Pre-set your exit criteria
Decide in advance what result would make you stop, continue, or scale up.
Why it works
Pre-committing to exit criteria prevents the sunk-cost fallacy — continuing a failing approach because stopping "feels like giving up." It also prevents premature abandonment: people who have not defined success often stop when motivation dips rather than when evidence warrants stopping. Clear criteria move the stopping decision from an emotional moment to a pre-planned evaluation.
How to do it
- Before starting, write three outcomes: the result that would make you stop, the result that would make you continue as-is, and the result that would make you scale up.
- Make each outcome specific enough to be measurable at the end of the trial period.
- Enforce the criteria at the review, even if your feelings in the moment push differently.
Evidence
Pre-commitment and prospective decision-making research shows that decisions made in advance of emotional states are better calibrated than in-the-moment judgments. Sunk-cost effects are robust and addressable by pre-commitment to stopping rules. (observational)
The specific application to personal behavior experiments is practitioner-level guidance; the underlying sunk-cost and pre-commitment mechanisms are well documented.
Sources
- Arkes & Blumer (1985), the sunk cost effect, Organizational Behavior and Human Decision Processes
Common mistake
Setting exit criteria that are vague ("if it feels right") or that require only one type of outcome (continue or stop), missing the possibility of modifying and re-testing.
Practice this with IX Coach
7 days free, then $40/month (~$1.30/day).
More practices for Tiny Experiments: Testing Change Without Committing to It
- The time-boxed trial
Run every new behavior as an explicit test with a fixed end date.
- Minimum viable version of the behavior
Reduce the new behavior to the smallest unit that still tests the hypothesis.
- Write the hypothesis before you act
State what you expect the experiment to reveal before it begins.
- Treat failure as data, not verdict
A failed experiment that teaches you something is a successful experiment.
- Run experiments in rapid iteration sprints
Cycle through short tests quickly to compress your learning rate.
- Design the experiment as an environmental probe
Test one specific context change rather than your willpower.