Sunk Cost: The Mental Model for Pricing the Past Out of a Decision

What is a sunk cost?

A sunk cost is a resource already spent that you cannot recover, no matter what you choose next. The mental model is simple to state: since the past is fixed for every option, it cancels out, so the only thing worth comparing is what each path costs and returns from today onward.

That makes sunk cost less a fact and more a discipline. You draw a line across your timeline at “now”. Everything left of the line is history. Everything right of the line is the decision.

This page is about using sunk cost as a tool. If you want to understand why people keep pouring money into failing projects, that is the sunk cost fallacy, the bias this model exists to correct.

Three things count as sunk:

• Money that is spent and not refundable or resellable.
• Time already given to a job, a degree, a codebase or a relationship.
• Effort and reputation already committed in public, like a launch you announced.

Three things that look sunk but are not:

• Salvage value: what you could sell, reuse or refund.
• Learning: skills and information you keep whatever you decide.
• Future obligations: a contract you still have to pay is a future cost, not a sunk one.

Where it comes from

The idea comes from economics, where the textbook rule is that rational choices depend only on marginal (future) costs and benefits. A cost that is the same under every option cannot help you choose between them.

Psychologists later showed how hard that rule is to follow. Hal Arkes and Catherine Blumer published “The Psychology of Sunk Cost” in 1985, a set of experiments in which people kept investing in options mainly because they had already paid for them. Evolutionary biologists had a related label, the “Concorde fallacy”, named after the supersonic airliner that the British and French governments kept funding after its commercial case had weakened.

So the model has two halves: an economic rule (ignore what can’t be changed) and a behavioral warning (you will find that rule hard to apply). Pair it with opportunity cost, which supplies the other half of the math: what the remaining resources could do elsewhere.

How to use sunk cost

Use these steps whenever you catch yourself saying “but I’ve already put so much into this”.

1. Draw the line at today. Write down what has been spent so far: money, months, effort. Label it “sunk”. You are not denying it; you are moving it out of the comparison.
2. Separate what is recoverable. List anything you can sell, refund, reuse or carry forward: equipment, skills, code, contacts, data. That is salvage, and it is fair game.
3. Rewrite the choice as two futures. “Continue” means: future cost X, expected return Y. “Stop” means: future cost A (exit fees, switching), expected return B (what the freed resources can do). Only these four numbers matter.
4. Ask the fresh-start question. If someone handed you this exact situation today, with nothing invested, would you choose to continue? If the answer is no, your history is the only thing holding you in.
5. Check the reversibility. Quitting is often easier to undo than it feels. See reversible decisions: if you can pause rather than kill, the bar for stopping drops.
6. Set a stop rule for next time. Before the next phase starts, write down the signal that would make you stop, and a date to check it. A rule written early is not contaminated by later spending.

A useful habit is to keep these entries in a decision journal, so you can see later whether your “continue” calls were made on future value or on past pain.

3 real-world examples

Business

Say a startup has spent nine months and most of its seed money building a feature for a customer segment that never signed up. The team argues for three more months “so the work isn’t wasted”. Applying the model, the founder asks only: what do three more months on this feature return, versus three months on the segment that is actually buying? The nine months are gone either way. The code that can be reused, and what the team learned about the market, carry over. The rest does not get a vote.

Personal life

Imagine you are two years into a degree you no longer want. The tuition and the two years are sunk. The real question is the next two years: what does finishing cost and return, compared with switching or leaving? Some of the past counts as salvage (credits that transfer, skills, the network). Framed this way, “I can’t quit now, I’ve come this far” turns into “is the remaining half worth it on its own?” Sometimes the answer is still yes, and that is fine. The point is to say yes for forward-looking reasons.

Investing

An investor bought a stock at $80 that now trades at $40. The purchase price is sunk; the market does not know or care what you paid. The live question is whether, with $40 in cash today, you would buy this stock over everything else available. If not, holding it is the same as buying it again at $40. Taxes are a legitimate exception that is not sunk: realizing a loss may have real future value, so it belongs in the comparison.

Common mistakes

• Treating learning as sunk. What you learned is not lost. Ignoring it can push you to quit things that are now cheaper to finish than to restart elsewhere.
• Confusing sunk with fixed. A gym contract you still owe is a future cost. You can’t escape it by quitting, but it is the same under both options, so it cancels too. Check which costs actually differ.
• Overcorrecting into quitting everything. Ignoring sunk costs does not mean abandoning projects at the first setback. Many good projects look bad halfway. The model says judge the future, not “always stop”.
• Ignoring switching costs. Leaving has its own future price: retraining, exit fees, lost momentum. These are not sunk and must be counted.
• Doing the math once. Sunk costs grow as you go. A choice that was right at month three may be wrong at month nine, so recheck on a schedule.

Sunk cost vs sunk cost fallacy

The Mentarium take

Knowing that sunk costs are irrelevant almost never stops anyone from honoring them. Most people can recite the rule and still finish the bad movie. What works better is structure decided before the money grows: a stop rule written at the start, a review date, and a record of what you predicted.

That is why we put sunk cost at the Review stage of The Mentarium Decision Loop (Frame → Map → Stress-test → Decide → Review). When you Decide, write the prediction and the signal that would make you stop. When you Review, compare the result with the prediction and ask the fresh-start question out loud. The sunk cost fallacy sits on the Bias Radar for career, relationship and business decisions for a reason: it shows up exactly where you have invested the most of yourself.

One more test we like: imagine your successor walks in tomorrow with no history and no ego in the project. What would they do first? If you can answer that quickly, you probably already know what to do. For bigger calls, borrow from the regret minimization framework: ask which choice your 80-year-old self would regret more, continuing or stopping, and leave the receipts out of it.