Sunk Cost in Research Projects

A researcher who has spent three months building a signal that still isn't profitable faces a choice: keep refining it, or kill it and move to a different idea. The sunk cost bias is the tendency to keep going because of the three months already spent, even when a fresh researcher looking only at the signal's current prospects would walk away. Those three months are gone regardless of what happens next — they cannot be recovered by continuing, and they say nothing about whether the remaining work is likely to pay off.

The correct question is always forward-looking: given only what is true today (the current state of the signal, the remaining time and capital needed, and the plausible payoff), is continuing the best use of that remaining budget compared with other ideas competing for it? Past effort should only matter insofar as it changed the estimate of future success — for instance, if those three months revealed a genuinely promising sub-lead, that is new information, not a reason rooted in the sunk cost itself.

Research teams manage this by setting kill criteria and time budgets before a project starts, so the decision to stop is made against a pre-committed bar rather than negotiated in the moment against how much has already been invested.