Sunk Cost

011 min

Sunk Cost at a glance

• What it is: Money, time, or effort already spent and gone, no matter what happens next.
• The rule: A rational decision weighs only the costs and benefits still to come.
• The mix-up: Sunk cost is the accounting fact. The fallacy is acting on it.
• Named case: Governments kept funding the Concorde programme years after its costs made the case uneconomical.

021 min

The problem Sunk Cost names

A product team spends eight months and a fixed budget building a redesign. Partway through, user testing shows the new design tests no better than the version it replaces, and a simpler fix would solve the original complaint in two weeks. The honest math says stop the redesign and ship the simpler fix. Most teams do not stop. The eight months already spent becomes an argument for finishing, even though that time is gone either way and the only real question is what the remaining budget should buy.

This is the failure sunk cost accounting exists to prevent: judging a project by how much money has already been spent on it, rather than by what finishing it would still cost and still return. The money already spent is gone in every version of the future. Only the money not yet spent is still a choice.

032 min

Understanding Sunk Cost

In economics and business decision-making, a sunk cost is a cost that has already been incurred and cannot be recovered. That is the field's own definition, and it matters because of what it excludes: a sunk cost carries no interest, no depreciation, and no ongoing claim on the future, and no decision made today changes that. Decision theory calls the rule that follows from this the bygones principle: a rational choice compares only the costs and benefits that are still ahead, and treats every dollar, hour, or unit of effort already spent as irrelevant to that comparison, whether the amount was five dollars or five million.

This is a rule about the correct way to make a decision. It is not a description of how people actually decide, and the gap between the two is large enough to have its own name: the sunk cost fallacy, the tendency for a past, unrecoverable cost to shift a present decision toward whichever option already has more money in it. Sunk cost is the accounting fact; the fallacy is what a mind does with that fact when the fact should carry no weight at all. An organization can have a correct sunk cost policy written down and still fall for the fallacy in every budget meeting, because the failure is behavioral, not procedural.

Two forces make the fallacy hard to resist. Loss aversion makes walking away from a project feel like locking in a loss, even though the loss already happened the moment the money was spent. A framing effect compounds it: the same remaining decision reads differently depending on whether it is presented as finishing what was started or as spending money again starting today, even though the arithmetic is identical either way.

041 min

Where Sunk Cost comes from

Sunk cost has no single named originator. It is part of the bygones principle, sometimes called the marginal principle, a standard result of rational choice theory that predates any single paper: a decision should weigh only the costs and benefits that are still avoidable, because a cost that is already fixed cannot be avoided by any current choice. Economists were applying this logic long before anyone tested whether people actually follow it.

The behavioral question came later, from two directions. Barry Staw's 1976 study, "Knee-Deep in the Big Muddy," was the first to test it inside an organization, using a simulated $10 million research-and-development decision with 240 business students. Nine years later, Hal Arkes and Catherine Blumer gave the individual-level failure its lasting name in "The Psychology of Sunk Cost" (1985). In one of their field studies, season-ticket holders who paid the full price for a theater series used significantly more tickets over the first half of the season, an average of 4.11 tickets, than holders who had received a $2 or $7 discount, 3.32 and 3.29 tickets, even though every ticket was already paid for and equally unrecoverable.

052 min

The numbers

Take a hypothetical case with round numbers. A company has already spent $5 million building a custom internal tool. Finishing it needs another $2 million. A ready-made alternative that does the same job costs $1.5 million to buy and install. Every number here is hypothetical, chosen only to make the arithmetic easy to follow.

The $5 million is a sunk cost the moment it is spent, whichever option the company picks next. It does not belong in the comparison at all.

Compared correctly, the ready-made alternative costs $500,000 less from this point forward, and the company should switch, even though switching means the $5 million already spent bought nothing that ships. Compared incorrectly, the company treats the built tool as five million in, only two million left, which makes finishing look cheaper than it is, and makes switching look like it wastes five million dollars instead of the two million that switching would actually cost on top of it.

People choose the second framing more often because it feels more responsible, not because the arithmetic is unclear. Nothing about finishing the custom build returns any part of the original five million dollars; it stays spent under every option on the table.

061 min

A second case

The clearest real-world case is the Anglo-French Concorde programme, the source of the term "Concorde fallacy" for this exact mistake. Britain and France signed a treaty on 29 November 1962 to jointly build a supersonic passenger jet, with an original programme cost estimate of £70 million. By the time Concorde entered commercial service on 21 January 1976, delays and cost overruns had pushed the total programme cost to between £1.5 billion and £2.1 billion, roughly twenty to thirty times the original estimate, and the aircraft never earned back what it cost to develop.

What makes Concorde the case worth naming is not the overrun by itself. It is that both governments kept funding the programme through the 1960s and 1970s after the commercial case for it had visibly weakened, in part because backing out would have meant writing off the hundreds of millions already committed. The 1962 treaty added a second, separate reason to keep going: it included a clause, originally asked for by the UK government, imposing heavy penalties for cancellation. That clause matters on its own, because it means part of Concorde's continuation was a real contractual cost of quitting, not only the sunk cost fallacy, a distinction the section on when applying this goes wrong returns to.

072 min

What this means for your work

A product manager six months into a redesign, watching usability tests come back flat, faces a real decision: not whether the redesign was worth it, which is unanswerable and irrelevant, but whether finishing the remaining two months costs less than switching to the smaller fix, given what both are now expected to return. The six months already spent answers neither question.

A founder who has personally invested a year and most of their savings into a product with flat growth faces the same test in a harder form, because their own sunk cost is also their income and their identity. The forward-looking question does not change: given only what it would cost to keep going from today, and what that spending is now expected to return, is this still the best use of the money that has not yet been spent.

An engineer three weeks into debugging a broken data migration, with a clean rewrite now visibly cheaper than finishing the fix, faces the smallest version of the same choice, and it is usually the easiest one to get right, because nothing except the three weeks already spent argues for finishing.

In all three cases, the practical move is the same: separate what has been spent from what remains to be decided, and make the remaining decision as if today were the first day, using only the costs and benefits that are still avoidable.

081 min

How to apply Sunk Cost

The reliable check is a single forward-only question: knowing everything you know now, if none of the money, time, or effort already spent existed, would you still choose to spend what remains? If the honest answer is no, the amount already spent is not a reason to say yes anyway. The check works because it removes the one input that should never have been part of the decision.

Two adjustments make the check hold up in practice. First, run a cost-benefit analysis of what remains on its own, before looking at what has already been spent on the project, so the number in your head is never anchored to it. Second, when the decision is significant, hand it to someone who was not involved in the original choice. Staw's organizational studies found that the person who made or defended the original decision is the one most likely to keep funding it past the point the numbers justify, because reversing the decision would also mean admitting the earlier one was wrong.

Neither adjustment requires ignoring genuine reasons to continue, such as a contract, or a real chance the situation improves. It only removes the one reason that is never genuine: what has already been spent.

091 min

When applying Sunk Cost goes wrong

Treating every sunk cost as a reason to quit is its own mistake. Quitting has real costs of its own, and when those costs are the actual reason to continue, that is not the sunk cost fallacy. The fallacy is specifically continuing because of the past spend, not for a separate, forward-looking reason that happens to point the same way.

Three forward-looking reasons are genuine. A contract can make quitting expensive on its own terms: large infrastructure and construction projects routinely carry penalty clauses, redundancy payments, and lease obligations that apply whether or not the original spending was wise. A damaged reputation is a genuine cost too: organizations known for walking away from commitments face worse terms in future negotiations, a cost that has nothing to do with what was already spent. And persistence sometimes produces information a quick exit would not: Apple's Lisa, a $10,000 computer that failed commercially in 1983, still produced the graphical interface Apple used in the Macintosh the next year.

The hard part is not the rule itself. It is telling a genuine forward-looking reason to continue apart from a disguised sunk cost, in the moment, before any more money is spent.

102 min

Sunk Cost vs. nearby concepts

Sunk cost is most often confused with the sunk cost fallacy itself. They are not interchangeable: sunk cost names a category of cost, and the fallacy names a specific error in judgment about that category. An organization can talk fluently about sunk costs and still commit the fallacy in the next meeting, because knowing the term is not the same as applying the rule when a real decision is on the table.

Sunk cost is also confused with opportunity cost, which points in the opposite direction. Sunk cost looks backward, at money, time, or effort already spent that no choice can recover. Opportunity cost looks forward, at the value of the next-best option a person gives up by choosing one path over another. The deciding question for sunk cost is whether the cost can still be recovered. For opportunity cost it is what a person is giving up by choosing this option instead of another.

112 min

Frequently asked questions about Sunk Cost

?7 questions

Questions people ask

§7 sources

Sources