Sunk costs are past expenditures of time, money, or effort that cannot be recovered. In economic logic, they are irrelevant to the current decision. What matters is the marginal path forward: the future costs and benefits of each option. The $10m already spent on a failing project doesn't disappear if you shut the project down — but it doesn't increase if you keep going. The only question is whether the next dollar and the next hour produce more value in this project or somewhere else.
The principle is simple. Rational choice depends on comparing the expected outcomes of available actions from this moment on. Sunk costs are the same regardless of which action you take; they cancel out of the comparison. So they should not tilt the decision. In practice, people and organisations routinely let sunk costs dictate continuation. "We've come this far." "We can't waste what we've invested." The emotional and political weight of the past distorts the marginal calculus. The discipline is to separate "what we've spent" from "what we'll get if we continue" and to decide only on the latter.
The strategic use: before committing more resource to any project, product, or strategy, ask what the incremental return is. If the next unit of investment has a higher return elsewhere, reallocate. The sunk cost is not a reason to continue; it is a reason to be clear-eyed about the marginal decision so you don't compound the error.
Salvage value is the exception. If stopping or switching frees resources that can be used elsewhere — selling equipment, reassigning people, repurposing code — that recovery is a future benefit of the exit option. It belongs in the marginal comparison. Only the part that cannot be recovered under any choice is truly sunk. In practice, people overstate how much is sunk (they forget salvage) and understate how much is salvageable (they don't look for it). Both errors bias toward continuing.
Example in practice. A company had spent three years and $15m building an internal ERP. The system worked but was costly to maintain and limited flexibility. The CFO argued for one more year to "get value from what we've built." The right question: from today, is the expected return from the next dollar and the next engineer-year higher in fixing the ERP or in migrating to a best-in-class SaaS product? The $15m is sunk. The migration cost is a future cost of the "quit" option; the benefit is future flexibility and lower TCO. The marginal comparison favoured migration. They migrated; the sunk cost was written off; the business improved.
Section 2
How to See It
You see sunk costs in play when past investment is cited as a reason to continue. A team that has spent two years on a feature argues for one more quarter. A fund holds a position "because we're down so much." A leader keeps a failing division "because we've invested so much in it." The diagnostic: strip out the past. Would you start this today with a clean slate? If not, the past is acting as a sunk-cost anchor.
The clean-slate test is the quickest filter. If you wouldn't start it today, the only reason to continue is if stopping has a future cost (e.g. contractual, reputational) that outweighs the marginal return from reallocating. Often the answer is "we wouldn't start it" and "stopping has no material future cost" — in which case the rational move is exit. The resistance is usually emotional or political: admitting the mistake, reallocating resource away from someone's fiefdom, or writing off the sunk amount in the accounts. The economics don't care; the decision should be made at the margin.
Business
You're seeing Sunk Costs (Economics) when a company keeps a product line alive because it has already spent heavily on R&D and marketing, even though forward-looking projections are negative. The right question is whether the next dollar spent on that product earns more than the next dollar spent elsewhere. The sunk spend is irrelevant to that comparison.
Technology
You're seeing Sunk Costs (Economics) when a team refuses to kill a legacy system or rewrite because "we've put years into it." The years are sunk. The decision is whether the next month of maintenance or the next sprint of new feature work has higher return in the legacy stack or in a replacement. Emotional attachment to the sunk investment blurs that comparison.
Investing
You're seeing Sunk Costs (Economics) when an investor holds a losing position in the hope of "getting back to even." The purchase price is sunk. The only question is whether the expected return from holding exceeds the expected return from selling and reallocating. The desire to avoid realising a loss is a sunk-cost trap, not a margin-based decision.
Markets
You're seeing Sunk Costs (Economics) when a firm continues in a market after entry costs are sunk but margins are poor. The entry cost is irrelevant to the exit decision. What matters is whether the firm can earn more by reallocating those resources (people, capital) to another market or use. Exit is often delayed because the sunk cost is misread as "we have to make it work."
Section 3
How to Use It
Decision filter
"Before adding more resource to any project or commitment, ignore what's already been spent. Ask: from today, what is the expected return on the next unit of resource here versus the best alternative? If the alternative wins, reallocate. Sunk costs are not a reason to continue; they are a reminder to decide at the margin."
As a founder
Every project and product has a sunk cost — time, money, reputation. The trap is letting that cost justify another round of investment. Run the marginal test: if we had no history, would we start this today? If the answer is no, the default should be stop or pivot, not "one more try." The same logic applies to hiring and org structure: the cost of a bad hire is sunk; the decision is whether the next dollar of payroll is better spent on that person or on a replacement or different role. Cut losses when the marginal return is higher elsewhere.
As an investor
Position size and entry price are sunk once you've bought. The decision to hold or sell should depend only on expected return from here. "Averaging down" can be rational if the expected return has improved; it's irrational if it's driven by the desire to improve the average cost. The same applies to doubling down on a portfolio company: the prior investment is sunk. The question is whether the next dollar into that company beats the next dollar into a new deal or another existing position.
As a decision-maker
When a team or executive argues for continuing a project "because we've invested so much," reframe: "Assume we've spent zero. Would we invest the next dollar here or somewhere else?" If somewhere else, the decision is reallocate or exit. Sunk costs often hide behind status and politics — killing a project feels like admitting failure. The economic discipline is to separate the marginal decision from the narrative. Document the marginal logic so that the choice is transparent and defensible.
Common misapplication: Treating all past cost as sunk. Some costs are recoverable (e.g. selling an asset, redeploying a team). Only truly irrecoverable costs are sunk. The test: can we get any of this back by stopping? If yes, that's salvage value, not sunk cost — it belongs in the marginal calculation. If no, it's sunk and should be ignored. People often assume the whole investment is sunk when part could be recovered (e.g. selling a division, licensing IP, reassigning key people). Counting salvage value correctly can make the "quit" option look better and the decision clearer.
Second misapplication: Using "sunk cost" as an excuse to quit too early. The principle says ignore the past when comparing options; it doesn't say the project is bad. Sometimes the marginal return on continuing is high. The discipline is to compute the marginal return honestly, not to assume that because money was spent, the project is doomed or that because it's sunk, we should stop. Decide on the margin; don't let the label "sunk" override the numbers.
When the model helps most: Use sunk-cost logic when you're deciding whether to add more resource to an existing project, hold or sell an investment, or continue or kill a product or division. It's especially valuable when the team or organisation is emotionally or politically committed to the past bet. The model is less relevant when the decision is truly one-off and there's no reallocation option — then there's no "elsewhere" to compare to. Even then, the habit of asking "would we start this today?" sharpens judgment.
Grove's decision to exit memory and bet Intel on microprocessors was a textbook marginal choice. Intel had invested heavily in memory; the business was being crushed by Japanese competition. Grove asked: if we were kicked out and the board brought in a new CEO, what would they do? The answer was exit memory. The sunk cost — decades of investment and identity in memory — was set aside. The marginal decision was to put resources where they could win. The result was dominance in microprocessors.
Buffett has repeatedly said that the right way to think about a holding is "would I buy it today at this price?" — not "am I up or down?" The purchase price is sunk. The decision to hold or sell depends only on current price versus expected value. He has sold positions at a loss when the marginal case no longer held. The discipline is to never let the sunk cost (what you paid) override the marginal question (what is it worth from here).
Section 6
Visual Explanation
Sunk costs: past spend is the same whether you continue or quit. The decision depends only on future (marginal) costs and benefits of each option. Reallocate when marginal return is higher elsewhere.
Section 7
Connected Models
Sunk costs (economics) sit with marginal analysis, opportunity cost, and the psychological sunk-cost fallacy. The models below either explain why sunk costs don't matter (marginal cost/benefit, thinking at the margin), capture the behavioural trap (sunk cost fallacy), or set the frame for reversible vs irreversible decisions. Use them in sequence: marginal cost/benefit and thinking at the margin give you the correct decision rule; opportunity cost gives you the comparison set; the sunk cost fallacy explains why you and others might violate the rule; reversibility reminds you that the next commitment creates new sunk cost, so choose it with the same marginal discipline.
Reinforces
Marginal [Cost](/mental-models/cost)/Benefit
Marginal cost/benefit is the rule that decisions are made at the margin: compare the incremental cost and benefit of the next unit. Sunk costs are the opposite of marginal — they're in the past and don't change with the decision. The reinforcement: the same framework that says "optimise where marginal cost equals marginal benefit" says "ignore sunk cost." Both are applications of "only the future margin matters."
Reinforces
Thinking at the Margin
Thinking at the margin means evaluating the next unit — the next dollar, hour, or hire — not the average or the total. Sunk costs are total past expenditure; they're irrelevant to the next unit. The reinforcement is that disciplined marginal thinking automatically excludes sunk cost from the decision. If you're asking "what does the next unit buy me here vs elsewhere?" you're already ignoring the sunk.
Leads-to
Opportunity Cost
Opportunity cost is the value of the best forgone alternative. The marginal decision — "should I put the next dollar here or elsewhere?" — is an opportunity-cost comparison. Sunk cost has no opportunity cost (it's already spent). The forward-looking resource does: the opportunity cost of continuing is the return from reallocating. Sunk-cost discipline makes the opportunity-cost comparison clean.
Leads-to
Reversible vs Irreversible Decisions
Section 8
One Key Quote
"The sunk cost fallacy is a tendency to continue an endeavour once an investment in money, effort, or time has been made. Economists say the sunk cost should be ignored; only marginal costs and benefits should matter. But people don't ignore them."
— Richard Thaler, Misbehaving (2015)
Thaler states the normative rule (ignore sunk cost) and the descriptive reality (people don't). The gap is where the opportunity lives: the decision-maker who can separate the marginal calculus from the emotional and political weight of the past will make better exit and reallocation choices. Use the quote as a reminder that the right rule is simple even when applying it is hard.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
The marginal question is the only question. Would you start this project, hold this position, or keep this division if you had no history? If no, the default is stop or reallocate. The sunk cost is not a reason to continue; it's a reason to be explicit about the marginal return so you don't double down on a loser.
Salvage value is not sunk. If you can recover something by stopping — sell an asset, reassign a team — that recovery is a benefit of the "quit" option. Only the truly irrecoverable part is sunk. Don't confuse "we've spent a lot" with "we can't get anything back." Sometimes the marginal decision improves once you count salvage.
Organisations institutionalise the sunk-cost trap. Projects get funding because "we've already put so much in." Careers get extended because "we've invested in this person." The fix is process: require a marginal case for every incremental commitment. "We've spent X" is not an argument; "the next dollar here earns Y versus Z elsewhere" is.
Emotional and political cost are real — but they're future cost. If "killing this project" has a future cost (reputation, morale, signalling), that cost belongs in the marginal comparison. It's not a reason to ignore the marginal logic; it's a factor in the marginal outcome. The mistake is letting the past spend drive the decision. The future cost of exiting is a valid input.
Document the marginal case. When the decision is contentious, write down the expected return from the next unit of resource here vs elsewhere. Making the marginal logic explicit reduces the chance that sunk cost and politics override it. It also creates accountability: if we're continuing, we're asserting that the marginal return here is higher. Revisit that assertion on a schedule.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A company has spent $20m on a product that has weak traction. The CEO argues: 'We need one more year and $5m to make it work. We can't walk away from the $20m.'
Scenario 2
An investor sells a position at a 30% loss and reallocates to a different stock. She says: 'The purchase price doesn't matter. What matters is expected return from here.'
Scenario 3
A team has built a feature over six months; it has low usage. The PM says: 'Let's give it three more months and a marketing push. We've already built it.' Engineering says the same engineers could ship a different feature with higher expected impact in that time.
Scenario 4
A company shuts a division and sells its assets. It uses the proceeds to pay down debt and reports a one-time loss. The CEO says: 'We're reallocating capital to higher-return businesses.'
Section 11
Summary & Further Reading
Summary: Sunk costs are past expenditures that cannot be recovered. They should not affect the current decision; only marginal (future) costs and benefits matter. The rational choice is to compare the expected return from the next unit of resource in each option and to reallocate when the return is higher elsewhere. In practice, people and organisations often honour sunk cost and continue losing endeavours. The discipline is to ask the marginal question explicitly and to separate it from the emotional weight of the past. Use the model to justify exit and reallocation and to avoid throwing good resource after bad.
Marshall's treatment of cost: distinction between fixed, variable, and sunk cost, and the principle that supply responds to marginal cost. Foundational for why past cost is irrelevant to current supply and allocation.
The FTN mental model on the psychological fallacy. Use with this page: economics says ignore sunk cost; the fallacy page explains why that's hard and how to counteract it.
Standard treatment: project evaluation uses incremental (marginal) cash flows; sunk cost is excluded. Reference any major corporate finance text (Brealey–Myers, Berk–DeMarzo) for the formal rule and worked examples.
Reversible decisions can be undone; irreversible ones cannot. Sunk costs are the result of past irreversible commitments (time and money spent). The link: for reversible decisions, the cost of trying and undoing is low. For irreversible decisions, the commitment creates a future "sunk" that you'll be tempted to honour. Good process treats reversible decisions as low-stakes experiments and irreversible ones as marginal bets — in both cases, past sunk cost is irrelevant.
Tension
Sunk Cost Fallacy
The sunk cost fallacy is the psychological tendency to continue because of past investment — the exact behaviour that economic logic says is wrong. The tension: the economics are clear (ignore sunk cost), but people and organisations keep honouring it. The model helps you name the trap and correct for it; the fallacy is what you're fighting.
Tension
Irreversibility
Irreversibility means some decisions cannot be undone. Sunk cost is the shadow of past irreversible commitments. The tension: once you've sunk cost, the marginal decision is still "what's best from here?" — but the next commitment may create new irreversibility. So you want to ignore past sunk cost while being deliberate about creating new sunk cost. Don't compound the first error by throwing good money after bad; do invest when the marginal case is strong.
