The Peak-End Rule: How Customers Actually Remember Your Experience

The Peak-End Rule, in One Sentence

Customers do not remember an experience as the average of every moment in it. They remember it almost entirely by two moments: the most emotionally intense point (the peak) and the final point (the end). Everything in between — including how long the experience lasted — is heavily discounted by memory. If you want loyalty, referrals, and repurchase, you must deliberately engineer the peak and the end, then research whether you got them right.

This is the single most actionable finding in experience design, and most teams ignore it because their dashboards measure averages while their customers' brains measure peaks and endings.

"We do not choose between experiences, we choose between memories of experiences." — Daniel Kahneman, Nobel laureate, Thinking, Fast and Slow

Key Takeaways

• Customers judge experiences by the peak (most intense moment) and the end, not the average — and they largely ignore how long it lasted (duration neglect).
• The effect is well established: Kahneman's colonoscopy study and a 2022 meta-analysis of 174 effect sizes both confirm it.
• Emotion predicts loyalty better than effort, so a single great or terrible moment can define the whole relationship.
• The practical playbook: map the emotional curve, build one intentional peak, strengthen every ending, and neutralize negative peaks.
• You cannot find peaks in an average — AI-moderated interviews surface them by probing the emotional story behind each moment.

---

The Science Behind the Peak-End Rule

The peak-end rule was established by Nobel Prize-winning psychologist Daniel Kahneman and colleagues. In a landmark 1996 study, Kahneman and Donald Redelmeier tracked patients undergoing colonoscopies, recording their real-time discomfort every 60 seconds and then asking them to evaluate the experience afterward. The counterintuitive result: patients whose procedures were longer but ended on a milder note remembered the experience as less painful than patients whose procedures were shorter but ended at a moment of high pain.

The total amount of suffering — what researchers call the "area under the curve" — barely predicted the remembered experience. The peak and the end predicted nearly everything. In the original studies, averaging the peak and end correlated with global memory in the 0.7 to 0.9 range (Wikipedia: Peak–end rule).

This is not a one-off finding. A 2022 meta-analysis published in Organizational Behavior and Human Decision Processes synthesized 174 effect sizes and found strong, robust support for the peak-end rule across domains (ScienceDirect).

Two related biases ride alongside it:

• Duration neglect: the length of an experience has surprisingly little effect on how it is remembered. A 20-minute wait that ends well can beat a 10-minute wait that ends badly.
• The peak can be positive or negative. A single catastrophic moment can define an otherwise good experience — which is why one rude support interaction can erase months of smooth service.

---

Why This Matters for Customer Experience and Revenue

Memory drives behavior. Customers re-purchase, renew, and refer based on what they remember, not what objectively happened. And memory is emotional. Qualtrics research across 350+ brands found that emotion predicts customer loyalty better than effort or effectiveness across every metric measured — trust, NPS, and repurchase intent — a result that has held consistently across 14 years of study (Qualtrics). Emotionally connected customers are 5.7x more likely to trust a brand: 80% trust the brand when they give high emotion ratings, versus just 14% at low emotion ratings.

The revenue stakes are real. Forrester has found that 41% of CX-focused companies saw at least 10% revenue growth in their most recent fiscal year, and that improving CX by a single point can add up to $1 billion in revenue for a large company (Forrester). Yet customer emotions are deteriorating — happiness, excitement, and relief have declined by up to 9 percentage points since 2017 — even as companies pour money into functional improvements. Translation: teams are optimizing the flat middle of the experience curve and neglecting the two moments that memory actually keeps.

---

How to Apply the Peak-End Rule

1. Map the emotional curve of your customer journey

Most journey maps plot steps. Peak-end mapping plots feelings. For each stage — onboarding, first value, a support contact, renewal — ask: how intense is the emotion here, and is it positive or negative? You are hunting for the highest peak and, critically, the true ending of each meaningful episode.

2. Engineer an intentional peak

Add one moment of disproportionate delight. Examples: a surprisingly personal onboarding message, an unexpected upgrade, a "you did it" celebration when a user hits their aha moment. The peak does not have to be expensive — it has to be emotionally salient.

3. Fix the ending — every ending

Endings are everywhere: the end of onboarding, the end of a support ticket, the end of a billing cycle, the last screen before logout. A weak or anxious ending poisons the memory of an otherwise good experience. Design endings to close on competence, gratitude, or progress.

4. Neutralize negative peaks

One furious moment can define the whole relationship. Find your worst moments — failed payments, error states, long waits, churn flows — and either remove the pain or add a recovery moment immediately afterward (a "better end").

---

The Modern Approach: Finding Peaks and Ends with AI Interviews

Here is the hard part: you cannot find emotional peaks in a satisfaction average. A CSAT score of 4.2 tells you nothing about which moment spiked and why. Traditionally, surfacing peaks meant recruiting participants, scheduling moderated interviews, manually transcribing recordings, and hand-coding emotional moments — weeks of work for a handful of conversations.

Koji collapses that into hours. As an AI-native research platform, Koji runs AI-moderated interviews — over voice or text — that probe the emotional arc of a customer's experience automatically. When a customer mentions a frustrating or delightful moment, Koji's AI asks the follow-up a static survey never would: "Tell me more about that moment — what were you feeling right then?" This adaptive probing is exactly how you locate peaks and endings, and it happens across hundreds of conversations in parallel.

Koji then runs automatic thematic analysis across every transcript, clustering the emotional high and low points so you can see which peaks and endings recur — not just one anecdote, but the pattern. Its real-time reporting turns that into a shareable narrative in minutes.

Where traditional survey tools like SurveyMonkey force you to guess in advance which moments to ask about, an AI-native platform like Koji discovers them from the conversation. And Koji's six structured question types let you quantify the emotion you uncover:

• scale questions to measure intensity of a remembered moment (the peak)
• single_choice to identify which stage felt most positive or negative
• open_ended to capture the story behind the emotion
• ranking to order moments by memorability
• multiple_choice to tag which touchpoints drove feeling
• yes_no to confirm whether the ending left them satisfied

Teams using AI-assisted research report dramatically faster time-to-insight — turning a multi-week peak-end study into a same-week deliverable, without a PhD in behavioral science. See the structured questions guide for how to combine these into a single conversation.

---

Common Mistakes

• Optimizing the average, not the peak. Raising every step by 5% does less than fixing the one moment customers actually remember.
• Ignoring endings. Teams obsess over onboarding (a beginning) and neglect renewal, offboarding, and ticket closure (all endings).
• Measuring only at the end. If you only survey after the experience, you capture the memory but not the moment. Pair retrospective interviews with in-the-moment signals.
• Treating a negative peak as "just one bad review." A single high-intensity failure is statistically the moment most likely to define the relationship.

---

Real-World Examples of the Peak-End Rule

• IKEA's exit. A long, tiring maze of a store ends with a cheap hot dog or cinnamon roll at the checkout — a deliberate positive ending that reframes the memory of the whole trip.
• Disney parks. Long queues (a sustained low) are punctuated with character interactions and themed details (peaks), and visits end on fireworks and a curated gift-shop moment (a strong end), so guests remember magic, not waiting.
• Hotels. A frictionless, warm checkout — the literal end of the stay — disproportionately shapes the review, which is why a single rude front-desk moment at departure can sink an otherwise excellent stay.
• SaaS onboarding. Ending setup with a visible "you're all set — here's your first win" celebration creates a positive peak-and-end at exactly the moment that predicts Day 1 retention.
• Support tickets. The resolution message is the ending of that episode. Closing with genuine confirmation and a thank-you, rather than an abrupt "ticket closed," is one of the cheapest CX wins available.

The pattern across all of them: spend disproportionately on the peak and the end, and stop trying to make the flat middle marginally better.

Related Resources

• Customer Journey Mapping: The Complete Guide — plot the stages where peaks and endings live
• The Aha Moment: Researching First Value — engineer a positive peak early
• Empathy Interview Guide — surface the emotion behind each moment
• Voice of Customer Research Program — operationalize ongoing emotional listening
• Customer Experience Benchmarking — track whether your peaks and ends improve over time
• Structured Questions Guide — quantify emotional intensity with scale, ranking, and single-choice questions