Peak-End Rule in Customer Experience: Kahneman's Research Explained

Ask a patient to rate a colonoscopy on pain, and the answer has almost nothing to do with how long the procedure lasted. It has everything to do with how it ended. Daniel Kahneman proved this in a hospital, not a boardroom — but the finding now explains why your CSAT score can rise for two straight quarters while churn quietly climbs alongside it.

The peak-end rule is a cognitive bias, identified by Nobel laureate Daniel Kahneman, which holds that people judge an experience not by its sum or its average, but almost entirely by two moments: its most intense point (the peak) and how it finished (the end). Everything else — the long, uneventful middle that makes up most of the journey — gets discounted or forgotten. For customer experience leaders, the implication is blunt: you can smooth every touchpoint in a journey and still lose, if you fail to engineer the one or two moments that get remembered.

What exactly is the peak-end rule?

The peak-end rule describes how human memory compresses an experience into a handful of representative moments rather than storing it in full. When someone recalls a flight, a hospital stay, or a call with customer service, they are not running a mental average of every minute. They are retrieving a snapshot: the worst or best instant, and the final one. Kahneman built this into a broader distinction he later popularised in Thinking, Fast and Slow (2011) — between the experiencing self, which lives each moment in real time, and the remembering self, which constructs the story that actually drives future choices, loyalty, and word of mouth.

That distinction matters more than most CX programmes admit. You do not get repeat business from the self that experienced the journey. You get it from the self that remembers it — and that self is a careless editor.

Where does the evidence actually come from?

This is not a borrowed metaphor; it is a tested finding. In a study published in Psychological Science in 1993, Daniel Kahneman, Barbara Fredrickson, Charles Schreiber and Donald Redelmeier had participants submerge a hand in painfully cold water twice: once for 60 seconds, and once for 90 seconds where the final 30 seconds included a slight, still-painful warming. Logically, the longer trial involved strictly more total pain. Yet when asked which trial they would prefer to repeat, most participants chose the longer one — because it ended on a less painful note. The researchers called this duration neglect: the near-total disregard for how long an unpleasant or pleasant experience lasted, in favour of its peak intensity and its ending.

A follow-up field study by Donald Redelmeier and Daniel Kahneman extended the finding into a real clinical setting: patients undergoing colonoscopy reported less retrospective distress when the procedure's final moments were made milder, even when the total duration — and therefore total discomfort — was longer. The medical implication was striking enough to influence how some procedures were later administered. The CX implication is just as striking and far less acted upon: a longer, more effortful journey can be remembered more fondly than a short, efficient one, purely because of how it closes.

Why does this break the way most companies measure experience?

Most voice-of-customer programmes still measure satisfaction as if the remembering self didn't exist — a single survey score meant to represent an entire journey, often triggered at an arbitrary moment, then averaged across thousands of respondents into one tidy number for a dashboard. That number treats a disastrous middle and a graceful recovery as equivalent to a smooth middle and a weak ending, provided the arithmetic mean comes out the same. It never does, in the customer's head.

This is why organisations routinely see CSAT and NPS hold steady, or even improve, while repeat purchase and advocacy quietly erode. The average experience got better; the remembered experience did not. Averages are a System 2 construct — careful, deliberate, built for spreadsheets. Memory of an experience is a System 1 output — fast, emotional, and anchored to extremes. A measurement framework built entirely on averages is measuring the wrong cognitive system.

Customers don't rate the journey. They rate the story their memory kept of it — and the story only has room for the sharpest scene and the last line.

Where do peaks and ends actually sit in a customer journey?

Every journey — booking a flight, opening a bank account, resolving a billing dispute — has a handful of moments that will disproportionately determine the story a customer tells afterwards. Identifying them requires mapping the journey properly, stage by stage, rather than guessing from a boardroom. Typical high-leverage moments include:

• The point of maximum emotional intensity — a declined payment, a flight delay announcement, a diagnosis, a resolved complaint. This is often negative by default, which is precisely why it deserves deliberate design, not neglect.
• The first few seconds of onboarding — not technically an "end," but often the first peak, and it anchors every expectation that follows.
• The handoff between channels — digital to human, self-service to call centre — a frequent source of friction that becomes the unintended peak of a bad journey.
• The final interaction before a decision point — checkout, contract renewal, the last message in a support thread, the moment a guest checks out of a hotel.
• Any moment involving a wait — queues compress poorly into memory on their own, but how a wait ends (with an apology, a small gesture, a resolution) is remembered far more than how long it was.

Service blueprinting is the discipline that surfaces these moments systematically rather than anecdotally — mapping every touchpoint, emotion, and point of friction across the full customer journey so that peaks and endings are identified by evidence, not instinct.

How should a CX leader actually design for the peak-end rule?

Knowing the bias exists is not the same as acting on it. Turning the peak-end rule into an operating discipline follows a reasonably consistent sequence:

1. Map the full journey before touching anything. You cannot engineer a peak or an ending you haven't located. Build or audit the service blueprint stage by stage, capturing emotional highs, lows, and the final interaction in each stage.
2. Score each touchpoint for emotional intensity, not just satisfaction. A touchpoint can score well on CSAT and still be emotionally flat — which means it has no chance of becoming a positive peak. Look for where intensity (positive or negative) actually spikes.
3. Identify your worst unmanaged peak. In most journeys, the most intense moment is a problem — a complaint, an outage, a failed payment — left to default scripts and tired agents. This is the single highest-leverage fix available, because an unmanaged negative peak defines the entire journey's memory regardless of what came before it.
4. Design the ending deliberately, every time. The last interaction in any journey or sub-journey — a call, a stay, a return process — should never be left to whatever happens to occur last. Close on resolution, on a small human gesture, or on clarity about what happens next; never on ambiguity or friction.
5. Build at least one positive peak on purpose. This does not require a grand gesture. A well-timed upgrade, a personalised note, or a moment of unexpected ease can become the peak that anchors the entire remembered story, provided it is genuinely surprising rather than scripted into meaninglessness.
6. Re-measure at the moment of recall, not just the moment of use. Pulse surveys fired mid-journey capture the experiencing self. A survey fired after the journey has concluded, once memory has consolidated, captures the self that actually decides whether to return.

This sequence is also where behavioural economics earns its keep in a CX function — not as an academic decoration, but as the lens that tells you which four or five touchpoints in a forty-touchpoint journey are actually worth the budget.

Why do endings carry so much extra weight?

Part of the answer is duration neglect, already established. But endings do additional work because of a second, related mechanism: the goal-gradient effect, the well-documented tendency for motivation and emotional investment to intensify as a goal draws closer. A customer nearing the end of a long support call, a multi-step checkout, or a loyalty-tier threshold is more emotionally activated than they were at the midpoint — which means friction or delight near the finish line lands harder than the same friction or delight would have landed earlier. The ending isn't just remembered more; it's also felt more intensely as it happens, which compounds the peak-end effect rather than existing separately from it.

This is why a botched return policy at the very last step of an otherwise smooth retail journey does disproportionate damage, and why a graceful final gesture — a handwritten note, a fee waived without being asked, a sincere closing line from an agent — can rescue a journey that went wrong somewhere in the middle. The data backs the intuition: the Nielsen Norman Group has long applied the peak-end rule to digital product design on exactly this basis — closing screens, confirmation moments, and final error states deserve design attention disproportionate to the time users spend on them.

What's the contrarian case most CX programmes are missing?

Here is the uncomfortable argument most maturity models avoid: trying to improve every touchpoint equally is not just inefficient — it is a worse strategy than concentrating resources on a small number of peaks and endings. The prevailing CX orthodoxy treats the journey as a long chain where every weak link needs reinforcing. The peak-end rule says otherwise: a flawless middle with a weak ending loses to an uneven middle with a strong ending, every time memory is doing the judging rather than a stopwatch.

Call this the peak-end allocation problem. Most CX budgets are still spread like insurance — a little improvement everywhere, proportional to volume or visibility. A peak-end-literate budget is spread like a portfolio — concentrated, disproportionate investment in the two or three moments per journey that will actually be remembered, with everything else held to "functional and frictionless" rather than "exceptional." This is not an argument for neglecting the middle; sludge and avoidable friction anywhere in a journey still erode trust and should be designed out. It is an argument against spending delight budget evenly when memory doesn't weigh it evenly. Leaders who run the numbers on this through a tool like the CX ROI Calculator typically find that re-allocating investment toward peak and closing moments outperforms flat, journey-wide spend of the same size.

This reframes a familiar CX debate. The question is rarely "should we invest in experience?" It is "where, specifically, does a unit of investment change the story a customer tells themselves afterwards?" The peak-end rule is the most direct answer behavioural science has produced to that question.

What does this look like in practice?

Consider two journeys with identical average quality. In the first, every stage is adequate, but the final interaction — a refund confirmation, a check-out, a discharge conversation — is handled by whoever is available, with no attention to tone or closure. In the second, the middle has one visible flaw, but the team has deliberately engineered a strong closing moment: a clear explanation, a small unprompted gesture, a genuinely warm sign-off. Both will likely score similarly on a mid-journey pulse survey. Only the second will generate advocacy, because only the second ends on a note the remembering self chooses to keep.

Signature moments — deliberately designed rituals at key points in the journey — are one of the more reliable ways to engineer a positive peak without relying on frontline improvisation. Renascence's work on customer rituals and ceremonies exists precisely because consistent, repeatable peak moments beat occasional, accidental ones. A peak you can only hope happens is not a strategy; it's luck wearing a CX label.

The same logic explains why brands with famously strong journeys often tolerate imperfections elsewhere without losing loyalty — their end-to-end journeys are engineered around a small number of unmissable peaks and clean endings, not around flawless consistency at every single step. Consistency matters, but it is not the same lever as memorability, and conflating the two is one of the more expensive mistakes a CX function can make.

Where this leaves CX leaders

The peak-end rule is not a reason to abandon rigorous, journey-wide measurement — it is a reason to stop treating the average as the thing that matters most. Map the full journey, find the moments of real emotional intensity, and never again leave an ending to chance. Customers will forgive an imperfect middle. They are far less forgiving of a story that ends badly, and far more loyal to a brand that knew exactly how to close the scene.

The organisations that will win the next decade of loyalty are not the ones chasing uniformly higher satisfaction scores across every touchpoint. They are the ones treating the ending of every journey — the call, the stay, the return, the renewal — as the single most valuable five minutes they have with a customer, and designing it accordingly. If you want to know where to start, a proper customer experience audit will tell you which peaks and endings in your journeys are currently being left to chance, and what it would take to make them deliberate instead.