Every resource you have already spent — every dollar, every hour, every ounce of emotional energy — is gone. It cannot be recovered regardless of what you decide next. The sunk cost fallacy is the systematic error of letting those irrecoverable expenditures influence future decisions, as though continuing to invest can somehow redeem what has already been lost.
The rational principle is stark: only future costs and future benefits should inform current choices. Past expenditures are economically irrelevant. A dollar spent yesterday has exactly zero bearing on whether spending another dollar today is wise. The decision framework should be identical whether you have invested $10 million or $10 in a failing venture — what matters is the expected return on the next dollar, not the accumulated weight of previous ones.
Economists have understood this since at least Alfred Marshall's Principles of Economics in 1890. The insight is trivial to grasp intellectually and nearly impossible to execute emotionally. Ask a room of MBA students whether sunk costs should influence decisions and every hand goes up to say no. Put those same students in a simulation where they've already "invested" significant resources in a failing project, and a majority will escalate their commitment. The gap between knowing the principle and living it is the entire story of the sunk cost fallacy.
The most famous case study in sunk cost reasoning carries the bias's alternate name. The Concorde supersonic jet — a joint venture between the British and French governments — consumed approximately $2.8 billion in development costs (over $13 billion in 2024 dollars) across two decades of engineering. By the early 1970s, internal analyses at both governments had concluded the aircraft would never achieve commercial viability. Sonic boom regulations, fuel consumption rates, and narrow passenger capacity guaranteed that airlines would not order enough planes to recoup the investment. Both governments knew this. Both continued funding.
The reasoning was transparent and irrational: they had already spent too much to stop. Cancellation meant accepting the loss publicly — admitting that billions had been wasted. Continuation preserved the fiction that the investment might eventually pay off. Only 14 Concordes entered commercial service. The last flight landed at Heathrow on October 24, 2003. Economists coined the term "Concorde fallacy" to describe the pattern, and it remains the purest example of the phenomenon at sovereign scale.
The same logic drove the escalation of the Vietnam War. Robert McNamara, U.S. Secretary of Defense from 1961 to 1968, later acknowledged in his 1995 memoir In Retrospect that by 1965, senior officials had serious doubts about the war's winnability. The United States had already deployed 184,000 troops and spent billions. The sunk cost argument — "we cannot let these sacrifices be in vain" — became the political and psychological justification for escalation. Troop levels peaked at 549,500 in 1969. Total U.S. expenditure exceeded $140 billion (roughly $1 trillion in 2024 dollars). The cost of the war wasn't incurred because the strategy was working. It was incurred because the prior costs made withdrawal feel like betrayal.
Hal Arkes and Catherine Blumer formalised the phenomenon in their 1985 paper "The Psychology of Sunk Cost," published in Organizational Behavior and Human Decision Processes. Through a series of ten experiments, they demonstrated that people who had invested more in an activity were more likely to continue it — regardless of whether the activity was producing returns. In one landmark experiment, they sold Ohio University theatre season tickets at three randomly assigned prices — full price ($15), a small discount, and a large discount. Over the following months, full-price buyers attended significantly more performances than discounted buyers. The shows were identical. The enjoyment was identical. The only difference was how much each person had paid — and that payment, already made, should have been irrelevant to the decision of whether to attend on any given evening. A $15 ticket to a bad play keeps more people in their seats than a free ticket to the same bad play. The investment creates a psychological debt that only continued participation can discharge.
The corporate version is more expensive. Kodak invested over $5 billion in digital photography research and development between 1975 and 2003 — more than any competitor. Steve Sasson, a Kodak engineer, built the first digital camera prototype in 1975. The company held thousands of digital imaging patents. Kodak had every technical advantage needed to lead the digital transition. The accumulated investment in film infrastructure — factories, supply chains, distribution partnerships, brand positioning — became an anchor. Each dollar sunk into the film ecosystem reframed digital as a threat rather than an opportunity. Kodak filed for bankruptcy in January 2012, killed not by a lack of technology but by an inability to abandon the past.
The fallacy operates at every level of human decision-making. A PhD student who has spent five years on a dissertation topic continues despite losing interest because abandoning it means "wasting" those years. A CEO who championed an acquisition continues defending it after integration fails because reversing course means admitting the original decision was wrong. A government that has spent $400 billion on a fighter jet programme refuses to cancel it despite decade-long delays and ballooning costs because the prior spending creates political pressure to see it through. In each case, the past consumes the future. The sunk cost doesn't change what the right decision is. It changes how painful the right decision feels.
The asymmetry between knowing and doing is the entire subject of the sunk cost literature. Every CEO can articulate the principle in a board meeting. Almost none can execute it when the $200 million project they championed is failing and the board is watching. The fallacy is not an information problem. It's an execution problem — and the execution gap widens in direct proportion to the size of the prior investment, the public visibility of the commitment, and the personal identification of the decision-maker with the original choice.
Section 2
How to See It
The sunk cost fallacy disguises itself as perseverance, loyalty, consistency, and fiscal responsibility. The language gives it away: "We've come too far to stop now." "We can't let that investment go to waste." "We owe it to the people who worked on this." Each statement treats past expenditure as a claim on future resources — as though the money, time, or effort already spent has created a moral obligation to continue.
Business
You're seeing the Sunk Cost Fallacy when a company continues funding a product that has missed every milestone for three consecutive years because "we've already invested $200 million." The $200 million is gone regardless of the next decision. The question isn't whether to waste it — that decision was made incrementally over three years. The question is whether the next $50 million has a positive expected return. Xerox PARC spent the 1970s inventing the graphical user interface, Ethernet, and the laser printer, then watched Apple and Microsoft commercialise them — partly because Xerox's sunk investment in copier infrastructure made the leap to personal computing feel like abandoning a fortress rather than building a new one.
Investing
You're seeing the Sunk Cost Fallacy when an investor refuses to sell a stock that has declined 60% because "I'll sell when it gets back to my entry price." The purchase price is a historical accident. The stock's future trajectory depends on current fundamentals, competitive position, and valuation — none of which are influenced by what any individual investor paid. Terrance Odean's 1998 study of 10,000 brokerage accounts found that the stocks investors sold outperformed the stocks they held by 3.4 percentage points annually — because investors systematically held losers and sold winners, anchoring to entry prices the market had already moved past.
Policy
You're seeing the Sunk Cost Fallacy when a government continues funding a defence programme because cancellation would "waste" prior spending. The F-35 Joint Strike Fighter programme, with a projected lifetime cost exceeding $1.7 trillion, has faced persistent calls for scaling back. The counterargument invariably centres on the hundreds of billions already spent — as though prior expenditure creates an obligation to continue regardless of whether the remaining spend is justified on its own terms. The rational question is whether the next dollar of spending generates sufficient military capability relative to alternatives, not whether the previous dollars would be "wasted" by stopping. The previous dollars are wasted either way — the question is whether you waste the next ones too.
Personal life
You're seeing the Sunk Cost Fallacy when someone stays in a career they dislike because they spent four years and $200,000 on a degree. The degree cost is sunk — it cannot be recovered regardless of career choice. A 2018 University of Chicago study by Steven Levitt found that people who made major life changes — including career switches — reported significantly higher happiness six months later than those who stayed. The years spent on a degree provide knowledge and credentials that transfer across fields. They do not create an obligation to spend forty more years in a profession that doesn't fit.
Section 3
How to Use It
Understanding the sunk cost fallacy has two distinct applications. The defensive use: recognising when past investments are distorting your own decisions and building systems to override the distortion. The offensive use: understanding when competitors, counterparties, or institutions are trapped by their sunk costs — and positioning yourself to exploit the resulting inefficiencies.
George Soros's 1992 bet against the British pound was partly a bet on sunk cost reasoning: the Bank of England would defend the ERM peg longer than rational analysis warranted because abandoning it meant accepting a visible, public loss. Soros positioned $10 billion against the pound, knowing the Bank's sunk cost psychology was creating the conditions for its own capitulation. The trade netted $1 billion in a single day. Institutional sunk cost traps are among the most reliable patterns in markets — and among the most profitable to exploit.
Decision filter
"Before committing additional resources to any endeavour, ask: if I were starting fresh today — with no prior investment — would I choose to enter this project at its current state? If the answer is no, the only reason you're continuing is sunk costs."
As a founder
The most dangerous sunk cost for a founder isn't money — it's identity. You named the company. You pitched the vision at fifty investor meetings. You recruited a team around a specific thesis. Pivoting feels like admitting that the version of you who made those commitments was wrong. The emotional cost of that admission keeps founders pouring resources into failing products long after the data has rendered its verdict.
Build a "kill criteria" list at the start of every major initiative. Before writing a line of code, define the specific, measurable conditions under which you would abandon the project — revenue below a threshold by a date, user retention below a percentage after launch, customer acquisition cost above a ceiling. The criteria must be set in advance, before sunk costs accumulate and before the emotional attachment crystallises.
Eric Ries codified this discipline in The Lean Startup (2011): the "pivot or persevere" meeting, conducted at regular intervals, forces a binary decision against predetermined benchmarks. The meeting works because it converts an emotional question ("should I give up on my dream?") into a mechanical one ("have we hit the numbers or not?"). Stewart Butterfield did this at Tiny Speck in 2012 — the gaming company was failing, but the internal communication tool the team had built was exceptional. Killing the game meant losing two years of work and the gaming identity. The pivot produced Slack, which reached a $27 billion valuation by 2020. The kill criteria told Butterfield the game was dead. The sunk cost in the game was irrelevant to the value of what was being built alongside it.
As an investor
Treat every position as though you are buying it today at the current price. If you wouldn't buy the stock at its current valuation with your current information, you should sell it — regardless of what you paid. The mental reframe eliminates the sunk cost anchor by replacing "should I hold or sell?" with "should I buy this today?" Peter Lynch formalised this in One Up on Wall Street (1989): "Know what you own, and know why you own it." The "why" must reference the future, not the past.
Position sizing discipline reinforces the point. George Soros, who generated a 30% annualised return at Quantum Fund over three decades, was famous for cutting losses with mechanical speed. His protégé Stanley Druckenmiller described the discipline: "Soros has taught me that when you have tremendous conviction on a trade, you have to go for the jugular. It takes courage to be a pig. But it also takes courage to take a loss."
The willingness to abandon a position — regardless of its cost basis — is the single most reliable differentiator between professional and amateur capital allocators. The market doesn't know what you paid. Your cost basis is irrelevant to the stock's future. The only question that matters is whether the position, evaluated at today's price, has a better expected return than the alternatives available today.
As a decision-maker
In organisations, sunk costs create political gravity. The executive who championed a project becomes its defender — not because the project is working, but because its failure reflects on the decision to fund it.
The structural fix: separate the decision to continue a project from the person who initiated it. Amazon's six-page memo format and Jeff Bezos's insistence on written arguments (rather than PowerPoint advocacy) make decisions about evidence rather than personality. Intel under Andy Grove went further: strategic reviews explicitly asked whether the company would enter a business today if it weren't already in it. The question strips away accumulated investment and forces evaluation on future merit alone. Grove's "new CEO" thought experiment — imagining a replacement executive unburdened by history — is the purest organisational tool for defeating sunk cost reasoning.
The same principle applies to programme management. When Google killed Google Reader in 2013 despite millions of loyal users, it framed the decision internally as a loss that was already happening — declining engagement, rising maintenance costs, strategic misalignment — rather than as a choice being imposed.
Whether you're restructuring a team, discontinuing a product, or changing a pricing model, the question isn't just "what's the right decision?" It's "how do I structure the decision process so that sunk costs can't distort the outcome?" The best decision architectures don't rely on overcoming the bias in the moment. They design the moment out of existence.
Common misapplication: Not all persistence is sunk cost fallacy. SpaceX's first three Falcon 1 rockets exploded between 2006 and 2008, consuming nearly all of Elon Musk's remaining capital. A surface reading would call the decision to fund a fourth launch sunk cost reasoning. It wasn't. Musk's analysis indicated that each failure yielded specific engineering data improving the next iteration — the technical problems were solvable and the underlying market thesis was intact. The decision to continue was based on updated future projections, not past expenditure.
The distinction is critical: continuing because the future looks promising despite past costs is rational. Continuing because past costs make quitting feel wasteful is the fallacy. The diagnostic test is always the same — would you start this project today, knowing what you now know, if you had no prior investment? If yes, continue. If no, the only thing keeping you in is the sunk cost. The difficulty lies in being honest about the answer, because the sunk cost fallacy's most effective trick is making the future look promising specifically because the past investment was large.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
The sunk cost fallacy reveals itself most clearly at the extremes — catastrophic failures where leaders couldn't abandon the past, and decisive pivots where leaders refused to let prior investment dictate future strategy. The cases below span semiconductors, consumer electronics, streaming, software, and e-commerce. In each, the central question is identical: did the leader evaluate the future on its own terms, or did the weight of the past distort the calculation?
The pattern is remarkably consistent. The leaders who built enduring companies share an unusual ability to treat their own prior investments as irrelevant when the evidence demands a change. They feel the pain — Grove described Intel's memory exit as a "valley of death," and Hastings watched Netflix's stock collapse 77% during the streaming transition — but they act on the analysis, not the attachment. Five cases illustrate both the cost of surrendering to sunk costs and the reward of ignoring them.
Intel's exit from memory chips is the definitive counter-example to the sunk cost fallacy — the case where a leader recognised the trap and walked through it.
By 1985, Intel had spent seventeen years building its memory business. The 1103 DRAM, introduced in 1970, was the company's first commercial hit. Engineering teams, manufacturing facilities, customer relationships, and corporate identity — all built around memory. Japanese manufacturers had destroyed Intel's competitive position through aggressive pricing and superior yields. The company was losing $173 million annually in the segment.
Grove's breakthrough was a question that neutralised the sunk cost anchor. He asked Gordon Moore: "If we got kicked out and the board brought in a new CEO, what would he do?" The hypothetical CEO had no sunk costs. No identity tied to memory. No seventeen years of history creating a psychological claim on the future. That imaginary executive would exit memory without hesitation.
Intel laid off 7,200 employees, shuttered fabrication plants, and redirected every resource to microprocessors. The transition nearly destroyed morale — engineers who had spent careers perfecting memory chips felt their life's work was being discarded. Grove later described the period as the most painful of his career.
Revenue grew from $1.9 billion in 1986 to $25 billion by 1999. The 486 and Pentium processors made Intel synonymous with personal computing. The company that sunk cost reasoning nearly killed became the most valuable semiconductor firm on earth — because one leader found a cognitive technique to strip away the past and evaluate the future on its own terms.
When Jobs returned to Apple in 1997, the company was ninety days from bankruptcy and selling over forty different products — desktops, laptops, servers, printers, the Newton PDA, the Pippin gaming console. Each product had a development team, a manufacturing line, a marketing budget, and an internal constituency.
Jobs eliminated 70% of the product line within his first year. The Newton — hundreds of millions in development investment and Apple's bet on handheld computing — was cancelled in February 1998. The Pippin was killed. The server line was gutted. Dozens of software projects were terminated. When engineers argued that years of work would be "wasted," Jobs responded with a question that sliced through sunk cost reasoning: "Which ones would you choose to start today?"
The result was a four-quadrant grid — consumer and professional, desktop and portable — that focused Apple's remaining resources on four products. The iMac G3, launched in August 1998, sold 800,000 units in its first five months and returned Apple to profitability. The discipline of killing products with massive sunk costs freed capital, engineering talent, and management attention for the iPod (2001), the iPhone (2007), the iPad (2010). Treating sunk costs as irrelevant was the precondition for everything that followed.
The Fire Phone was Amazon's most ambitious and most public hardware failure. Announced in June 2014 after four years of secretive development, the device featured a 3D "Dynamic Perspective" display powered by four front-facing cameras — a technical achievement that solved no meaningful customer problem. Amazon invested an estimated $170 million in development. Hundreds of engineers were consumed for years.
The phone flopped immediately. AT&T dropped the price from $199 to $0.99 within two months. Amazon took a $170 million write-down in Q3 2014. The conventional sunk cost response would have been to iterate — release a Fire Phone 2, improve the software, protect the investment.
Bezos killed the programme entirely and redirected the hardware team to a new project that became the Amazon Echo. The Echo, launched in November 2014, used voice recognition to create a genuinely useful interface. By 2020, Amazon had sold over 100 million Echo devices and established Alexa as the dominant voice assistant platform. The $170 million spent on Fire Phone was treated as tuition, not as a debt requiring repayment. Bezos framed it explicitly: "If you're going to take bold bets, they're going to be experiments. And if they're experiments, you don't know ahead of time if they're going to work." Past bets are information, not obligations.
In 2007, Netflix was a profitable, growing business with 7.5 million DVD-by-mail subscribers and a dominant position in physical media rental. The company had spent eight years building the most sophisticated DVD logistics network in the world — 58 distribution centres, proprietary sorting technology, and supplier relationships optimised for next-day delivery. The sunk cost in that infrastructure was enormous, both financially and organisationally.
Hastings bet the company on streaming anyway. The DVD network — the thing that made Netflix Netflix — was treated as a depreciating asset rather than a permanent identity. The reframe was explicit: either Netflix would cannibalise its own DVD business, or a competitor would do it for them. Blockbuster, Amazon, Apple, and Hulu were all circling the streaming opportunity. Clinging to DVDs to "protect the investment" would have meant entering streaming late, without the subscriber base or content relationships to compete.
The 2011 Qwikster debacle — when Netflix tried to separate DVD and streaming into distinct brands — showed the messiness of the transition. The stock dropped 77% in four months. 800,000 subscribers cancelled. The sunk cost pressure to reverse course and recommit to DVDs was overwhelming. Hastings held the strategic line. By 2023, Netflix had 260 million streaming subscribers and a market capitalisation exceeding $250 billion. The DVD business was shut down in September 2023, having been irrelevant for years. Blockbuster, which clung to its retail model and rejected a chance to acquire Netflix for $50 million in 2000, had filed for bankruptcy in 2010. The difference between the two companies was not technology or talent. It was the willingness to treat past investment as irrelevant to the next decision.
When Nadella became CEO in February 2014, he inherited a $7.6 billion problem. Steve Ballmer's 2013 acquisition of Nokia's phone business represented Microsoft's largest bet on mobile hardware — the culmination of years of Windows Phone development, app store subsidies, and the conviction that Microsoft needed its own device ecosystem. Sunk costs extended far beyond the acquisition price: billions more in development, marketing, and the organisational disruption of integrating 25,000 Nokia employees.
Within eighteen months, Nadella wrote off the entire Nokia acquisition and eliminated 7,800 jobs from the phone division. The write-down was the largest in Microsoft's history. The sunk cost argument against it was formidable: years of platform development, billions in acquisition spending, enormous political capital invested in the mobile strategy.
Nadella's reframe was structural, not incremental. He redirected Microsoft toward cloud computing and enterprise services — Azure, Office 365, LinkedIn — where the company had genuine competitive advantages that no amount of phone hardware investment could replicate. The capital, talent, and management attention freed by killing the phone business funded the cloud pivot. Azure grew from a marginal product in 2014 to a $60 billion annual revenue business by 2024. Microsoft's market capitalisation rose from $300 billion to over $3 trillion.
The $7.6 billion write-down looks, in retrospect, like the most profitable loss Microsoft ever took — because it purchased the organisational clarity needed to pursue a strategy worth trillions. The lesson is precise: the cost of the write-down was $7.6 billion. The cost of not taking it — of continuing to pour resources into Windows Phone to "protect the investment" — would have been measured in trillions of foregone enterprise value.
Section 6
Visual Explanation
The sunk cost fallacy distorts decision-making by introducing a variable — past investment — that has no bearing on optimal future choices. The rational framework evaluates only the expected future value of continuing versus stopping. The biased framework adds past expenditure to the ledger, creating a phantom obligation that tips the scale toward continuation even when the expected return is negative. The diagram below illustrates the two frameworks side by side: the rational decision-maker who asks "would I start this today?" and the biased decision-maker who says "we've invested too much to quit." Same project. Same future prospects. Radically different conclusions — because one framework includes a variable that the other correctly excludes.
Sunk Cost Fallacy — The rational decision evaluates only future costs and benefits. The biased decision adds irrecoverable past costs, distorting the choice toward continuation.
Section 7
Connected Models
The sunk cost fallacy does not operate in isolation. It is powered by deeper psychological biases, reinforced by cognitive patterns that protect the ego from uncomfortable truths, and held in tension by frameworks specifically designed to override exactly this kind of distortion. Understanding these connections reveals how a single decision error — weighting past costs in a forward-looking decision — cascades through an entire system of reasoning, recruiting allies from loss aversion, escalation dynamics, and confirmation bias until the original error becomes self-sustaining.
Reinforces
Loss Aversion
Loss aversion is the psychological engine behind the sunk cost fallacy. Kahneman and Tversky's finding — that losses are felt roughly twice as intensely as equivalent gains — explains why abandoning a failing investment feels so disproportionately painful. The money is already gone, but the brain doesn't process "spent" and "lost" as equivalent categories. Spending is a completed transaction. Recognising the spend as a loss is a fresh wound.
Continuing to invest postpones the moment of recognition — keeping the loss in a psychological limbo where it feels provisional rather than final. An unrealised loss feels revocable, like a stock that hasn't been sold yet. Pulling the plug makes it permanent — and permanence is what the brain resists.
Without loss aversion, sunk cost reasoning would lack its emotional fuel. The Concorde's British and French backers weren't miscalculating the economics. They were avoiding the pain of publicly recognising a loss that had already occurred. The £2.8 billion was gone whether they flew the planes or not. What they were really protecting was the illusion that it wasn't.
Reinforces
Escalation of Commitment
Barry Staw's 1976 research demonstrated that people personally responsible for an initial investment are significantly more likely to commit additional resources when the investment is failing — compared to those who inherited the same failing position. The effect is not small: in Staw's experiments, personal responsibility increased escalation by roughly 40%. The implication is profound — the person best positioned to evaluate a project (the one who understands it most deeply) is the person most psychologically incapable of abandoning it.
The sunk cost creates the trap; escalation is the spiral that deepens it. The Concorde programme, the Vietnam War, and countless corporate acquisitions follow the same trajectory: initial investment, negative feedback, doubling down, further negative feedback, further doubling down. Each round of additional spending creates new sunk costs that make the next exit point harder to take. The interaction produces a ratchet effect — the door to retreat closes a little more with each incremental commitment, until the accumulated sunk costs make abandonment feel psychologically impossible even when it is economically obvious.
Section 8
One Key Quote
"Should you find yourself in a chronically leaking boat, energy devoted to changing vessels is likely to be more productive than energy devoted to patching leaks."
— Warren Buffett, 1985 Berkshire Hathaway Shareholder Letter
Section 9
Analyst's Take
Faster Than Normal — Editorial View
The sunk cost fallacy is the most expensive cognitive bias in business — not because each individual instance is catastrophic, but because it operates continuously, at every level of every organisation, on every decision that involves prior investment. It is the background radiation of bad decision-making. And unlike many cognitive biases, awareness provides almost no protection. Knowing about the sunk cost fallacy doesn't make the $200 million you've already spent feel any less like a claim on the future.
Here's what makes it uniquely dangerous: the fallacy mimics responsible management. "We need to protect our investment" sounds like fiduciary duty. "We owe it to our team to see this through" sounds like leadership. "We've come too far to turn back" sounds like determination. Every one of these statements is sunk cost reasoning dressed in professional language. The bias doesn't announce itself as irrational. It arrives wearing the vocabulary of accountability and fiscal prudence.
What makes this particularly insidious is that the people most susceptible are often the most conscientious. The executive who cares deeply about stewardship of capital is the one most likely to feel that abandoning a $200 million investment is "irresponsible." The founder who feels genuine obligation to early employees is the one most likely to keep funding a failing product because "we owe it to the team." The bias exploits virtue. It turns conscientiousness into a mechanism for misallocation.
The pattern I observe most frequently: sunk cost reasoning accelerates precisely when it should decelerate. The bigger the initial investment, the more urgently leaders feel the need to justify it — and the larger the subsequent commitments become. Barry Staw's escalation research predicts this precisely. A $10 million investment that's failing generates a $5 million follow-on. A $100 million investment that's failing generates a $50 million follow-on. The ratio stays roughly constant; only the scale changes.
The result is that the worst decisions attract the most resources — a systematic misallocation engine running silently inside every organisation that lacks explicit countermeasures. The corporate graveyard is filled with companies that didn't die from a single catastrophic error but from the slow, compounding effect of continuing to invest in things that weren't working because the prior investment made stopping feel impossible.
The leaders who handle sunk costs best share a specific trait: they institutionalise the exit decision. They don't rely on individual courage to walk away from failing investments. They build processes that make abandonment a scheduled, depersonalised event. Grove's strategic review at Intel, where every business unit was evaluated as though the company were deciding whether to enter it today. Bezos's framing of failed experiments as "tuition" — a linguistic device that converts sunk costs from losses into learning. Nadella's willingness to take a $7.6 billion write-down on Nokia within eighteen months of becoming CEO — a signal to the entire organisation that killing the past was not failure but strategy.
Section 10
Test Yourself
The scenarios below test your ability to distinguish sunk cost reasoning from rational persistence, strategic patience, and legitimate reinvestment. The critical question in each case: is the decision being shaped by past investment that cannot be recovered, or by a forward-looking evaluation of expected returns? The distinction matters enormously — misidentifying rational persistence as sunk cost fallacy is just as costly as failing to recognise the fallacy when it's operating.
Is this mental model at work here?
Scenario 1
A biotech company has spent $800 million over eight years developing a drug that just failed its Phase III clinical trial. The CEO announces an additional $200 million for a modified Phase III trial, arguing that 'we cannot walk away from $800 million in research.'
Scenario 2
A venture capitalist with a $2 million investment in a struggling startup invests an additional $500,000 after the company pivots to a new market showing early traction — three enterprise customers paying $50,000/year each, plus a pipeline of twelve qualified leads.
Scenario 3
A couple has spent $40,000 on non-refundable wedding deposits — venue, catering, photographer — when they realise, three months before the date, that they are no longer compatible. They proceed with the wedding because 'we've already paid for everything.'
Section 11
Top Resources
The sunk cost literature sits at the intersection of cognitive psychology, behavioral economics, and organisational theory. Start with Arkes and Blumer for the foundational experimental evidence, then move to Kahneman for the theoretical framework and Thaler for the economic applications. Grove's memoir provides the best first-person account of overcoming the bias at organisational scale. The primary sources are unusually accessible for academic research — both Kahneman and Thaler write for general audiences with rare clarity.
The foundational experimental paper on the sunk cost effect. Arkes and Blumer's Ohio University theatre ticket experiment — full-price buyers attending more shows than discounted buyers despite identical performances — remains the clearest demonstration that past expenditure distorts future behaviour. Ten experiments across money, time, and effort investments established the phenomenon with rigour that subsequent research has replicated and extended. Dense but accessible, and available in most university databases.
Kahneman's treatment of loss aversion and prospect theory provides the theoretical foundation for understanding why sunk costs exert such pull. Chapters 25–29 explain the value function — the S-shaped curve showing that losses hurt roughly twice as much as equivalent gains — which is the engine behind the sunk cost effect. The chapter on the planning fallacy is directly relevant: it explains why forward-looking estimates are systematically overoptimistic, making the sunk cost anchor even more distortive when people believe the next round of investment will be the one that turns things around.
Thaler's intellectual memoir traces how sunk cost reasoning, mental accounting, and the endowment effect moved from laboratory curiosities to the foundations of a new economic discipline. The chapters on mental accounting are particularly illuminating — Thaler explains how people create mental "accounts" for investments and treat closing an account at a loss as psychologically unbearable, even when the economic reality hasn't changed. The concept of "closing an account at a loss" is the precise mechanism by which sunk costs distort behaviour. Written with humour and honesty about the decades-long battle to get mainstream economics to acknowledge that people aren't rational.
The best first-person narrative of a leader overcoming sunk cost reasoning at organisational scale. Grove's "strategic inflection point" framework and the "new CEO" thought experiment provide practical tools for stripping away accumulated investment when evaluating strategic direction. His honesty about the emotional difficulty of the decision — laying off 7,200 employees, abandoning the company's founding identity, navigating what he called a "valley of death" — makes the book uniquely valuable. This is not a sanitised case study but a record of how painful it actually is to walk away from sunk costs when your identity is invested in what you're abandoning.
The longest-running record of a practitioner grappling with sunk costs in real capital allocation. The 1985 letter's "leaking boat" metaphor is the most quoted articulation of the principle. The discussion of the Dexter Shoe mistake (1993 letter, revisited in subsequent years with escalating calculations of opportunity cost) shows Buffett confronting his own sunk cost reasoning with rare candour. The letters on Berkshire's original textile operations — which Buffett held for twenty years despite knowing the business had no future, partly out of loyalty to employees and partly because closing it meant accepting a loss — are the most honest account of sunk cost reasoning by a master allocator. Buffett acknowledges that even he succumbed. Free, online, and indispensable.
Tension
[Inversion](/mental-models/inversion)
Charlie Munger's inversion technique — "tell me where I'm going to die, so I'll never go there" — is the most direct antidote to sunk cost reasoning. Inversion forces you to evaluate the future by asking what would guarantee failure, rather than defending the past by asking how to salvage an existing investment.
Andy Grove's "new CEO" thought experiment is inversion applied specifically to sunk costs: by imagining a decision-maker with no history, no emotional attachment, and no accumulated investment, the technique strips away exactly the variables the sunk cost fallacy introduces. The hypothetical new CEO doesn't know about the $200 million already spent. Doesn't know about the three years of engineering effort. Doesn't know that the board publicly committed to the strategy. That executive sees only the future — and decides accordingly. The two frameworks cannot coexist in the same decision. Sunk cost reasoning anchors you to the past. Inversion forces evaluation purely on future terms. The discipline is choosing inversion.
Tension
Regret Minimization Framework
Jeff Bezos's regret minimisation framework resets the temporal reference point from "what have I already invested?" to "what will I wish I had done at age 80?" The reframe directly counteracts sunk cost reasoning by replacing backward-looking anchors with forward-looking ones.
A founder trapped by sunk costs asks: "What about all the time and money I've spent?" The framework replies: "Will you regret spending more time on something that isn't working, or will you regret not redirecting those resources to something that could?" The framework doesn't eliminate the pain of abandonment — it redirects it, making the cost of continuation (wasted future) feel heavier than the cost of stopping (wasted past). The two models fight over the same psychological territory: which loss looms larger? Sunk cost reasoning says the past investment is the bigger loss. Regret minimisation says the unlived future is. The framework works precisely because it uses loss aversion against the sunk cost fallacy — redirecting the bias rather than trying to eliminate it.
Leads-to
Opportunity Cost
Every dollar, hour, and unit of attention committed to a failing project because of sunk costs is a resource unavailable for the next opportunity. The sunk cost fallacy's most devastating consequence is not the continued waste on the failing venture — it's the invisible loss of the ventures that never happened because those resources were consumed.
Kodak's billions in film infrastructure didn't just prevent a digital pivot; they prevented investment in adjacent markets where imaging expertise could have created entirely new value. Microsoft's years defending Windows Phone meant years not spent building Azure's cloud lead earlier. Nokia's refusal to abandon Symbian meant missing the smartphone revolution it was positioned to lead. In each case, the visible cost was the money wasted on the failing strategy. The invisible cost — the opportunity foregone — was orders of magnitude larger. The sunk cost fallacy leads directly to opportunity cost blindness: fixation on what you've already given while missing what you're giving up with every day of continued commitment.
Leads-to
Confirmation Bias
Once sunk costs lock a decision-maker into a course of action, confirmation bias arrives to reinforce the commitment. The executive who championed a failing acquisition selectively seeks positive customer feedback, optimistic revenue projections, and favourable analyst opinions — while discounting or ignoring evidence of failure. You attend to data that supports your current strategy and dismiss data that contradicts it. You surround yourself with advisers who agree with you and avoid those who challenge you.
The sequence is predictable and self-reinforcing: sunk costs create the emotional motivation to continue, and confirmation bias provides the cognitive mechanism to maintain the illusion that no loss has occurred. A 2005 study by Andrea Frazzini found that fund managers holding losing positions consume significantly more bullish research on those positions than bearish research. The combination forms a feedback loop that can sustain a failing commitment for years, insulating the decision-maker from the very information that would trigger the rational response: stop. The sunk cost opens the door. Confirmation bias locks it shut.
The common thread: each separated the person who made the initial decision from the decision to continue. When the same person is both investor and evaluator, sunk cost reasoning is almost impossible to defeat. The emotional investment compounds the financial investment. The reputational cost of admitting error compounds both. The only reliable fix is structural — make the continuation decision someone else's job, evaluated against criteria set before the emotional attachment formed.
The most dangerous variant of this dynamic occurs in organisations where "commitment" is culturally rewarded and "quitting" is culturally punished. In such environments, the leader who kills a failing project is seen as disloyal, while the leader who pours another $50 million into it is seen as dedicated. The culture itself becomes an amplifier of sunk cost reasoning. The antidote is to build cultures where the highest-status move is not defending a past decision but making the best possible decision today — regardless of what was decided yesterday.
For investors, the prescription is brutally simple: automate your sell discipline. If you wouldn't buy the stock at today's price, sell it. Your entry price is irrelevant to the stock's future. The disposition effect costs retail investors an estimated 4-5% annually in foregone returns, per Odean and Frazzini's research. Over a thirty-year career, that compounds to roughly 70% of terminal wealth. The difference between disciplined and undisciplined exits, accumulated across decades, is the difference between financial security and financial mediocrity. Jim Simons at Renaissance Technologies engineered this at a systematic level — removing human judgment from trade execution entirely. Algorithms don't feel the pain of realising a loss, which means they don't hold losers or sell winners prematurely. The Medallion Fund's 66% annualised return before fees over three decades is, among other things, a testament to what happens when sunk cost reasoning is designed out of the investment process.
One nuance that deserves serious attention: not all past investment is psychologically irrelevant, even when it's economically irrelevant. The years a founder spent building domain expertise while working on a failed product are genuinely valuable — they represent learning, relationships, and pattern recognition that transfer to the next venture. The sunk cost fallacy applies to the financial and time commitment to the specific endeavour. It does not apply to the human capital accumulated along the way.
Bezos was right to frame Fire Phone as tuition. The engineers who built the 3D camera system learned hardware skills that powered the Echo. Stewart Butterfield's failed game at Tiny Speck produced Slack — the $27 billion company that emerged from the internal communication tool his gaming team had built. The game was a sunk cost. The tool was a byproduct. The distinction matters: abandon the project, but carry the learning. The money is gone. The knowledge compounds.
The question I ask every leadership team I study: "What are you funding today that you would not choose to start if you were beginning fresh?" The honest answer, pursued with genuine rigour, is worth more than most strategy consulting engagements. It is also the question that boards, executives, and founders are least willing to ask — because the answer might require them to admit that the thing they've spent years building, defending, and championing has run its course. The sunk cost fallacy's deepest trick isn't making you spend money on something that isn't working. It's making you unable to see that it isn't working — because seeing it would mean accepting that the past was a loss, and loss is the one thing the human brain is wired to avoid at almost any cost.
Scenario 4
A software company has spent two years building a proprietary database engine. A superior open-source alternative emerges that is faster and free. The CTO argues for switching. The CEO overrules her: 'Our team put two years of their lives into this engine — we owe it to them to ship it.'
