| name | apply-peak-end-experience-design |
| description | Use when designing a customer, patient, or user experience with multiple sequential moments — a service journey, a medical procedure, a product onboarding flow, an event — to deliberately design the most intense moment and the final moment, because people's remembered evaluation of an entire experience is dominated by its peak and its ending, not by its average moment-to-moment quality or duration. |
| source | Fredrickson & Kahneman, "Duration Neglect in Retrospective Evaluations of Affective Episodes", Journal of Personality and Social Psychology (1993); Redelmeier & Kahneman, "Patients' Memories of Painful Medical Treatments", Pain (1996); Kahneman, Fredrickson, Charles Schreiber & Redelmeier, "When More Pain Is Preferred to Less", Psychological Science (1993) |
| tags | ["customer-experience","service-design","ux","peak-end-rule","product","psychology"] |
| related | ["apply-availability-heuristic-correction"] |
Apply Peak-End Experience Design
Design the single most intense moment and the final moment of a multi-step experience with disproportionate care, because people's remembered evaluation of the entire experience is dominated by those two points, not by its average quality or total duration.
Why This Is Best Practice
Why best: This is not a general claim that "endings matter" — it is a specific, measured finding that the remembering self (the version of a person that evaluates an experience afterward, and that decides whether to repeat it, recommend it, or complain about it) constructs its judgment almost entirely from the peak (best or worst) moment and the final moment, largely discarding information about how long the experience lasted or what the average moment felt like. Because future decisions (repeat business, recommendations, satisfaction scores) are driven by the remembering self's summary, not by a moment-by-moment average, the two most valuable design targets are structurally identifiable in advance: the peak and the end.
Fredrickson & Kahneman (1993): Subjects rated their moment-to-moment affect while watching a series of short film clips, then gave a retrospective overall evaluation. The retrospective evaluations correlated strongly with the peak affect and the affect at the end of each clip, and showed near-zero correlation with the clip's duration — establishing "duration neglect" and the disproportionate weight of peak and end moments as a general property of retrospective evaluation, not specific to any one type of experience.
Redelmeier & Kahneman (1996): Had patients undergoing colonoscopy report their real-time pain at regular intervals during the procedure, then give a retrospective global rating of how painful the procedure was. The retrospective rating was predicted much more strongly by the peak pain level and the pain level in the final minutes of the procedure than by the procedure's total duration or average pain level — direct clinical evidence that the peak-end pattern applies to a real, high-stakes, involuntary experience, not just to voluntarily-viewed film clips.
Kahneman, Fredrickson, Schreiber & Redelmeier (1993) — "when more pain is preferred to less": In a follow-up experiment, subjects underwent one trial with high peak discomfort followed by an abrupt end, and a second, objectively longer trial with the same peak discomfort but ending with a period of gradually decreasing discomfort. Subjects reliably preferred to repeat the longer trial with the better ending, despite it involving strictly more total discomfort by any moment-by-moment sum — demonstrating that a better-designed ending can make an objectively worse (longer, more total discomfort) experience preferred over a shorter one with a worse ending.
Adopted by: Peak-end-rule-informed design is standard practice in customer-experience and service-design consulting (e.g., deliberately engineering a strong final moment in retail checkout, hospitality checkout, and post-purchase follow-up); healthcare providers increasingly design the final minutes of uncomfortable procedures (e.g., tapering discomfort, providing reassurance and clear information at the end) based directly on the Redelmeier & Kahneman colonoscopy findings; theme parks and experience-design firms are widely documented as deliberately sequencing the most intense moments and the finale of a ride or show for peak emotional impact rather than spreading intensity evenly.
Impact: Redelmeier & Kahneman found retrospective procedure ratings were predicted far more strongly by peak pain and end-of-procedure pain than by total duration, directly informing subsequent clinical practice around procedure sequencing; Kahneman et al.'s "more pain preferred to less" experiment demonstrated a reversal of a basic intuition (that less total discomfort is always preferred) purely through end-of-experience design, showing the effect size is large enough to override an objectively worse total outcome.
Steps
-
Map the full sequence of the experience and identify its likely peak moment(s). For any multi-step experience (a service visit, a medical procedure, a product onboarding flow, an event), identify where the most intense positive or negative moment is likely to occur — this is often not something explicitly designed, and finding it is a prerequisite to intentionally shaping it.
-
If the likely peak is negative, redesign to blunt or reposition it; if positive, design to guarantee and amplify it. A predictable negative peak (a painful step, a frustrating friction point, a moment of maximum uncertainty) is the highest-leverage single point to fix — fixing it changes the entire retrospective evaluation disproportionately relative to fixing an average moment elsewhere in the sequence.
-
Design the final moments of the experience with at least as much care as the peak. The end of the experience — the last few minutes of a procedure, the final screen of an onboarding flow, the last interaction before a customer leaves — carries disproportionate weight in the remembered evaluation regardless of what preceded it; ensure the final moments are calibrated to be at least as good as, and ideally better than, the middle of the experience.
-
Where a negative experience cannot be avoided, engineer a de-escalating rather than escalating sequence toward the end. Following the Redelmeier & Kahneman and Kahneman et al. findings, ending a necessarily-uncomfortable experience with gradually decreasing intensity produces a better remembered evaluation than an abrupt stop at the same or lower peak intensity — sequence the unavoidable discomfort earlier when possible, tapering toward the end.
-
Do not over-invest in improving the average, moment-to-moment quality of the middle of the experience at the expense of the peak and the end. Resources spent smoothing out moderate, non-peak moments produce comparatively little improvement in remembered evaluation and satisfaction; check that peak and end design have been addressed before allocating further resources to average-quality improvements.
-
Validate with retrospective (not just moment-by-moment) satisfaction measurement. Since what drives repeat behavior and recommendations is the remembering self's summary, not the moment-by-moment experience, measure post-experience retrospective satisfaction directly (not just an average of in-the-moment ratings) to check whether peak and end design changes are actually improving the metric that predicts real-world behavior.
Rules
- Identify and deliberately design the peak moment and the final moment before investing in improving the average quality of the middle of an experience — those two points carry disproportionate weight in the remembered evaluation that drives repeat behavior and recommendations.
- When an experience must include an unavoidable negative moment, sequence it earlier and taper the intensity down toward the end, rather than allowing it to occur near or as the final moment.
- Measure retrospective, post-experience evaluation directly — a moment-by-moment average satisfaction metric, however precisely measured, does not predict repeat behavior or recommendation as well as the peak-end-weighted remembered evaluation does.
- Do not assume a longer, objectively "better on average" experience is automatically preferred over a shorter one with a worse average but a better ending — the research finding is specifically that ending quality can outweigh total or average quality.
Examples
Healthcare: A clinic performing a necessarily uncomfortable procedure restructures the sequence so the most invasive, uncomfortable steps happen in the middle rather than at the very end, and ensures the final few minutes involve tapering discomfort, clear communication, and reassurance — following the pattern in Redelmeier & Kahneman's colonoscopy findings, patients' retrospective ratings of the procedure improve without reducing the procedure's actual total discomfort or duration.
Retail/hospitality: A hotel chain identifies checkout (the final interaction before a guest leaves) as a peak-end design priority independent of room quality during the stay, and invests specifically in a fast, warm, personalized checkout experience — guest satisfaction and likelihood-to-return scores improve disproportionately relative to the cost of the checkout-specific investment, compared to equivalent spending on marginal improvements to average in-room amenities.
Product onboarding: A software product's onboarding flow has a technically difficult step in the middle (peak friction) followed by an anticlimactic final screen. Redesigning to end the flow with a clear, celebratory "you're set up" moment and a visible next action — while leaving the difficult middle step's actual difficulty largely unchanged — improves completion and satisfaction survey scores, because the redesigned ending dominates the remembered evaluation more than the unchanged peak friction does.
Common Mistakes
- Spending design and engineering effort smoothing the average quality of an experience while leaving an unaddressed negative peak or a weak ending untouched. This produces comparatively little improvement in remembered satisfaction relative to the effort spent, because the average moment carries little weight in retrospective evaluation.
- Ending a necessarily difficult experience on its most intense note rather than tapering. An abrupt stop at high intensity, even if brief, produces a worse remembered evaluation than a longer experience that tapers down before ending.
- Relying only on moment-by-moment or average satisfaction metrics to evaluate experience design changes. These metrics can show no change or even improve while the retrospective, decision-driving evaluation (which determines repeat behavior) moves in the opposite direction.
- Assuming that reducing total experience duration always improves satisfaction. The evidence specifically shows people can and do prefer a longer experience with a better-designed ending over a shorter one with a worse ending.
When NOT to Use
- For experiences with no meaningful peak or distinct ending to design around (e.g., continuous, low-variance background services where no single moment stands out) — the framework requires an identifiable sequence with distinguishable moments.
- When the goal is optimizing objective outcomes (e.g., total pain minimized, total task time minimized) rather than subjective remembered satisfaction — peak-end design specifically targets the remembering self's retrospective evaluation, which can diverge from objective moment-by-moment outcomes; be explicit about which one is the actual design target.
- In safety-critical procedures where sequencing choices driven by remembered-satisfaction optimization would conflict with clinical or safety best practice — safety and correctness take priority over experience design whenever they conflict.
