Sunk Cost Fallacy
Past investment is gone either way. Decide based on future costs and benefits only.
To an executive
Money already spent is gone whatever you decide. Judge a struggling programme on what it will cost and deliver from here, not on what it has consumed.
To an engineer
Ask whether you would start this path today knowing what you know now. Make it safe for sponsors and teams to stop, switch or shrink work.
The idea
Money, time and effort already spent cannot be recovered, whichever way you decide next. Yet people keep investing in failing courses of action to avoid "wasting" what was spent. Psychologists Hal Arkes and Catherine Blumer showed this tendency in a series of experiments in the 1980s: the more people had already invested, the more likely they were to continue.
A rational decision looks only forward: given where we are today, which option gives the best future costs and benefits? In large organisations sunk costs are also political. Someone sponsored the programme, and stopping it can feel like admitting failure in public.
Why it works
It removes a cost that cannot change. The money already spent is the same under every option, so it cannot help choose between them.
It reframes stopping as a decision, not a failure. Asking "would we start this today?" puts continuing and stopping on equal footing.
It surfaces the real future cost. Programmes that have run long often hide how much more they need. A forward-looking view forces that estimate into the open.
Where it applies
- Long-running transformation programmes, such as a custom ERP or core-system replacement, that keep missing milestones.
- Vendor and platform commitments, where a large licence or implementation spend makes switching feel wasteful.
- Build versus buy reviews, where an in-house build continues because of the effort already put in.
- Portfolio governance, deciding which initiatives to stop at stage gates.
How to apply
- Separate the past from the future in the business case: show spend to date, but compare options on remaining cost and benefit only.
- Ask the zero-based question: if we were starting today, would we choose this?
- Compare realistic alternatives, including stopping, pivoting and buying.
- Make stopping safe, by agreeing kill criteria in advance and protecting the people involved.
Example scenario
Several years into a custom ERP build, the programme still could not close the books. The board kept funding it "to protect the investment", and each review compared the remaining cost against everything already spent.
An enterprise architect reframed the next review to compare only the future: the cost and time to finish the build, against the cost and time to move to a packaged SaaS ERP. Seen that way, the packaged option was cheaper and faster to go live, and the board changed course.
Questions to ask
- If we were starting today, with what we now know, would we choose this path?
- Are we continuing because of the evidence, or because of what we have already spent?
- What would make it safe, for the sponsor and the team, to stop?
When not to use it
- When past spend has created real assets, such as skills, data or reusable components, that genuinely lower the future cost of continuing. Count those as future benefits, not sunk cost.
How it misleads you
Abandoning work at the first sign of difficulty. Some investments are close to paying off, and switching has its own costs and risks. The model says to ignore the past spend, not to ignore the future cost of starting again.
Further reading
Last reviewed · Suggest an improvement