Memory and attention
Peak-end rule
How an experience ends can influence how it is remembered. Make completion useful and honest.
The idea in plain language
What it means.
Research on retrospective evaluation shows that salient moments and endings can influence people's later assessment of an experience. That does not make the rest of the journey disposable. A clear ending can confirm what happened, explain remaining work and give the person a useful next step.
Evidence and interpretation
What supports it.
Experiments on retrospective affect and observational research on memories of medical procedures. Applying these findings to product journeys is an inference, not a promise that a pleasant finish outweighs earlier friction.
From principle to interface
Where you could use it.
Proposed applications, not tested outcomes. The expected effects are hypotheses to check with realistic tasks and users.
Application 1
Finish an application
Show the submission reference, what was received, the expected review process and a way to retrieve the record.
Expected effect to test
The applicant leaves with a clear account of the completed step.
Keep in view
Submitted is different from approved; keep those states distinct.
Application 2
Recover from a failed import
Explain which rows were imported, preserve the rejected rows and provide a corrected retry path.
Expected effect to test
The person can finish with a workable recovery plan.
Keep in view
A friendly message cannot compensate for missing or silently changed data.
Turn the idea into a check
Use it deliberately.
- Define a truthful completion state for the task.
- Give people a useful receipt, result or recovery action.
- Study the whole journey as well as the final screen.
Where the rule gets complicated
Keep the limits attached.
- Peak and end effects are context-dependent, not an exact formula for every experience.
- Do not prolong an unpleasant task to improve its remembered ending.
- Completion quality includes correctness and recoverability, not just celebration.
A named principle does not settle an accessibility requirement, prove a conversion lift or replace testing the particular interface.
Application and implementation boundaries
What the application check shows.
The local success and failure paths used accurate completion language, retained input after failure and allowed synthetic recovery. No retrospective satisfaction was measured.
Methods: Local application example. A completed check covers the stated sample; it is not a safety certification.
- Target
- Original IonBin synthetic interface example; no external product or human-participant study.
- Version or capture
- 2026-10-19 local fixture
- Integrity scope
- SHA-256 f85d1056afd667225a1e0364edfb8f25b4693ffa7b2b8c7d7af5b92338b85ee4 identifies the exact HTML fixture only.
Test conditions
- 155.0.8059.82 Chromium; fresh context with all network requests blocked.
- Synthetic inputs, memory-only state, reduced-motion preference, 1280px and 320px viewports.
Individual checks and observations
Recovery
passLocal application example: Exercise the original local application fixture and inspect the resulting controls or DOM state.
Observed: Confirmed deletion changed only memory state; Restore brought it back.
Honest failure
passLocal application example: Exercise the original local application fixture and inspect the resulting controls or DOM state.
Observed: Failure was stated in text and did not erase the input or claim a successful save.
Honest completion
passLocal application example: Exercise the original local application fixture and inspect the resulting controls or DOM state.
Observed: Completion offered review/edit wording after the simulated success state.
Security findings and boundaries
Review of the local example’s stated interaction boundaries; no backend, authentication, production permissions or external service was exercised.
Practical advice · caution
Completion must match what actually happened
Do not celebrate a submission that is pending, rejected or only partially processed. Show what remains and how to recover.
Practical caveats
- Test the failure ending as carefully as the successful one; a pleasant finish does not repair lost data.
What remains untested
- This checks an authored application example, not the truth, effect size or generality of the named principle.
- No users, assistive-technology participants, physical devices or production task outcomes were evaluated.
Security limits
- The example has no real authorization or persistence layer; its UI behavior cannot establish the security of a deployed workflow.
Evaluation sources
- Fredrickson and Kahneman: Duration Neglect (1993)
Author abstract describing two affective-experience experiments.
- Redelmeier and Kahneman: Patients' Memories of Painful Medical Treatments (1996)
Original observational study linking retrospective evaluation with peak and final pain; author abstract reviewed. This guide provides no medical advice.
Go to the source
Research and further reading.
- Fredrickson and Kahneman: Duration Neglect (1993)
Author abstract describing two affective-experience experiments.
- Redelmeier and Kahneman: Patients' Memories of Painful Medical Treatments (1996)
Original observational study linking retrospective evaluation with peak and final pain; author abstract reviewed. This guide provides no medical advice.
Related questions
More on memory and attention.
Research finding
Miller's Law and working memory
Keep needed information available instead of making people hold it in memory.
Research finding
Von Restorff effect
A meaningful difference can help something stand out in memory. Make the difference worth noticing.
Research finding
Serial position effect
Order can affect recall. Choose an order that also makes the task easier to complete.