Peak-End Composition

Author the experience’s most intense moment and its final moment together, and accept that the rest will be largely forgotten.

Understand This First

• Peak-End Rule — the cognitive finding the pattern operationalizes; without it, the budget-reallocation move below has no warrant.
• Experiencing Self vs. Remembering Self — the dual-target framework the pattern serves; the peak and the end are addressed to one of the two selves, and the trade-off below is between them.
• Duration Neglect — the companion finding that licenses the under-investment in the middle.

Context

A designed experience of bounded duration: a hotel stay, a museum visit, a themed-attraction ride, a tasting menu, an immersive-theatre run, a guided tour, a brand activation, a flagship-store visit. The operator has a finite budget of attention, money, staff training, lighting, scent, choreography, and architectural attention to spend. The guest’s evaluation of the experience will be sampled at least twice: once during, by whatever real-time mechanism the operator runs, and once after, by review, by net promoter score, by booking behavior, by what the guest says to a friend. The two samples will disagree, and the second one is the one that books the next visit.

The pattern lives at the composition scale, between strategic positioning at the macro end (an Aman versus a Marriott; a Royal Academy exhibition versus a kunsthalle program) and the moment-scale tactical patterns (a single threshold, a single signature scent, a single greeting line) at the micro end. Composition is where the operator decides which moments are load-bearing and what each one’s budget share looks like. The experience-economy framing makes the priced product legible; this pattern is how that product is actually staged.

Problem

Designers and operators want every moment of an experience to be equally good. The instinct is honest. Even pacing reads as professional, and uneven pacing reads as failure of attention somewhere along the line. The result is a budget allocation that resembles a flat line: a similar amount of design care, staff hours, and operator focus on every interval of the duration.

The cognitive evidence says this is the wrong loss function. The remembered quality of an extended experience tracks an average of two anchor points (the peak and the end), with duration neglected and the rest contributing little to the summary judgment. A flat budget produces a flat curve, and a flat curve loses to a curve with two well-engineered spikes. The same total spend, redistributed, produces a stronger memory and a higher rebooking probability.

So the practitioner faces a working dilemma. Even pacing feels like the conscientious choice and reads as such on the on-property pulse survey. Uneven pacing feels like a confession that some moments matter more than others, which is true, but politically uncomfortable to defend in an operations review. The operator who wants to act on the cognitive evidence has to accept a budget allocation that won’t plot as a smooth line on any in-stay metric.

Forces

• Average-utility instinct vs. peak-end evidence. The professional instinct toward even pacing optimizes the experiencing self’s running average; the cognitive evidence says the remembering self prices the curve’s anchors instead.
• Operational consistency vs. compositional emphasis. Operations rewards predictable delivery at every interval; composition asks for deliberate amplification at two intervals and deliberate restraint at the rest.
• In-stay metrics vs. retrospective metrics. The on-property pulse samples the experiencing self; the post-stay survey samples the remembering self. Optimizing one degrades the other unless the brief names the trade.
• Politically defensible budget vs. compositionally honest budget. A budget that treats every moment as equally important is easier to defend in a review meeting; a budget that explicitly deprecates the middle is easier to defend on the rebooking dashboard.
• Restraint vs. saturation. Two strong moments of equal intensity blur into each other. The pattern asks the operator to choose which is the peak and which is the end and to differentiate them by intensity, register, or sensory channel.

Solution

Author the peak and the end deliberately, share the budget asymmetrically across them and the rest, and treat the middle as the operational floor rather than the design ceiling.

A working application has four mechanical steps and one editorial one.

The first step is to name the two anchor moments the brief is optimizing for. Specify the peak by its location in the experience’s sequence, by the sensory channels it engages, and by the intended emotional register (awe, recognition, surprise, catharsis, intimacy, release). Specify the end the same way. The pair should not be of identical intensity or register; an experience composed of two equal-and-similar peaks summarizes as one peak, with the second moment dampening rather than reinforcing the first.

The second step is to budget against the pair. The exact split is setting-specific, and the literature doesn’t publish a universal ratio, but the working heuristic across hospitality, themed entertainment, and tasting-menu service has converged on something like a 40 / 30 / 30 distribution: roughly forty percent of marginal compositional spend at the peak, thirty at the end, thirty across the rest. The middle should not collapse to zero (operational consistency is what holds the average above the floor), but the differential is real, and the brief should make it explicit.

The third step is to engineer the peak’s delivery against an operational floor at the rest. The middle is where the experience’s reliability is purchased. Service-blueprint discipline, staff training, baseline cleanliness, predictable wayfinding, the absence of friction: these are the floor. They are not where the peak lives; they are what makes the peak safe to spend on. An operator whose middle is unreliable cannot afford to redirect budget to the peak, because the in-stay average will fall below the threshold the experiencing self forgives.

The fourth step is to differentiate the peak and the end by register. If the peak is high-arousal (a fireworks finale, a chef-out-of-kitchen moment, a reveal), the end should run at a different temperature: lower-arousal-but-high-recognition (a hand-written note, a personal escort, a small ritualized gesture). If the peak is intimate (a private one-on-one moment, a recognition ritual), the end can run higher. The two anchor points are most memorable when they differ; the variance between them is part of what the remembering self encodes.

The editorial step that holds the four mechanical ones together is naming the trade. The brief explicitly states that the average in-stay rating will be lower than it would be under flat pacing, and that the post-stay rating, the review distribution, and the rebooking conversion will be higher. The trade is real; making it explicit prevents the operations review from second-guessing the composition every time the in-stay pulse dips in the middle.

Sensory Channels

• Primary at the peak: setting-specific. Visual at the Disney fireworks ($5{,}000 in pyrotechnics across roughly 12 minutes); gustatory at the Eleven Madison Park signature dessert reveal; auditory at the Punchdrunk Sleep No More waltz finale; haptic at the Aman in-stay turndown ritual.
• Primary at the end: typically lower-arousal multimodal. Visual + auditory at the Disneyland Main Street exit (warmer color temperature, instrumental music); haptic at the Aman send-off (paper bag in hand, escort to the elevator); olfactory at the high-craft restaurant’s parting petit-four (a scent the guest carries on their hands into the cab).
• Tertiary across both anchors: explicit register differentiation between peak and end. If the peak runs at, say, 65 dB ambient and 50 lux, the end should sit roughly 10–15 dB lower and 20–30% dimmer. The exact numbers are setting-specific; the directional rule is constant.

Inheres-In

• Primary: transposable. The pattern’s canonical form lives nowhere; it is a cross-setting compositional move.
• Settings where the pattern is canonical: hospitality, themed-entertainment, immersive-theatre, museum, service-flow.
• Settings where the pattern transposes with care: retail (a flagship visit’s “peak” is often the product try-on or the personalization moment, and the “end” is the bag-hand-over and the parting line); brand-experience (a pop-up’s peak is typically the shareable moment, and the end is the giveaway hand-off); mixed-channel-cx (the digital-physical seam needs an end-of-session anchor that the channel switch tends to flatten).
• Does not transpose: open-ended attractions without a definable end (a public square, a continuously running ambient installation) — the pattern presupposes a bounded duration with a recognizable terminus.

How It Plays Out

Three cases at three durations and three sensory registers illustrate the same compositional move.

Disney’s “Happily Ever After” / “Wonderful World of Animation” park-close fireworks (Walt Disney Imagineering, Walt Disney Studios, current rotation). The Walt Disney Company’s parks have been composed around a deliberate end-of-day peak since the 1955 opening of Disneyland, with Walt Disney’s instruction that the close of the day “must equal the first.” The contemporary version at Magic Kingdom (Happily Ever After, 2017–2023, returning in 2025) and at Disney Hollywood Studios (Wonderful World of Animation, 2019–) is a roughly 18-minute coordinated peak: pyrotechnics, projection-mapped facade animation, a scored John Williams–and-Alan Menken cue list, and choreographed crowd lighting. The peak is followed, deliberately, by a quieter end: the warm-color-temperature ramp on Main Street, U.S.A.; a bandstand instrumental version of the closing music; cast members thanking departing guests by name where possible — the “kiss goodnight” sequence documented in The Imagineering Field Guide to the Magic Kingdom at Walt Disney World and traceable across forty years of trade-press coverage. The pair is exactly the pattern: a high-arousal peak (visual + auditory + chest-thump) followed by a lower-arousal end (warm light, slower walk, named recognition), with the middle of the day’s parade and queue programming explicitly under-budgeted relative to the close. The operator’s working evidence is per-guest spend on next-visit booking and pass-renewal lift, both tracked internally and frequently described in Cornell Hospitality Quarterly features on themed-attraction guest-experience design.

Eleven Madison Park’s tasting menu close (Daniel Humm and Will Guidara at Eleven Madison Park, NYC, 1998–present in current ownership form; Guidara’s Unreasonable Hospitality, 2022, as the playbook). The Eleven Madison Park tasting menu runs roughly two-and-a-half hours across nine to twelve courses. The composition treats the courses themselves as the experiencing-self middle: a flat operational floor where every course meets a defensible standard. The peak is a single signature reveal — historically the savory cheddar-and-apple cracker that opens an “Eleven Madison Park, The Story” reveal of the restaurant’s New York lineage; in subsequent menus, the dessert presentation that arrives with table service from the chef-de-cuisine. The end is a hand-off the guest carries home: a small box with a granola or chocolate, sometimes accompanied by a card noting the night’s date, occasionally — when the staff recognized a celebration during service — a note from the table’s server. Guidara’s published account of the in-house service practice (Unreasonable Hospitality, Chapter 7 on signature moments and Chapter 12 on the parting gesture) names the discipline directly: the budget for staff attention, plating elaboration, and chef visibility is asymmetrically loaded toward the signature reveal and the parting hand-off, with the courses in between operating to a tightly held floor. The peak is high-arousal (table service, recognition, theater); the end is low-arousal-high-recognition (a small object, a name on a card). The pair differs by register, in line with the differentiation rule.

Punchdrunk’s Sleep No More waltz finale (Punchdrunk and Emursive, the McKittrick Hotel, NY, 2011–2024). The 13-year New York run of Sleep No More composed a roughly three-hour immersive-theatre experience around a single peak and a single end that were spatially co-located but emotionally distinct. The peak was the third-act “Macbeth banquet” sequence — the entire cast converging in a single ballroom for a slow-motion choreographed scene with live cello, peaking on the Macbeth-Lady Macbeth waltz that the production was named for among returning audiences. The end was the lobby exit through the Manderley Bar: a 1930s-styled jazz lounge where the show’s masks came off, a live torch-singer set continued the music’s mood at lower intensity, and audiences could either leave or sit with their experience for another twenty minutes. The peak ran at high arousal (full cast, live music, near-sacrificial slow-motion); the end ran at lower arousal but at the highest intimate recognition the show offered: the unmasking, the quiet bar, the chance to talk to a stranger about what one had just witnessed. The middle of the experience — the wandering, the partial scenes, the missed encounters — was deliberately unbudgeted in the average-utility sense; many audiences saw less than half of the available scenes on a given visit. Studies in Theatre and Performance and Theatre Journal coverage of the production through its decade-long run cited the composition as the production’s central design move; The New York Times and The New Yorker features through the same period treated the pair as the show’s actual product.

The three cases run at three durations (an evening at a theme park; two-and-a-half hours at a restaurant; three hours at an immersive-theatre run) and at three price tiers (≈ USD 130 to 250 a head; ≈ USD 365 a person at the time of EMP’s three-Michelin-star tenure; ≈ USD 200 to 350 a head). They are the same move at three settings: load the budget on a peak and an end, differentiate the two by register, hold the middle to an operational floor, and accept that the average in-stay metric will read lower than a flat-budget alternative would have produced.

Consequences

What the pattern buys, what it costs.

The pattern reliably lifts the metrics that price the remembering self: post-stay surveys, reviews, NPS at one-week and one-month, rebooking conversion, the friend-told-a-friend chain. The lift is largest in experiences of moderate length (a few hours to a day or two) and a single dominant valence; multi-day experiences with multiple peaks see a smaller per-peak lift but still benefit from a strong final morning.

It also lets the brief act as a budget instrument rather than a slogan. Once the peak and the end are named in the brief, every cost line becomes traceable: this lighting spec is for the peak; this staff training is for the floor; this furniture purchase is for the end; this signage is for the floor. Reviewers can ask whether a cost belongs to the peak budget or to the floor budget and get a defensible answer.

The pattern carries three honest costs.

It depresses the in-stay running-average metric. The middle is, by design, less spent-on than the peak and the end. An operator who measures only the on-property pulse will see numbers below those of a flat-budget competitor and will be tempted to redirect spend back toward the middle. The trade is fine, but it has to be defended on the right metric.

It raises the consequences of an end that goes wrong. A flat budget distributes risk across the duration; a peak-and-end-loaded budget concentrates it. A failed end (a disorganized checkout, a forgettable curtain call, a fireworks rain-out without recovery) hits the remembered curve harder than a flat-budget competitor’s failed-anywhere-in-the-middle would. The risk is asymmetric and known; the recovery patterns (Service Recovery Theatre) carry disproportionate weight here.

It tempts saturation. Once peak-engineering becomes the in-house craft, every moment of the experience starts to read as a candidate peak, and the operator drifts toward two, three, four peaks in a row. The drift is the antipattern: a saturation curve in which every moment claims to be the most important, with the result that no moment is. The differentiation rule above is the discipline that holds saturation off; the section’s Ritual Saturation entry names the failure mode at the service-ritual scale.

Failure Modes

• Two-peak saturation. The composition runs two equal peaks of similar register. The remembering self averages them into one diffuse peak; the differentiation rule has been violated.
• Peak without floor. The middle is allowed to run below the operational threshold. The peak is impressive in isolation but does not survive contact with an unsatisfied experiencing self; the in-stay frustration corrupts the post-stay summary. Aman’s five-star floor is what makes Aman’s send-off legible as a peak; a comparable send-off at a three-star property fights against the in-stay average and often loses.
• End without peak. The composition has a strong farewell but no engineered peak earlier in the experience. The end has nothing to anchor against; the remembered curve summarizes as “a mediocre experience with a memorable parting moment,” which is the failure mode the sector calls “all end, no journey.”
• Mis-transposition into open-ended formats. The pattern is applied to an experience without a definable end (a public space, a continuously running installation, a brand pop-up with rolling entry). The end-anchor has nowhere to land, and the peak collapses into a single attraction rather than a composed pair.
• Cultural miscalibration. The composition assumes the guest’s individual remembered curve as the unit of analysis; in cultures where the remembering is socially mediated (a family unit, a religious community, a business delegation), the composed peak and the composed end need to land at the social-narrative scale, not the individual-narrative one. A composed peak that the group cannot retell collectively is a peak that won’t compound.
• Saturation drift over operational time. The operator’s craft improves; every moment becomes a candidate peak; the composition drifts toward continuous peak-engineering and into Ritual Saturation. The discipline that holds saturation off is the explicit reservation of two anchor moments and the explicit deprecation of the rest.
• Optimizing for the surveyed peak instead of the felt peak. The operator engineers the peak that scores highest in the post-stay survey (the photogenic moment, the social-share-friendly reveal) rather than the one that does the most for the guest’s actual remembering self. The risk is the Shareable Moment drift: the peak optimized for the share is sometimes — not always — a peak the guest’s remembering self files away as performance rather than as experience.

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 cold-pressor study and the founding statement of the rule the pattern operationalizes.
• 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; the source for the dual-target framing the pattern’s solution depends on.
• B. Joseph Pine II and James H. Gilmore, The Experience Economy: Competing for Customer Time, Attention, and Money (Harvard Business Review Press, updated edition 2019), originally published 1999, especially the chapter on “memorabilia” and the staging-design discussion in the second-edition afterword. The framing that lets a hotel director, a museum director, or a producer treat the staged composition as the priced product.
• Will Guidara, Unreasonable Hospitality: The Remarkable Power of Giving People More Than They Expect (Optimism Press, 2022), Chapters 7 and 12 in particular. Eleven Madison Park’s working playbook for the signature-moment / parting-gesture composition; the source for the in-house translation of peak-end thinking into a hospitality-service operating procedure.
• 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 that established the rule’s robustness outside the lab and grounded the practical recommendation that extending an experience with a milder ending improves remembered evaluation.
• Bernd H. Schmitt, Experiential Marketing: How to Get Customers to Sense, Feel, Think, Act, and Relate to Your Company and Brands (Free Press, 1999), the chapters on experience modules and on the retail and entertainment cases. The first widely-cited translation of peak-end thinking into a working brand-and-experience composition vocabulary; the source the pattern’s “register differentiation” rule is most directly indebted to.