept because of the years together. The Stoic way to fire a client covers the exit.
• A content channel you have posted to for two years with no measurable leads.
• A product you keep discounting instead of retiring.

In training

The body is a business asset, and it gets the same bad accounting. You keep a program because you paid for the coaching block, or keep running through a nagging pain because you are eight weeks into a plan. The plan already happened. The question is what serves the next eight weeks. A sunk program is not a reason to load an injury.

In the AI and software stack

Tool migrations are classic sunk costs. You spent a month setting up a platform, so you build workarounds instead of admitting a different tool fits. Run the same zero-base question: if you were choosing today, with your current workflows, would you pick this tool? AI is leverage only when it is the right tool for the job you have now, not the job you had when you bought it.

Thinking one step further helps here. Second-order thinking asks what happens after you continue, and after that. Sunk cost reasoning almost never survives the second "and then what?"

Frequently asked questions

What is the sunk cost fallacy in simple terms?

The sunk cost fallacy is continuing something because of what you have already put into it, rather than because of what it will return from here. Money, time or effort already spent cannot be recovered whatever you decide next, so it should not decide the next move. A sound decision compares only future costs with future benefits, and asks what else those resources could do.

What is an example of the sunk cost fallacy in business?

A founder spends $30,000 building a course, launches it, and sells almost nothing. Instead of testing whether the market wants it, they spend another $20,000 on ads to avoid "wasting" the first $30,000. The original spend is gone either way. The only real question is whether the next $20,000 has a better expected return in the course or somewhere else in the business.

Is the sunk cost fallacy always irrational?

No. Continuing can be rational when you hold real information that a project is about to turn, when stopping carries its own costs such as penalties or broken contracts, or when finishing has value beyond the project itself. The error is only when the past investment is the reason. If your case for continuing is about future outcomes and you can state it plainly, it is not the fallacy.

Is the sunk cost fallacy the same as the Concorde fallacy?

Yes, the terms are used for the same pattern. The name comes from the British and French governments continuing to fund the Concorde supersonic jet after the economic case for it had collapsed, citing what had already been spent. Escalation of commitment is a closely related idea from organisational research that focuses on decision-makers who keep investing in their own failing choices.

How do you avoid the sunk cost fallacy?

Ask whether you would start the project today from where it stands, with no history. Price only future costs and benefits, and compare them with the best alternative use of the same money and hours. Set measurable continue-or-stop criteria with a date before you commit, so the decision is made before past investment exists to argue with you. Review open projects one at a time each week.

Spend the present, not the past

The sunk cost fallacy asks you to honour money and time that are already gone by spending more of the only resource you still hold. The Stoic answer is short. The past has been returned. The present decision is the only one in your control.

Make it a habit, not an occasional reckoning. Pick one project this week, run the four steps in your Stoic weekly review, and either set its kill date or execute its first shutdown step within 48 hours. The capacity you free up compounds.

If you want the daily structure that makes reviews like this automatic, start the free 5-Day Stoic Operator Challenge. Five days, one principle each day, built for operators who want clarity in their decisions and discipline in their execution.