Sunk Cost: Recognising It in Your Own Reasoning

A sunk cost is money, time, or effort you have already spent and cannot get back, regardless of what you decide to do next. The sunk cost fallacy is letting that unrecoverable past spending pull on a decision that should only be weighed by what happens from here forward. Almost everyone can recite this definition without hesitation; it is one of the better-known ideas in this field. What almost no one is good at is catching the fallacy operating inside their own reasoning while it is happening, because it never announces itself as “I am now committing the sunk cost fallacy.” It shows up dressed as a perfectly reasonable-sounding argument: we’ve come this far, it would be a waste to stop now, think about what we’ve already put into this. Each of those phrases feels like a legitimate reason in the moment. This piece is about spotting the disguise, using a real worked example rather than just restating the definition once more.

A worked example: continue or pivot

Suppose a small team has spent six months building a feature that has not gained the traction anyone expected. The question in front of them: continue, or pivot the effort toward something else. Using the weighted pros and cons scorer, someone drafts a first pass at the case for continuing, weighting each point from 1 (minor) to 5 (decisive):

• Pros of continuing — “Six months already invested” (weight 5), “Team is familiar with the codebase” (weight 3), “Some customers are waiting for it” (weight 2).
• Cons of continuing — “Market signal is weak” (weight 4), “Opportunity cost of the higher-demand feature” (weight 4).

Scored as written, the pros total 10 against 8 for the cons, for a lean of 55.6% toward continuing — the tool’s verdict is lean-yes. Now strike the “six months already invested” line and rescore with nothing else changed: the pros total drops to 5 against the same 8 for the cons, the lean falls to 38.5%, and the verdict flips to lean-no — in the plain language of the decision, pivot rather than continue.

Sit with what actually changed between those two runs: nothing about the market, the codebase, the waiting customers, or the opportunity cost moved by a single point. The entire flip from lean-yes to lean-no came from removing one item — a backward-looking fact about money and time that already happened — from a list that is supposed to describe the forward-looking merits of continuing. That one item was worth almost as much, on its own, as the entire case against continuing.

The test that separates a real point from a sunk cost

Notice that not every item connected to “we’ve already started this” belongs on the chopping block. “Team is familiar with the codebase” sounds superficially similar to “six months already invested” — both exist only because of the six months of history — but they are not the same kind of claim, and the second run above correctly kept the first while cutting the second. The test that tells them apart: if you were making this decision completely fresh today, with the situation exactly as it currently stands but with no history behind it, would this fact still be true and still count as a reason? “Six months already invested” fails immediately — it is a statement purely about the past, and it would not exist at all in a decision that started today, because there would be no past spending to refer to. “Team is familiar with the codebase,” on the other hand, passes: even in a hypothetical fresh decision made today, a team that already knows a codebase well is a real, present-tense asset that will reduce future ramp-up time, regardless of how that familiarity was acquired. The history explains how the fact came to be true, but the fact itself describes something real about today and tomorrow, not merely about the past — which is exactly why it belongs in the list and the sunk cost does not.

Where the fallacy likes to disguise itself

A handful of phrases are worth treating as a flag to run the test above, precisely because they so often smuggle a sunk cost in under a more presentable name. “We’ve come this far” is almost always a sunk cost restated as momentum. “It would be a waste to stop now” has the causality backwards — the spending is already gone whether you stop or continue, so continuing does not un-waste it; only a good outcome from here forward would, and that has to be judged on its own future merits. “Think of the team’s morale” and “we already told the board this would work” deserve slightly more care, because they can be pointing at something genuinely forward-looking — a real, present-tense cost to credibility or trust that would exist even in a fresh decision. But they are also exactly the kind of claim people reach for to dress a sunk cost in more defensible clothing, so run the same fresh-decision test on them rather than accepting them at face value: would this reputational or morale cost still be a real, sizeable factor if you were deciding today with no history to protect, or is most of its apparent weight actually coming from discomfort about admitting the past six months did not pan out as hoped?

Sunk cost and opportunity cost are not the same thing

It is worth being precise about a second concept that sits right next to sunk cost in the worked example above, because the two are easy to blur together despite pointing in opposite directions. A sunk cost looks backward: it is spending that has already happened and cannot be changed by any choice available to you now, which is exactly why it should carry zero weight in a forward-looking decision. An opportunity cost looks forward: it is the value of the next-best alternative you give up by choosing one path over another, and it is a completely legitimate, forward-looking factor that belongs in the decision. In the worked example, “opportunity cost of the higher-demand feature” earns its place in the cons column precisely because it describes something that has not happened yet and depends entirely on which choice gets made today — continuing the current feature means the team’s time is not available for the other one, and that trade-off is real and current, not a leftover from the past six months. The two concepts sound similar because both involve the word “cost” and both often appear in the same decision, but only one of them changes based on what you choose next, and that is the one worth weighing.

Why this feels so wrong to override

Understanding the logic of sunk cost intellectually does not make ignoring it feel comfortable, and it helps to know why. Part of it is loss aversion, covered in the biases guide linked above — walking away from six months of work registers as a loss to be avoided, even though that loss already happened the moment the six months were spent, regardless of what you decide today. Part of it is a more specific discomfort sometimes described as waste aversion: stopping a project feels like actively choosing to waste the resources already put into it, when in fact those resources are equally “wasted” whether you stop today or in another six months of continued underperformance — the only thing actually still within your control is whether you spend more time and money before recognizing the same conclusion. And part of it is social: admitting that continuing is no longer the right call can feel like admitting the original decision to start was wrong, even though a decision that was reasonable given what was known at the time can still be reasonably reversed once new information arrives. None of these feelings are irrational to have; they are simply not evidence about the decision in front of you, and treating them as though they were is precisely how a sunk cost sneaks back into the reasoning after being logically ruled out.

Why knowing the concept doesn't stop it

The field guide to common decision-making biases covers the sunk cost fallacy as one entry among twelve, and reading that entry will not, by itself, stop you from falling for it, for the same reason reading any of the other eleven does not immunize you against them: these are automatic patterns of reasoning, not beliefs you hold consciously and can simply choose to abandon once you know better. What actually works is a habit applied at the moment of listing your reasons, not a fact recalled afterward once the decision is already made. Before finalizing any pros-and-cons list, a decision matrix, or any other structured comparison, go through every item on it one at a time and ask the fresh-decision test explicitly: would this still be true and still matter if I were deciding today, with no history behind it? Anything that only survives because of what has already been spent gets struck, the same way “six months already invested” was struck above — not because past effort does not matter emotionally, but because it cannot be recovered by any choice available to you now, and a forward-looking decision should only be weighed by the costs and benefits that are still actually in play.

This is the same discipline, applied to a different failure mode, as checking a weighted decision matrix for rigged inputs: in both cases, the arithmetic is only as honest as the list of items you feed into it, and the moment to catch a bad input is before you look at the total, not after you have already seen a result you like and gone looking for reasons to keep it.