Peak-End Rule
Date: 2026-08-17
Memory of an experience is dominated by its most intense moment and its ending, not by its average or its duration. It means the last thing that happens disproportionately determines whether someone comes back — and for a retailer, the ending is usually delivery, not checkout.
The peak-end rule is the finding that people judge a past experience largely by how it felt at its most intense point and at its end, rather than by the sum or average of the whole.
Duration neglect is the companion finding: the length of an experience has surprisingly little effect on how it’s remembered.
The consequence
| Experience A | Experience B | |
|---|---|---|
| What happens | Mild discomfort for 8 minutes | The same, then a short period of less discomfort before ending |
| Objectively | Better — shorter, less total discomfort | Worse |
| Remembered as | Worse | Better |
Adding a better ending improves the memory even when it makes the experience objectively longer and worse in total. That’s counterintuitive enough to be worth stating twice.
This is Redelmeier and Kahneman’s colonoscopy study, and the design is worth knowing precisely: one group had a standard procedure, the other had the same procedure with the scope tip left in place for three extra minutes — no inflation, no suction, so mildly uncomfortable rather than painful. The longer group rated the whole experience as less unpleasant, and total duration barely predicted the remembered rating at all. That last part is duration neglect, measured.
Where the peak and the end actually are
For a retailer, both are usually after the part teams optimise.
THE JOURNEY
browse → choose → checkout → WAIT →
delivery → unbox → use → reorder
↑ ↑
most optimised the actual
peak and end
Checkout is the middle of the experience, not the end. The remembered ending is the parcel arriving — or failing to — and the peak is frequently the moment of anxiety during an uninformative delivery window.
LIKELY PEAKS (positive) LIKELY PEAKS (negative)
finding exactly the right "out of stock" at
thing checkout
unexpected extra in the an unexpected cost
parcel at step 4
a support reply that a delivery window
solved it fast that passed
a refund that took
three weeks
Designing the ending
The cheapest interventions in the whole journey sit here:
- The confirmation page and email. Reassurance, clear expectations, what happens next
- Delivery communication. The wait is the longest part of the experience and usually the least designed — information removes the anxiety even when it can’t speed anything up
- Unboxing. The physical ending, and disproportionately memorable
- The follow-up. Timed to arrive after they’ve used it, not after it shipped — Lifecycle Messaging
- Returns and refunds. The ending for a dissatisfied customer, and where a lost sale becomes a lost customer — Return Rate and Reverse Logistics
Refund speed is the highest-leverage ending nobody owns. It’s a finance and operations decision that determines whether someone shops with you again, and it’s invisible in every conversion metric — Service Blueprints.
Recovery is worth more than prevention sometimes
The related and well-supported finding: a problem resolved well can leave a stronger impression than no problem at all.
That is not licence to create problems. It does mean:
- Resource recovery properly. The support interaction after a failure is a peak, and it’s the one you fully control
- Fail visibly and fix fast rather than failing quietly
- Empower the resolution. A refund granted immediately by the first person contacted beats the same refund after three escalations
Where it’s oversold
- It doesn’t excuse a bad middle. A great ending after a genuinely broken checkout does not produce a happy customer
- It applies to remembered experience, which drives repeat purchase and word of mouth — not to the current conversion. Both matter and they’re different metrics
- The original research is on discrete, bounded episodes. A shopping journey spread over weeks with multiple touchpoints is a looser fit than a single continuous experience
- It’s not a licence for a manufactured “delight” moment. A gimmick at the end that solves nothing is noticed as a gimmick
Measuring it
The gap this predicts is measurable:
CONVERSION high — the checkout works
REPEAT RATE low — the experience after
it doesn't
That combination points at the post-purchase stretch, and it’s diagnosable with delivery-time data, support contact reasons and review text rather than with more funnel analysis — Repeat Purchase Rate, Voice of Customer Data.