Peak-end rule
People remember an experience by its emotional peak and its ending — not by its duration or its average. A core finding from Daniel Kahneman’s work on the “remembering self” vs the “experiencing self.” The 1993 cold-water study is canonical: subjects who held their hand in painfully cold water for 60 seconds, then 30 additional seconds at a slightly less cold temperature, reported the second trial as less unpleasant overall than the first — despite the second trial being objectively worse (60s of equivalent pain plus 30s of additional pain). The ending was milder, so the memory was milder.
The rule generalizes broadly:
- Same total pain, different ending → different memory
- Same total joy, different peak → different memory
- Duration almost doesn’t matter (“duration neglect” is part of the same finding)
For product designers, the implication is severe: what users remember of an experience determines whether they return — and what they remember is shaped much more by the moments of peak emotion and the closing beat than by anything else.
In onboarding — Marathon’s progress bar
Per The Hidden App Growth Killer (video), Tim Gabe cites the peak-end rule (alongside Goal-gradient effect) for Marathon’s top progress bar during onboarding. The reasoning:
Visual progress cues deliver motivation in bite-sized doses. Show just enough progress to build momentum.
The progress bar shapes the emotional arc of onboarding — small peaks at each completed step, a clear high-note ending when the bar fills. Without it, the experience is undifferentiated and the user’s memory of it is unfavorably blurred.
Operational implications
- Don’t end onboarding on a chore. End on the Aha moment, on a celebratory beat, or on a success state — whatever the experience’s peak is, anchor it near the end.
- Don’t fight duration when you can shape the curve. A 60-screen Duolingo flow that ends on first-lesson satisfaction beats a 4-screen flow that drops the user into emptiness. The remembered experience is the curve, not the integral.
- A bad finish poisons a good middle. If the last interaction is a permission denial, an error, or a confusing screen, the entire flow is recalled as confusing.
Where the peak-end rule fits in Kahneman’s apparatus
The peak-end rule sits alongside the other Kahneman/Tversky biases this wiki tracks:
| Bias | Locus |
|---|---|
| Loss aversion | Outcome valence asymmetry |
| Framing effect | Linguistic presentation |
| Anchoring bias | First-information bias |
| Certainty effect | Deterministic-vs-probabilistic asymmetry |
| Peak-end rule | Memory of experience vs experience itself |
It’s the most directly experience-design-relevant of the five — the others govern point-in-time decisions; the peak-end rule governs the recall that drives return behavior.
Related
- Daniel Kahneman — the upstream theorist; “remembering self” vs “experiencing self”
- Aha moment — the natural “peak” candidate for a product experience
- Goal-gradient effect — stacks with peak-end in onboarding (Marathon case)
- Onboarding flow — peak-end is one of the central design constraints for the flow’s shape
- Gift vs receipt — Tim Gabe’s three-stage framework (anticipation → reveal → afterglow) is essentially a peak-and-ending design pattern: the afterglow stage is the ending engineered to be remembered
Sources
- The Hidden App Growth Killer (video) — Marathon onboarding progress bar; direct attribution