CompanyGraph
Why past spending can distort the next decision, and how to tell escalation from a justified continuation.
The money is gone; the pressure is not
A sunk cost is a past expenditure that cannot be recovered by choosing differently now. The rational question for the next decision compares future costs, future benefits, risk, and alternatives. Past spending may still provide information about what has been learned, but it is not a future cash flow that the next phase can earn back.
In practice, the expenditure changes the social and organizational conditions around the choice. A manager may have announced the project, a team may identify with it, and a board may fear admitting error. The pressure can make continuation feel safer even when the next dollar has a poor expected return.
Research distinguishes effect from rule
Arkes and Blumer's 1985 experiments documented a sunk-cost effect: people were more likely to continue an endeavor after an unrecoverable investment, even when the prior expenditure should not affect the new choice. Staw's work on escalation of commitment examined how personal responsibility and negative outcomes can increase commitment to a chosen course. These studies establish tendencies under particular tasks; they do not show that every continuation is irrational or that a project should be abandoned at the first setback. Arkes and Blumer's study and Staw's escalation study provide the provenance and boundaries.
Organizations add their own lock-in
Personal responsibility can create reputational pressure. A team may filter evidence to protect its work or reinterpret each delay as a reason to fund one more phase. Budgets and reporting systems can also reward spending completion rather than useful output. The result is not just a cognitive bias; it is a governance problem about who can stop a project and how the decision is reviewed.
Escalation becomes more likely when the objective is vague, milestones are moved, the same sponsor controls both investment and review, and the alternative uses of the resources are not shown. It becomes less likely when an independent group reviews evidence, a new decision-maker receives the same data without the history, and stop criteria are set before the next commitment.
Future evidence can justify continuation
Not every unfinished project is a failure. A prototype may have produced knowledge that improves the next design. A regulatory approval may be delayed but still likely. A market may have changed in the project's favor. Continuing is rational when new information raises the expected value of the next phase, not when the organization merely wants to avoid recognizing the past loss.
Where escalation appears
Infrastructure projects can receive successive approvals after cost overruns because abandonment would leave a partially usable asset and a visible loss. Acquisitions can receive more integration budget after customers and employees depart. Product teams can keep adding features after usage shows that the product has not found a customer problem. In each case, the next decision should be made from the current condition, not the size of the original cheque.
How to protect the next decision
Before funding, state the problem, expected benefit, measurable milestones, time limit, and conditions for stopping. At each gate, report total spending separately from the expected value of the remaining work. Invite a reviewer who is not rewarded for the original decision. Compare continuation with sale, closure, redesign, and alternative projects.
For investors, look for repeated funding after missed milestones, shifting definitions of success, and management explanations that refer to what has already been spent rather than what the next phase can earn. Also look for the opposite error: abandoning a valuable capability because a short-term metric disappointed. The concept is a discipline for future choices, not a command to write off every difficult investment.