Sunk Cost Fallacy
Past investment that cannot be recovered still drives decisions, in users and in product teams alike.
Definition
The tendency to continue an endeavour because of previously invested resources, despite those resources being unrecoverable and irrelevant to the decision at hand. Arkes and Blumer demonstrated it experimentally in 1985.
Established research. Originates in published research and has been replicated. The effect is real. Its size in your product still needs measuring.
On this plate
Reach for this when
- Reviewing whether to continue a project.
- Designing long flows.
- Evaluating features for deprecation.
- A decision to continue is being justified by what has already been spent.
Common mistakes
- Exploiting the bias to increase completion.
- Citing past effort in a forward-looking decision.
- Confusing commitment with evidence.
Two places it matters
In users, it keeps people in flows they should abandon, and it is exploited by designs that make abandonment feel wasteful after a long form or a lengthy configuration. That is a deceptive pattern when it is deliberate.
In teams, it is why features nobody uses survive, why a failing redesign gets defended, and why a research finding that contradicts six months of work gets explained away. This is usually the more expensive of the two.
Do
- Save progress so leaving a flow is not a loss, which removes the pressure honestly.
- Set decision points in project plans where continuing has to be re-justified.
- Evaluate features on current value, not on what they cost to build.
Do not
- Design flows that make abandonment costly in order to increase completion.
- Use progress investment as a retention mechanism.
- Defend a decision by citing the effort already spent.
Two places it operates
| Who | What it looks like | What helps |
|---|---|---|
| The reader | Continuing a long form or setup that is no longer worth finishing | Save progress, so returning is free and abandoning is not final |
| The reader | Staying on a plan because of what was already paid | Clear, cheap exits, and no penalty for changing |
| The team | Shipping a feature because it is nearly finished | Judge on remaining cost against expected value, not on spend to date |
| The team | Defending a design because of the effort in it | Have someone without the investment run the critique |
In practice
- i.
Saved progress removes the trap
Long formsIf a user can leave and return without losing work, the sunk cost pressure disappears. Completion driven by that pressure is a worse outcome than a later, willing return.
- ii.
Feature deprecation
Product managementThe cost of building a feature is irrelevant to whether it should stay. Usage, maintenance burden and strategic fit are the only inputs that matter.
- iii.
Nearly finished
The team versionA feature at eighty percent was completed and shipped to a usage rate near zero, for the stated reason that it was nearly done. The eighty percent was gone either way, and it was the only argument made.
Local chart
BIAS·014 · Region I · Grid A2
This entry and the entries it is stated to connect with. Each line is labelled with what the connection means.
Sources
Where a source establishes something narrower than the popular reading of it, the note says so.
- [1]
The psychology of sunk cost
H. R. Arkes and C. Blumer, Organizational Behavior and Human Decision Processes, 1985
Continue from here
Each link says what the connection is, so you can tell a principle from an alternative from a thing people mix this up with.
Explains
The mechanism underneath these entries.
- Deceptive PatternsPRIN·011 · Principle
- Pareto PrinciplePRIN·002 · Principle
Applied through
Where this shows up as a concrete interface decision.
- Prioritising Design WorkFLOW·006 · Workflow
Check your work
Review passes that test whether this was done properly.
- Form UX ChecklistCHK·001 · Checklist
Short definition
The one-paragraph version, for when that is all you need.
- Technical DebtTERM·097 · Term
Related concept
Connected closely enough to change how you apply this.
- Confirmation BiasBIAS·009 · Cognitive bias
- Endowment EffectBIAS·007 · Cognitive bias
- Loss AversionBIAS·002 · Cognitive bias