Sunk Cost Fallacy

Why Past Investment Holds Future Decisions Hostage

Also known as: Concorde Fallacy · Escalation of Commitment · Throwing Good Money After Bad

The sunk cost fallacy is the behavioral-economics finding that prior investment produces continued commitment to losing trajectories — audiences and operators stay with bad investments specifically because the prior cost is psychologically irreversible, even though rational analysis says only forward-looking expected value should govern. The framework was crystallized by Hal Arkes and Catherine Blumer's 1985 Organizational Behavior and Human Decision Processes paper "The Psychology of Sunk Cost," extended by Richard Thaler's 1980 mental-accounting work, and embedded in practitioner literature through Kahneman's 2011 Thinking, Fast and Slow and Dan Ariely's Predictably Irrational. The strategic question for brand work is whether subscription retention, abandoned-cart recovery, and feature-bloat decisions should be designed against documented sunk-cost dynamics rather than rational forward-looking evaluation.

The intellectual lineage runs through behavioral economics. Richard Thaler's 1980 Journal of Economic Behavior & Organization paper "Toward a Positive Theory of Consumer Choice" introduced sunk-cost effects into the mental-accounting framework that earned him the 2017 Nobel Prize in Economics. Hal Arkes's Ohio University and Ohio State work since 1975 — including the 1985 paper with Blumer — established the empirical base. Barry Staw's 1976 Organizational Behavior and Human Performance "Knee-Deep in the Big Muddy" introduced escalation of commitment as the organizational variant. Daniel Kahneman's 1979 prospect theory work with Tversky gave the broader loss-aversion frame that sunk cost operates inside. Dan Ariely's 2008 Predictably Irrational carried the lesson into popular practitioner orthodoxy.

How it works

Sunk cost operates through three distinct mechanisms that distort forward-looking decisions.

The first is loss aversion against an unrealized loss. Abandoning a prior investment forces the decision-maker to recognize the loss; continuing keeps it notional. Loss-averse audiences preferentially keep losses notional, even when continuing increases the eventual realized loss. Arkes and Blumer's 1985 theater-ticket experiments are the canonical demonstration: subjects who paid for a ticket were more likely to attend a play in bad weather than subjects who received the same ticket free, despite the forward-looking expected utility being identical. Prospect Theory (entry 95) describes the underlying loss-aversion frame.

The second is commitment-and-consistency pressure. Continuing a prior commitment maintains a self-narrative of consistency; abandoning it requires admitting the prior decision was wrong. Cialdini's commitment-and-consistency principle compounds the loss-aversion effect — the cognitive cost of revising the self-narrative is itself a sunk-cost-like force. Cialdini Influence Principles (entry 99) describes the broader frame.

The third is ego investment. For organizational sunk cost — Staw's "knee-deep in the big muddy" finding — the decision-maker's identity is bound up in the prior decision. Acknowledging the project is failing is functionally equivalent to acknowledging "I was wrong," which often runs against career incentives, reputation maintenance, and self-image. The result is escalation of commitment — doubling down rather than cutting losses. The Concorde program is the canonical case at sustained scale.

There's a fourth feature operating in 2026: AI-mediated personalized retention. Subscription-retention systems now algorithmically surface a user's accumulated history at cancellation moments — "you've watched 1,247 hours, are you sure?" — engineering sunk-cost salience into the cancel flow. The line between welfare-aligned reminder and manipulative retention is contested, and regulators in several jurisdictions have begun pushing back on dark-pattern cancellation flows.

Variants

Subscription Retention

The most-discussed variant: streaming, software, and gym-membership categories all operate inside sunk-cost retention dynamics. Adobe Creative Cloud's accumulated PSD/AI/PRPROJ files lock users into a forward-looking switching cost that compounds the prior subscription investment. Netflix's accumulated viewing history and recommendation tuning function similarly. Gym memberships famously trade on January-signup intent that becomes year-long passive retention.

Abandoned-Cart Recovery

E-commerce abandoned-cart email sequences operate inside browsing-and-selection sunk cost — the user already invested decision time in selecting items, and the recovery message surfaces that investment as a retention prompt. Goal Gradient Effect (entry 105) describes the parallel progress-toward-completion dynamic.

Loyalty Tier Lock-In

Sephora Beauty Insider, Starbucks Rewards, and airline-mileage programs build sunk-cost into their tier architecture. Accumulated points, status levels, and member-only access function as switching costs that increase with tenure.

Concorde-Fallacy Strategic Decisions

Strategic-decision sunk cost — the Concorde program (1962-2003), Quibi (April-October 2020), Meta's Reality Labs investment (2014 onward) — produces escalation of commitment at organizational scale. The pattern is consistent: forward-looking analysis suggests cutting losses; accumulated investment, organizational ego, and sunk capital push toward continuing.

Free-Trial-to-Paid Conversion

Free-trial product onboarding installs accumulated investment before the paid decision: configured profiles, uploaded data, integrated workflows. Endowment Effect (entry 102) describes the parallel ownership-feeling dynamic. The forward-looking cost of switching at trial-end includes the discarding of all that onboarding investment.

When it breaks

The primary failure is retention engineering without product substance. When sunk-cost retention is the only thing keeping users on the platform, eventual reckoning is just delayed — accumulated discontent compounds into a single bad cancellation experience and broader reputation damage. Capital Inflation describes the parallel signal-depreciation dynamic.

The second failure is audience detection of manipulative retention. Dark-pattern cancellation flows, hidden cancel buttons, and engineered cancellation friction trigger backlash that costs more than the retention preserves. The FTC's 2023 click-to-cancel rule and similar regulatory moves are direct responses to this failure mode.

The third is cultural variation in sunk-cost intensity. Sunk-cost effects are robust across Western experimental samples but show variable magnitudes across cultures with different relationships to face-saving and commitment. Strategic moves that depend on sunk-cost-driven retention transfer inconsistently across markets.

The most expensive failure is strategic lock-in through accumulated escalation. Brands that have built years of capital, organizational identity, and stakeholder narrative around a strategic bet face structural difficulty cutting losses when the bet stops working. Kodak's digital-photography case is the canonical example: the technology was invented in-house in 1975 but sunk-cost commitment to the film business delayed strategic pivot until bankruptcy in 2012.

In the wild

Played straight. Sephora Beauty Insider and Adobe Creative Cloud both run sunk-cost retention with operational substance behind it — the loyalty rewards and product capabilities make the retention welfare-aligned rather than purely extractive. The accumulated-investment story works because the forward-looking value is real.

Inverted. Disruptor positioning explicitly attacks incumbent sunk costs. "Switching is easier than you think" / "we'll buy out your existing contract" / "no long-term commitment" all work by reframing the sunk cost as a trap rather than an asset. T-Mobile's "Un-carrier" positioning and various SaaS migration-incentive programs run this play.

Subverted. Work that comments on the framework directly — content that names the sunk-cost trap, retention flows that explicitly tell users "your past usage shouldn't determine your future" — uses audience awareness of the mechanism as creative material.

Averted. Pure-transactional categories where there is no accumulated investment to leverage. Commodity-purchase decisions reset every transaction, and sunk-cost retention machinery doesn't have anything to grip.

Canonical examples

Arkes-Blumer 1985 OBHDP foundational research

Hal Arkes and Catherine Blumer's 1985 Organizational Behavior and Human Decision Processes paper "The Psychology of Sunk Cost" is the canonical empirical foundation. The theater-ticket experiments — subjects who paid for tickets attended in worse weather than subjects given equivalent tickets free — established the effect across multiple replications. The paper has accumulated several thousand citations and remained a reference in behavioral-economics teaching .

Concorde program (1962-2003)

The Anglo-French Concorde supersonic-aircraft program is the canonical case of strategic sunk-cost escalation, to the point that "Concorde fallacy" is a synonym for the bias. The 1962 intergovernmental agreement committed both governments to the program; cost overruns, weak commercial demand, and accumulating evidence that the economics didn't work failed to override the political and capital sunk costs. The program ran for 27 years of commercial service and was retired in 2003 following the Air France 4590 disaster in July 2000. Initial development costs ran into the billions of pounds . Canonical case of national-scale sunk-cost lock-in.

Quibi (April 2020 - October 2020)

Quibi, the Jeffrey Katzenberg / Meg Whitman short-form streaming venture, launched April 2020 and shut down October 2020 — six months of operation at significant cumulative loss. The case is structurally instructive: the founders raised and committed roughly $1.75B before launch, kept investing through the launch period despite weak engagement signals, and ultimately closed only when continuing was no longer financially possible . Canonical case of sunk-cost dynamics at the venture-capital scale.

Meta Reality Labs (2014 onward)

Meta's Reality Labs (formerly Oculus, acquired 2014) is the canonical contemporary case of strategic sunk-cost dynamics still in progress. The division has accumulated substantial cumulative operating losses across the post-acquisition period . The strategic question of whether continued investment is forward-looking-rational or sunk-cost-driven is contested in real time, with quarterly earnings calls functioning as the public debate. Canonical case of strategic sunk-cost dynamics under active scrutiny.

Sephora Beauty Insider (2007 onward)

Sephora's Beauty Insider tier-loyalty program (launched 2007, with VIB and Rouge tiers added subsequently) is the canonical case of welfare-aligned sunk-cost retention at commercial scale. Accumulated points, tier status, and member-only access install switching costs, but the program substance — sample programs, exclusive products, beauty-class access — provides genuine forward-looking value. Active membership runs in the tens of millions and accounts for a dominant share of revenue .

Adobe Creative Cloud (2013 onward)

Adobe's transition from perpetual-license Creative Suite to subscription Creative Cloud in 2013 (under Shantanu Narayen) is the canonical contemporary case of subscription sunk-cost retention at scale. Accumulated PSD/AI/PRPROJ files in proprietary formats compound the prior subscription investment with a forward-looking switching cost. Subscriber base runs in the tens of millions with revenue in the multiple tens of billions annually .

Kodak digital-photography (1975-2012)

Eastman Kodak's digital-photography trajectory (already canonical for Counter-Positioning entry 74) deserves a second mention here for the sunk-cost dimension specifically. Steven Sasson invented the digital camera at Kodak in 1975 — Kodak had the technology decades before consumer adoption — but accumulated investment in film, paper, and chemistry created strategic sunk-cost gravity that delayed the pivot until the company filed Chapter 11 in 2012. Canonical case of sunk-cost lock-in producing strategic failure in a category the company had invented.

Netflix subscription retention (1998 onward)

Netflix's subscription operations (already canonical for Mere Exposure Effect entry 97 and Spacing Effect entry 111) deserve a second mention here for the sunk-cost dimension specifically. Accumulated viewing history, recommendation tuning, downloaded content, and watch lists all install retention pressure that compounds with subscription tenure. Subscriber base ran at roughly 270M by 2024 with revenue in the tens of billions . Canonical case of sunk-cost retention at platform scale.

Sunk cost fallacy is the behavioral-economics finding that prior investment produces continued commitment to losing trajectories, with the underlying mechanisms being loss aversion against unrealized losses, commitment-and-consistency pressure, and ego investment in prior decisions. The strategic implication is that brand operations face sunk-cost dynamics as a structural feature of retention, churn, and strategic-pivot decisions — both as something to leverage in customer-relationship architecture and as something to defend against in internal strategic decision-making. Contemporary AI-mediated personalized retention has substantially extended the framework's reach, with regulators now actively engaging the line between welfare-aligned reminder and manipulative retention. The brands that accumulate advantage in sunk-cost-engaged categories tend to be the ones that pair retention machinery with operational substance, calibrate to cultural variation, and maintain the discipline to cut their own losses before sunk-cost lock-in becomes terminal.

Related insights

Sunk Cost Fallacy operates inside Foundational as one of the field's core behavioral-economics frameworks. Prospect Theory (entry 95) describes the loss-aversion dynamic that sunk cost operates through. Anchoring Bias (entry 96) describes the parallel reference-point dynamic. Mere Exposure Effect (entry 97) describes the parallel exposure-frequency dynamic. Cognitive Dissonance (entry 98) describes the parallel post-decision rationalization dynamic that sunk cost amplifies. Cialdini Influence Principles (entry 99) describes the adjacent persuasion architecture, particularly the commitment-and-consistency principle. Peak-End Rule (entry 100) describes the parallel experience-evaluation dynamic. Endowment Effect (entry 102) describes the ownership-feeling dynamic that compounds with sunk cost. IKEA Effect (entry 104) describes the co-creation dynamic that operates inside sunk cost. Goal Gradient Effect (entry 105) describes the progress-toward-completion dynamic. Decision Fatigue (entry 106) describes the cognitive-resource dynamic that makes sunk-cost retention more powerful on tired audiences. Default Effects (entry 107) describes the adjacent default-acceptance dynamic. Framing Effects (entry 108) describes the parallel presentation dynamic. Spacing Effect (entry 111) describes the parallel exposure-distribution dynamic. Confirmation Bias (entry 112) describes the belief-congruent filtering that compounds sunk-cost reasoning. Zeigarnik Effect (entry 114) describes the incomplete-task dynamic that interacts with sunk cost in serialized commitment. Loyalty Programs (entry 64) operates inside sunk-cost dynamics through reward-cadence and tier-progression engineering. CAC-LTV Economics (entry 85) operates inside sunk-cost dynamics through retention assumptions in lifetime-value modeling. Brand Architecture (entry 81) operates inside sunk-cost dynamics through portfolio-retention choices. Crisis Communications (entry 80) operates inside sunk-cost-failure contexts when accumulated investment requires public retraction. Marketing Mix Modeling (entry 84) has to wrestle with sunk-cost retention at the attribution layer. Algorithmic Curation (entry 63) describes the AI-mediated infrastructure that personalizes sunk-cost retention. Heritage Brand Positioning (entry 51) operates inside sunk-cost dynamics through long-horizon equity that compounds with audience tenure. Founder Mythology (entry 72) operates inside sunk-cost dynamics through accumulated narrative investment. Counter-Positioning (entry 74) explicitly weaponizes incumbent sunk costs as a competitive lever. Costly Signals and Commitment Durability describe the operational substance that lets retention engineering land instead of feeling extractive. Manufactured Authenticity describes the failure mode when retention is engineered without that substance. The broader pattern is that sunk-cost dynamics operate whether brands acknowledge them or not, and the brands that pair retention machinery with reliable forward-looking value accumulate advantages over the ones running pure extractive retention or pure ignore-history rational-actor models.