Strategy ⚡ Low

Peak-End Rule

People judge an experience by its most intense moment and how it ended — not by averaging every step. Every option you add after the peak is cognitive debt the ending must repay.

Onboarding FlowsCheckout & PaymentEmpty StatesError RecoveryOffboardingConfirmations

01 — TL;DR

Two sentences.

People do not average their way through an experience. They remember the moment that hit hardest — positive or negative — and the feeling they had when it ended. Everything in between has far less influence on whether they rate it good, come back, or tell a friend.

Daniel Kahneman and Barbara Fredrickson established this through the cold pressor experiments in the early 1990s, where subjects preferred longer pain if the final moments were slightly less intense. For designers, the implication is direct: your resources are finite, so you have to be deliberate about which moments get elevated and what users feel when they leave.

02 — When to Use

Apply this when…

→Designing a multi-step onboarding sequence where you need to decide which screen gets the animation, the confetti, or the celebratory copy

→Engineering a checkout confirmation page — the end of a high-stakes transaction

→Writing the final screen of an offboarding flow to reduce churn regret and preserve brand goodwill

→Deciding which error state to invest in — prioritise the one users hit most often or most painfully

→Shaping the emotional arc of a feature launch, where the first and last impressions anchor perception

When NOT to apply it

Skip it for single-screen interactions with no temporal arc (a tooltip, a toggle, a snackbar), purely informational pages with no user goal to resolve, dashboards refreshed constantly with no discrete end moment, or experiences so short that peak and end are the same moment.

03 — How It Works

The mechanism

The brain does not store experiences as a continuous recording. It samples moments and compresses them into a narrative, leaning heavily on the most emotionally charged point and the final state. The average duration and the texture of the middle are largely discarded.

Duration neglect and affective sampling

Kahneman's research showed that retrospective judgements of an experience are almost entirely predicted by two data points — peak valence and end valence. Duration has almost no effect. In one study, subjects rated a shorter but uniformly painful procedure as worse than a longer one that ended with reduced pain, even though the longer procedure involved more total pain.

What it means in practice for interfaces

Every flow has a natural peak — the moment of highest cognitive or emotional load. In a checkout flow, that is usually the payment screen. In onboarding, it is the moment a user completes their first meaningful action. The end is whichever screen they see last. Both deserve disproportionate design attention. The screens in between can be functionally competent without being emotionally engineered.

A negative peak does not automatically ruin an experience

Duolingo routinely ends a lesson with a streak celebration and a bold reward screen, even when the lesson itself included frustrating errors. The recovery and the ending override the pain of the middle. This is why error recovery design matters as much as error prevention: a well-handled recovery can actually become the peak that defines the memory positively.

How to measure it

Run a memory interview, not a satisfaction survey, immediately after a usability session. Ask users: "What moment stands out most from that experience?" and "How did you feel when you finished?" Map those answers against your flow. If the peak they recall is unintentional — a loading spinner, a confusing field — you have found where to invest next.

Memory vs. real-time experience

The Peak-End Rule describes memory, not real-time experience. A user can feel frustrated throughout a flow and still rate it highly if the ending is strong. This is not an excuse to let the middle be bad — high friction in the middle increases drop-off before users ever reach your engineered ending. Fix the middle for completion rates; engineer the peak and end for memory and return.

04 — Real Example

Duolingo's lesson completion screen and the memory it manufactures

Duolingo ends every lesson with an animated celebration screen: a large XP count, a streak flame, sound effects, and a prominent visual reward. This screen appears regardless of how many errors the user made during the lesson itself. It is a textbook engineered ending.

The deeper insight is that Duolingo has identified its end moment with precision and invested heavily in making it emotionally positive. The lesson content — the actual drilling of vocabulary — is often tedious. Duolingo does not try to make every screen delightful. It concentrates emotional investment at the end, where memory formation happens, and lets the middle be functionally adequate.

Duolingo · Lesson Completion

Engineered ending overrides a tedious middle

Engineered ending → memory of delight

05 — In the Wild

Test yourself & see real examples

No examples yet — be the first.

Spotted a product that nailed its ending — or buried it under a generic "You're all done" message? Submit a screenshot. Every approved example gets attributed to you.

✓ Reviewed before publishing✓ Your name on every example you submit✓ Violation or fix — both welcome

Seen Peak-End Rule violated in a real product? Help grow the evidence base.

06 — Common Mistakes

Where teams go wrong

✗Investing equally across every screen in a flow. Teams often treat all steps as equally important and spread design effort uniformly. This produces flows where every screen is adequate and none is memorable. The Peak-End Rule says memory is not democratic — weight your effort toward the peak and the end, and let the middle be functionally solid rather than emotionally polished.

✗Treating the confirmation page as a receipt, not a moment. Confirmation screens — order placed, account created, form submitted — are nearly always the end of a flow. Most teams build them as functional receipts: an order number, a timestamp, a back-to-home link. This is a missed opportunity. The confirmation screen is the last thing a user sees and the moment most likely to colour their memory of the entire transaction.

✗Letting a negative peak go unaddressed because it is rare. Teams often deprioritise error states because they affect a small percentage of users. But a painful, confusing error recovery is precisely the kind of peak that sticks. Users who hit a broken payment flow and recover badly will rate the entire checkout negatively — even if 19 of 20 steps were smooth.

✗Engineering a peak that feels manipulative rather than genuine. Confetti on every action, celebrations for mundane tasks, and forced enthusiasm at the end of a generic form feel hollow — and users notice. A manufactured peak that does not match the actual weight of the moment creates cognitive dissonance. Reserve strong emotional moments for genuinely significant user milestones.

07 — Variations & Related Principles

Connected ideas

The Peak-End Rule sits inside a broader cluster of principles about how users form judgements and memories over time. It is most useful when paired with principles about cognitive load (which affects how painful the middle feels) and emotional design (which gives you the toolkit to engineer peaks).

The most important pairing is Peak-End Rule with Cognitive Load Theory. The Peak-End Rule tells you where to invest emotional design effort — at the peak and the end. Cognitive Load Theory tells you what is making the middle painful enough to cause abandonment before users ever get there. Address both and you have a flow that is completable and memorable.

08 — 10-Min Exercise

Run it right now

⏱ 10 minutes · Solo · No prep

The Memory Audit

1. Open a flow in your product — onboarding, checkout, or a core task — and write down every screen in order. Label each one with a rough emotional weight: neutral, positive, or negative, from the user's perspective.

2. Circle the screen with the highest emotional charge (positive or negative). That is your current peak. Now circle the last screen a user sees before exiting. That is your current end. Ask: are either of these screens receiving disproportionate design investment right now?

3. If your peak is unintentional — a loading state, a form validation error, a confusing decision point — you have found your highest-leverage design problem. If your end screen is a generic confirmation with no emotional signal, that is your second target.

4. Write one specific change you could make to either the peak or the end screen this week. Not a redesign — a single change: a headline, an animation trigger, a piece of copy, a visual reward. Ship that change first.

10 minutes