What you’ve already spent is gone. Decide based on what comes next.
The Idea
A sunk cost is money, time, or effort already spent that you can’t recover. Economic theory says it should play no part in what you do next, because no future choice can change it. The sunk cost fallacy is letting it play a part anyway: continuing something because of what has gone into it, rather than because of what it will deliver from here.
The evidence that people do this is long-standing. In a 1985 field study, psychologists Hal Arkes and Catherine Blumer gave some buyers of university theater season tickets an unexpected discount. Those who paid full price attended more plays in the first half of the season, though the shows were identical for everyone. In a companion scenario, most respondents chose a $100 ski trip over a $50 one they had said they would enjoy more.
Organizations show the same pattern at larger scale. Biologists Richard Dawkins and T. R. Carlisle called it the “Concorde fallacy,” after British and French investment in a supersonic airliner that continued long after its commercial prospects had faded. Organizational researchers call the broader pattern escalation of commitment, a term associated with Barry Staw’s 1976 study, in which people who felt responsible for a failing decision tended to put more money into it. Later replications of Staw’s experiment have produced mixed results, a reminder that the conditions behind escalation are still debated.
In ITAM Terms
ITAM is full of big, visible past commitments: multiyear enterprise agreements, perpetual licenses, custom integrations, data center build-outs. The more you’ve invested, the harder it feels to walk away, even when the numbers say you should. Technology is not immune: a 2000 study by Mark Keil, Joan Mann, and Arun Rai estimated that 30 to 40 percent of information systems projects show some degree of escalation.
One caution cuts the other way. Not everything that feels sunk is sunk. Exit penalties, migration effort, retraining, and data extraction are future costs, and they belong in the decision. The discipline is to separate the money that is gone from the money a change would still require.
Where You’ll See It
- Software: “We own perpetual licenses, so we should keep running it.” Maybe, but annual support, patching labor, and security risk are future costs, and they’re the ones that matter.
- Hardware: Keeping a storage array alive because it cost $2 million five years ago, while its maintenance now exceeds the cost of replacement capacity.
- Cloud: Staying on a reserved instance family that no longer fits the workload because “we committed.” Check whether you can exchange or resell the commitment before you bend the architecture around it. The FinOps Foundation’s overview of commitment discounts notes that these options vary by provider and can change.
- AI: Continuing to fund an in-house model after a commercial option overtakes it, because of the months already invested. The FinOps Foundation’s guide to choosing an AI approach and infrastructure strategy sets out the cost trade-offs between self-managed and managed options, which is the comparison an honest review needs.
Governance Takeaway
Build retirement and exit reviews into your lifecycle process, and make them compare only future costs and benefits. Separate “Was this a good purchase?” (useful for learning) from “Should we keep it?” (the actual decision). Where you can, include someone in the review who wasn’t part of the original decision, since the people who made a commitment have the most reason to defend it.
Try This
Pick an asset your organization defends mainly by pointing at what it cost. Model the next three years of keeping it against switching, excluding anything already spent but including every cost of leaving.
