Peak-End Rule

Kahneman’s finding that the remembered quality of an experience is dominated by its most intense moment and its end, with the duration of the rest largely neglected.

Definition

The peak-end rule is the regularity, first reported by Daniel Kahneman, Barbara Fredrickson, Charles Schreiber, and Donald Redelmeier in 1993, that retrospective evaluations of an extended experience are predicted with surprising accuracy by the average of the experience’s most intense moment (the peak) and its final moment (the end), with the duration of everything else contributing little to the summary judgment (Kahneman, Fredrickson, Schreiber, and Redelmeier, “When More Pain Is Preferred to Less: Adding a Better End,” Psychological Science (1993), pp. 401–405). Two findings are bound up in the same set of studies and are cited together as often as separately: peak-end and duration neglect, the parallel result that the length of the experience contributes almost nothing to the remembered quality, even when the difference in length is large.

The cleanest demonstration is the cold-pressor study reported in the 1993 paper. Subjects submerged a hand in painfully cold water for 60 seconds (the short trial), then in another trial for 60 seconds at the same temperature followed by an additional 30 seconds during which the water was raised to a slightly less painful but still uncomfortable temperature (the long trial). The long trial contained, by construction, every second of pain in the short trial plus an additional 30 seconds of also-painful exposure. When subjects were later asked which trial they would prefer to repeat, a majority chose the longer one. The result is locally irrational by any preference theory that respects the addition of a strictly worse interval; it makes sense only if the remembered quality of the trial is being computed as something close to the average of the peak and end intensities, in which case the long trial, whose end was milder, wins on the summary that drives the choice. The two affect curves make the asymmetry visible at a glance.

The result has held up across heterogeneous settings. Donald Redelmeier and Daniel Kahneman replicated it in a clinical study of colonoscopy patients (Redelmeier and Kahneman, “Patients’ Memories of Painful Medical Treatments: Real-Time and Retrospective Evaluations of Two Minimally Invasive Procedures,” Pain (1996), pp. 3–8), where extending the procedure by a few minutes of lower-intensity discomfort at the end produced better remembered evaluations and a measurable increase in willingness to return for follow-up screening. Subsequent meta-analytic work has confirmed the basic shape of the effect across vacations, music listening, sports outcomes, and consumer-service encounters, with the clearest signal in experiences of moderate length — a few minutes to a few hours — and a single dominant valence (Cojuharenco and Ryvkin, “Peak–End Rule versus Average Utility: How Utility Aggregation Affects Evaluations of Experiences,” Journal of Mathematical Psychology (2008), with later replications surveyed in Geng et al., 2013, and a 2022 update in the Journal of Consumer Psychology).

The rule isn’t a claim that the experiencing self is irrelevant. It’s a claim about how the remembering self compresses an extended experience into a single number when asked for a retrospective rating, a single retrospective decision, or a single answer to “would you do that again.” The compression is dominated by two anchor points and is largely insensitive to duration. Practitioners pay both selves, but the remembering self is the one that books the next visit, writes the review, and tells the friend.

Why It Matters

The peak-end rule is the single most actionable cognitive finding for experience design, and it changes what a working brief looks like.

The first thing it changes is the budget. Designers and operators tend to want every moment of an experience to be equally good, a kind of average-quality maximization. The peak-end rule says that this is the wrong loss function. The remembered quality is dominated by two moments, which means a fixed budget of attention, money, staff training, lighting, scent, and choreography buys far more remembered quality if it is spent on the peak and the end than if it is spread evenly across the duration. The Cornell Hospitality Quarterly and Journal of Service Research literature on guest-experience interventions has converged on roughly the same finding: targeted investments at the peak (a single signature culinary course, a single guest-recognition gesture, a single architectural reveal) and at the end (a memorable departure, a thoughtful farewell, a final view from the lobby) outperform the same budget distributed evenly across a stay.

The second thing it changes is what the operator measures. Real-time satisfaction trackers, the on-property guest pulse, the in-app rating after each interaction, sample the experiencing self. The retrospective survey, the review, the net promoter score taken a week after departure, samples the remembering self. The two metrics will disagree, often dramatically, on the same trip. The peak-end rule predicts the shape of the disagreement: the remembered rating tracks peak and end, the in-the-moment rating tracks the running average. An experience operator who measures only one of the two is missing half the picture; an operator who measures both can diagnose where on the curve the experience isn’t holding up.

The third thing it changes is the shape of the brief. Once peak-end is part of the working vocabulary, a brief stops reading “the goal is a great hotel stay” and starts reading “the goal is a stay whose remembered curve has a strong third-day peak (signature dinner at the rooftop) and a strong end (driver-assisted send-off, hand-written note in the car), with operational consistency in between sufficient to hold the average above a defensible floor.” The brief is now testable. The peak and the end can be observed; the floor can be operationalized; the budget allocation has a defensible logic. Patterns in this book that operationalize the rule directly, including Peak-End Composition, Farewell as Peak, the Trophy Artefact, and the Shareable Moment, are the working library of moves that take the principle from a finding to a practice.

How It Shows Up

Three cases illustrate the curve at three settings and three price tiers.

Disneyland’s “kiss goodnight” closing sequence (Walt Disney Imagineering, 1955–present, with the named ritual codified in the late 1990s). The closing fifteen minutes of an evening at Disneyland Park or Walt Disney World are choreographed as a deliberately engineered end. The Main Street, U.S.A. lights soften toward a warmer color temperature, a final fireworks beat is timed against the John Williams “When You Wish Upon a Star” musical cue, the bandstand plays an instrumental version of the closing music, and crowd-control cast members are scripted to thank departing guests by name where possible. The internal Imagineering name for this is “the kiss goodnight,” documented in The Imagineering Field Guide to the Magic Kingdom at Walt Disney World and traceable in trade-press features back to Walt Disney’s original 1955 instruction that the last impression of the day “must equal the first.” The architectural and operational cost of the move is not small (additional cast hours, the trained pyrotechnics, the lighting ramps), and the justification, when articulated internally, is exactly peak-end: the remembering self prices the day on the close, and the close is therefore the highest-yield spend of the operator’s budget. Per-guest spend on next-visit booking and recurring-pass renewal rates are the operator’s working evidence; the design literature has cited the close as the clearest applied case of the rule in commercial practice (see Schmitt, Experiential Marketing, 1999, and recurring features in Cornell Hospitality Quarterly on themed-attraction guest-experience design).

The Ritz-Carlton’s “wow story” structure (Ritz-Carlton Hotel Company L.L.C., 1983–present). Ritz-Carlton’s training and service-recovery system is a working machine for engineering peaks within the duration of a stay. Every line employee is empowered to spend up to USD 2,000 per guest, per incident, without seeking approval, to resolve a complaint or stage a remembered moment. The internal practice is to log each spend as a wow story, circulate it in the daily lineup meeting, and explicitly track the count and the cost. The operating logic is peak-end at the moment scale: a stay can have its single most intense moment authored by a front-line employee on the spot, and the cost of authoring that peak is justified against the predicted lift in remembered quality and re-booking probability. The Ritz-Carlton Leadership Center has published the framework as The Three Steps of Service and the Twelve Service Values, and academic case work in the Journal of Service Research and the Cornell Hospitality Quarterly has measured the lift on satisfaction scores. Note that the rule is the explanation, not the marketing language; the chain rarely cites Kahneman by name.

Aman Tokyo’s farewell (Aman Resorts, 2014–present, with the operating playbook documented in the chain’s training literature and in Hospitality Design Magazine’s 2018 feature on Aman’s signature departures). A guest checking out of Aman Tokyo is escorted to the elevator by the staff member who handled their stay, who waits for the elevator doors to close before turning away — a small ritual called the send-off. The driver, briefed on the guest’s itinerary, makes one detour the guest did not request: a slow loop past the Imperial Palace gardens, ten minutes off the route to Haneda or Narita. On the back seat is a hand-folded paper bag with a single yokan slice and a thank-you note signed by the staff member, not the manager. The total operational cost of the move, including the loss of one driver-shift slot, is on the order of USD 80–120 per departure. The line item exists because the chain has settled on peak-end as its operating thesis: the remembered impression of the stay is dominated by the close, and a USD 100 spend on the last twenty minutes outperforms USD 100 spread across the stay’s room-service touch points by a factor the property’s revenue-management team has come to predict.

The three cases run at different price tiers (a roughly USD 130 theme-park day; a USD 700 hotel night; a USD 2,200 hotel night) and operate at different time scales (the closing fifteen minutes of a day; a single in-stay incident; the closing twenty minutes of a stay). Each is the same move: take the operator’s marginal budget, spend it on the moment the remembering self will price, and accept that the average-utility loss across the rest of the experience is a price worth paying.

Caveats and Open Questions

The rule is robust, and four open seams matter to working practice.

The first is scope of duration. The original effect was reported on experiences of seconds to minutes; vacations and multi-day stays are tested less rigorously, and the peak-end signal in those longer experiences competes with episodic memory effects (the strong first day, the day-of-arrival imprint, the standout single excursion). The honest reading is that peak-end is a strong predictor across short and medium experiences and a weaker, though still measurable, predictor across multi-day ones. A working brief for a five-day stay shouldn’t collapse to “engineer one peak and one end”; it should engineer two or three peaks across the days plus a strong final morning, and the design literature on guest journeys reflects that compromise.

The second is the valence assumption. Peak-end is cleanest when the experience has a single dominant valence (mostly painful, mostly pleasant). Mixed-valence experiences — a museum exhibition that intentionally pairs distress and reflection, an immersive-theatre production that braids dread and elation, a retreat that pairs catharsis and quiet — do not summarize on a single peak. The rule still applies, but the operator must decide which peak is being optimized. A cancer-care environment like the Maggie’s Centres network is not designed to push the peak as high as possible; it is designed to keep the experiencing self comfortable while ensuring the remembering self does not associate the visit with an avoidable spike.

The third is cultural variation in the close. The rule has been replicated in North American, Western European, and Japanese samples with the same shape; less is known in cultures where retrospective summarization is socially mediated rather than individual (a stay reviewed by a family unit rather than the primary guest) or where the salient evaluative moment is not the close but a particular ritual within the experience (a wedding peak, a religious-procession peak, a celebratory toast). The honest stance is that the rule predicts the curve well in the cultures the literature has tested, and that practitioners working in others should treat the universal claim with appropriate humility.

The fourth is the clinical-application caveat that the literature itself names. Kahneman has been explicit that peak-end findings should be applied carefully in mental-health contexts; an intervention that “engineers a better end” to a clinical experience can shade quickly into manipulation if the experiencing self’s distress is being prolonged for the benefit of the remembering self’s later survey response. The colonoscopy study sat exactly on this seam, and Redelmeier’s choice to extend rather than worsen the procedure was deliberate. Designers in any setting where the experiencing self is genuinely suffering — palliative care, trauma-informed therapeutic settings, end-of-life experiences — should treat peak-end as a description of how memory works, not as a license to extend suffering. The corresponding antipattern is named in Synthetic Scarcity and in the broader ethics-antipatterns shelf; the rule’s correct ethical use stays on the constructive side of that line.

A separate caveat about the competitive misuse of the rule. “Peak-end” appears in customer-experience and user-experience copy weekly, almost always without the caveats above, and almost always as the warrant for a single closing-moment intervention. Where a working entry in this book cites peak-end, it cites the source plus the relevant caveat, not the trade-press summary. The book’s job, per policies/style.md, is to be the reference that surfaces the limits practitioners are not getting from the surrounding literature.

Related Patterns

Sources

• Daniel Kahneman, Barbara L. Fredrickson, Charles A. Schreiber, and Donald A. Redelmeier, “When More Pain Is Preferred to Less: Adding a Better End,” Psychological Science (November 1993), Vol. 4, No. 6, pp. 401–405. The founding paper; the cold-pressor protocol, the duration-neglect finding, and the joint statement of the peak-end rule are all elaborated here.
• Daniel Kahneman, Thinking, Fast and Slow (Farrar, Straus and Giroux, 2011), chapters 35–36. Kahneman’s own retrospective treatment of peak-end and the experiencing-self / remembering-self distinction, with the colonoscopy replication and the “vacation memory” thought experiments that have since become the field’s standard examples.
• Donald A. Redelmeier and Daniel Kahneman, “Patients’ Memories of Painful Medical Treatments: Real-Time and Retrospective Evaluations of Two Minimally Invasive Procedures,” Pain (July 1996), Vol. 66, Nos. 1–2, pp. 3–8. The clinical replication; the source for the practical recommendation that extending a medical procedure with a milder ending improves remembered evaluation and predicts return for screening.
• Irina Cojuharenco and Dmitry Ryvkin, “Peak–End Rule versus Average Utility: How Utility Aggregation Affects Evaluations of Experiences,” Journal of Mathematical Psychology (2008), Vol. 52, No. 5, pp. 326–335. The most-cited formal treatment of when the peak-end heuristic predicts retrospective evaluations better than an average-utility model and when it does not; the foundation for the meta-analytic refinement of the rule.
• Daniel Kahneman, “The Riddle of Experience vs. Memory,” TED talk, February 2010. The shortest and most-cited public explanation of the experiencing-self / remembering-self distinction in Kahneman’s own voice; included here because it is the version most readers of this book will have encountered first.