Sunk Cost Fallacy
Sunk Cost Fallacy
What is the sunk cost fallacy?
- Exclude unrecoverable costs from the decision
- In organizations, avoiding blame amplifies the error
- Best practice: set stop criteria at kickoff
Understanding the sunk cost fallacy changes the first question asked in a meeting. Instead of 'How much have we spent so far?', people ask 'How much more will it take to finish, and what will that money get us?' The concept comes down to moving your baseline from past spending to remaining value versus remaining cost. The KRW 1.8 billion (≈ US$1,346,700) already spent stays on the books whether you continue or stop, so it does not separate the two options, and a number that does not separate the options belongs off the decision-making table.
A common mistake in practice is to treat everything already spent as a sunk cost. If purchased equipment can be resold for 40% of its cost, that 40% is not sunk. The same goes for the labor cost of developers who can be reassigned to other projects. Only what remains after you subtract the recoverable portion is a sunk cost, and this calculation shows what stopping would really cost. You also need to separate it from neighboring concepts. Loss aversion is the tendency to avoid loss itself, whereas escalation of commitment means pouring in more resources to justify your own decision, and in organizations the latter is usually the more dangerous.
A stop criterion works only when it is a number, not a sentence. A rule like 'review if performance is poor' is one nobody can trigger. But a rule that spells out the metric, threshold, and timing, such as 'reopen the gate if, at 60% of budget spent, the pilot defect rate has improved by less than half the target,' lets the calendar, not an individual, call the meeting. Writing down a clause in advance that the project’s failure will not count against the performance review of the person who proposed stopping changes the temperature of what gets said in gate meetings.
The most common mistake in practice is reading 'ignore sunk costs' as 'ignore everything that has happened so far.' Contract termination penalties, penalties for missed client delivery dates, and the cost of reassigning staff who have completed training are future costs that newly arise if you choose to stop, so they must be included in the calculation. Lump them in with sunk costs and you miss that stopping may actually be the more expensive decision. Conversely, judging on short-term metrics alone leads to premature cancellation, with long-payback projects dropped after just two or three quarters.
John KimInstructor John Kim, author of B2B Marketing Design and CEO of Nanum Consulting
Ara LimSpecialist in global leadership, AI, and business communicationSince Hal Arkes and Catherine Blumer systematically demonstrated the sunk cost effect in a 1985 experiment, it has been treated as a classic bias in behavioral economics.
Company A, an auto parts manufacturer, has 420 employees and annual revenue of KRW 120 billion (≈ US$89,780,000). Two years ago it started an MES upgrade and has put in a cumulative KRW 1.8 (≈ US$1) billion, yet only 1 of its 3 lines is integrated, and the target of a 50% cut in the defect rate has produced only a 20% improvement. Executive meetings have ended three times in a row without a decision on whether to spend the remaining KRW 700 million (≈ US$523,700) budget. (Hypothetical example)
- At the start of the meeting, the head of production said, 'Let’s wipe the KRW 1.8 billion (≈ US$1,346,700) and start from there,' and had the room write on the whiteboard only the defect-rate improvement that the remaining KRW 700 million (≈ US$523,700) plus the additional KRW 400 (≈ US$1) million, KRW 1.1 billion (≈ US$823,000) in total, would buy.
- A finance manager worked out the resale value of the servers and the refunds from canceling licenses, and brought the numbers to the meeting: KRW 320 million (≈ US$239,400) of the KRW 1.8 billion (≈ US$1,346,700) was recoverable, so the actual sunk amount was KRW 1.48 billion (≈ US$1,107,300).
- When the head of information strategy asked, 'If this item were brought to you for the first time today, would you approve KRW 1.1 billion (≈ US$823,000)?', 5 of 7 executives answered that for the same money, logistics automation is better.
- The PMO lead pulled out the gate criterion agreed at kickoff (if 60% of the budget is spent and goal attainment is under 50%, review the project) and reported that the trigger condition had already been crossed 8 months earlier.
- The CEO decided against a full shutdown and chose to wind down with a reduced KRW 200 million (≈ US$149,600) budget, stabilizing only 1 line, and had the minutes state that the person who proposed stopping would not be held accountable.
- They redirected the remaining KRW 900 million (≈ US$673,400) to logistics automation, and six months later shipping lead time fell from 3.1 days to 2.2 days. Gate reviews became a standing quarterly agenda item, and two underperforming projects were later closed after 5 and 7 months.