Berlin’s new airport was supposed to open in 2011 on a budget of roughly 2 billion euros. It opened in 2020, having cost more than 6 billion. Across nine years of delays, fire-safety failures, and reworked ductwork, someone in a decision seat kept reaching the same conclusion: we have already put too much into this to walk away now. So they put in more. The logic that kept that project alive is the same logic that keeps a doomed software rollout, a bad hire, or a stalled reorg limping through your quarter.
That pull has a name. The sunk cost fallacy is the tendency to keep investing in something because of what you have already spent on it, rather than what you will actually get back from here. And it rarely announces itself as a bias. In the moment it feels like commitment, discipline, and follow-through. It looks like exactly the traits you were promoted for.
I want to argue the opposite of what most management advice implies. Some of the most valuable decisions you make as a manager are the ones where you stop.
The research is older and steadier than most of what you were taught
This is not a productivity-blog concept. It has a fifty-year paper trail.
In 1976, organizational psychologist Barry Staw published a study with the memorable title “Knee Deep in the Big Muddy,” borrowed from a Vietnam-era protest song. He put people in a simulated manager’s chair, made them responsible for an earlier investment decision, and rigged that decision to fail. The finding held up again and again: the people who had made the original call poured more money into the failing course than people who came to it fresh. Staw named this “escalation of commitment.” The deeper you are, the harder you push, precisely because backing out means admitting the first bet was wrong.
A few years later the economist Richard Thaler gave the everyday version. A man pays $300 for a tennis-club membership, develops tennis elbow after two weeks, and keeps playing through the pain because he doesn’t want to waste the $300. The money is gone either way. Playing in pain doesn’t recover it. It just adds a second cost on top of the first.
The most cited demonstration came from psychologists Hal Arkes and Catherine Blumer in 1985. They asked people to imagine they had bought a $100 ticket to a Michigan ski trip, then found a better $50 trip to Wisconsin for the same weekend and bought that ticket too. The trips conflict; neither ticket is refundable. Which do you take? Fifty-four percent chose the Michigan trip, the one they expected to enjoy less, purely because it cost more. They were willing to have a worse weekend to honor money that was already spent.
Underneath all of it sits loss aversion, the finding from Daniel Kahneman and Amos Tversky’s prospect theory that we feel a loss roughly twice as hard as we feel an equivalent gain. Writing off what you have invested registers as a loss right now, today, on your watch. Continuing lets you defer that feeling. So the brain quietly chooses the option that hurts less this afternoon, even when it costs far more over the year.
Why managers get hit harder than anyone
An individual falling for the sunk cost fallacy is fighting one bias. A manager is fighting three at once.
The first is authorship. You did not just observe the failing project; you approved it, staffed it, defended it in the budget meeting. Staw’s whole point was that the person responsible for the original decision is the most likely to escalate. Ownership makes you the worst-positioned person in the building to call it.
The second is audience. Individual sunk cost is a private error. A manager’s version is public. Killing a project you championed means telling your boss, your peers, and your team that you were wrong in front of everyone. That is why escalation of commitment is a broader, nastier problem than the tidy cognitive bias: face-saving, accountability structures, and quarterly reviews all reward the person who “stays the course” and punish the one who admits a miss, even when admitting it is correct.
The third is the story. By the time a project is deep enough to be a problem, it has a narrative. It has a name, a Slack channel, a roadmap slot, a team that has given it eighteen months. Killing it doesn’t feel like editing a spreadsheet. It feels like ending something. That emotional weight is real, and it has nothing to do with whether continuing is a good idea.
What it actually costs to keep going
The Standish Group has tracked technology projects for decades. Their CHAOS research found that a large share of IT projects are eventually cancelled outright, and the ones that limp to completion routinely run well past their original cost estimates. A meaningful piece of that overrun is not bad estimating. It is escalation: money that went in after the writing was on the wall, spent to protect the money that went in before.
You do not need a research sample to see it at scale. Meta’s Reality Labs division, the bet on the metaverse, lost $17.7 billion in 2024 alone and has run up cumulative operating losses north of $70 billion since 2020. Reasonable people can argue whether that is visionary patience or the largest sunk cost trap in corporate history. The point for the rest of us is smaller and more useful: if a company with that much analytical firepower can spend years defending a losing position because of what it has already spent, your team is not immune. You are running the same brain on a smaller budget.
The one I should have killed
Early in a fractional COO engagement, I inherited a systems rollout that was already about fourteen months old and several hundred thousand dollars deep. The vendor kept missing dates. Every workaround spawned two more. The honest read, if I had walked in fresh that morning, was that the platform was wrong for the business and we should cut it.
I did not cut it. I spent another quarter and a good chunk of budget trying to save it, and I can tell you exactly why, because I was thinking it in plain words at the time: we cannot have spent all of this for nothing. That sentence is the fallacy in its purest form. The money spent was gone regardless of what I chose next. The only real question was whether the next dollar and the next month would return more on this platform or somewhere else. They would not. When we finally switched, the replacement was live in about ten weeks. The lesson was not that I misjudged the vendor. It was that I let a decision I had inherited become a decision I was defending.
After that, I stopped trusting my own gut on projects I was personally attached to, and started building the decision to quit into the plan before there was anything to quit.
Practices that beat the pull
You will not out-discipline this bias in the moment, because in the moment it does not feel like a bias. It feels like resolve. So you have to set the traps in advance, when you are calm and uninvested.
Write the kill criteria before you start, not after you’re stuck. Every project of any size should ship with the conditions under which you would stop it: “If we are not at X by this date, we shut it down.” Decide that on day one, in writing, while you can still think clearly. Defining what failure looks like in advance is most of the battle, and it is the same discipline that makes any decision framework worth using.
Run the “would I start this today” test. Forget what you have already spent. If this project did not exist and someone pitched it to you right now, with the current information, the current team’s morale, and the current results, would you fund it? If the answer is no, you are not deciding whether to continue. You are deciding whether to start it fresh, and you already know the answer.
Hand the review to someone who wasn’t in the room. Because authorship is the core problem, the fix is to route the call through people who carry none of it. Andy Grove’s version at Intel was to ask what a brand-new CEO would do if the board fired the current one, then do that yourself. Bring in a peer with no stake, or make the go/no-go a group call among people who did not approve the original bet. This is one more reason to be clear about who actually owns a given decision.
Name the sunk cost out loud. In the review, separate the two numbers explicitly. One column: money and time already spent, which is gone and irrelevant to the choice. Another column: the expected return on the next investment only. Escalation thrives when those two get blended into a single “we’ve come too far” feeling. Split them on a whiteboard and the fallacy has nowhere to hide.
Frame continuing as a cost, because it is. Every quarter you keep a losing project alive is a quarter your best people are not on your best opportunity. That is the real bill, and it is invisible on any budget line. The team stuck saving a dead initiative is the team you can’t put on the work that would actually move the number, which is the same math behind why an over-scheduled team ships less, not more.
Treat a good stop as a win in your own after-action reviews. If your culture only celebrates finishing, you have trained your managers to escalate. Make “we killed this early and redeployed the budget” a story you tell with pride, the same way you would treat any decision worth learning from after the fact. What gets praised gets repeated.
The reframe that makes it easier
The hardest part is emotional, not analytical. Quitting a project feels like failure, and no framework fully dissolves that feeling.
So change what quitting means. Stopping a losing course is not the failure. The failure already happened, back when the project went sideways and the evidence started piling up. Continuing does not undo it. Continuing just funds it. The manager who stops is not the one who failed; that manager is the one who stopped paying for a failure that already occurred.
The money is spent. The time is spent. Neither comes back no matter what you choose next. The only thing you actually control is the next dollar and the next month, and those you can still put somewhere that pays. Grit is knowing when to keep going. Judgment is knowing when what feels like grit is just the fear of admitting you were wrong.