Peak-End Rule: How People Remember an Experience

The peak-end rule says people remember an experience by its most intense moment and its end, not its length. The evidence, its limits, and how to apply it.

info Quick Definition
The peak-end rule says people judge and remember an experience mainly by how they felt at its most intense moment and at its end, not by the average of every moment or by its total length. It is a memory bias, and in design it helps decide which moments of a flow deserve the most care.

Where does it come from?

The rule comes from work by Daniel Kahneman and colleagues in the 1990s and 2000s. The most cited reference is a 1993 paper with a cold-water experiment. In one version, participants kept a hand in water at 14 °C for 60 seconds. In another, after those same 60 seconds, the water was warmed to 15 °C for 30 more seconds. When given a choice, subjects chose to repeat the longer trial, because it ended on a less uncomfortable stretch (Wikipedia: Peak-end rule).

An earlier study, by Redelmeier and Kahneman (1996), asked patients about their memory of a colonoscopy: they judged the discomfort mainly by the worst moment and the final one, regardless of how long it lasted. The best-known study is from 2003, with 682 patients. One group had a longer procedure, but with a final stretch of lower discomfort. According to the Coglode summary, that group reported 10% less pain and a 10% increase in attending the follow-up procedure (Coglode: Peak-End Rule).

What does it mean for design?

It means your product’s memory is not the average of all its screens. It is mostly the peak of intensity and the final moment. A five-step flow can feel good if the tensest step is resolved clearly and the ending confirms that everything worked, even if the rest is ordinary.

That is why, when you review a flow, it helps to read it the way the person will remember it: what was the worst part, and how did it end? That reading often shows different priorities than a list of completed tasks does.

This matters in onboarding, in purchases, and in any flow with a friction point. If someone reaches the end frustrated, that frustration becomes the main memory, even when the other screens were good.

What does the evidence say, and what does it not?

The rule is not a law of nature, and its limits need careful reading:

  • A 2008 study (Kemp, Burt and Furneaux) found some support, but called it “not an outstandingly good predictor” of the remembered value of an experience, according to the Wikipedia summary.
  • Miron-Shatz (2009) found that retrospective evaluations of day-long experiences did not always follow the rule. That casts doubt on applying it directly to, for example, hotel stays.
  • Duration matters little, but not nothing. The effect is described as “extremely slight.”

UX Tigers makes a practical point: the rule describes the memory of an episode, not immunity to repetition. A daily-use product accumulates memories. A final confetti screen will not offset a rage-inducing failure in the middle of the journey, and celebrating every trivial click flattens the peaks: UX Tigers sums up the advice as “ration the fireworks.” So the rule is a heuristic for prioritizing, not an argument for hiding problems.

How do you apply it without manipulating users?

UX Tigers proposes design guidelines you can use directly, including:

  • Repair the most negative moment. If the peak is bad, give the person a clear way out.
  • Never end on an error. The last message should confirm the result.
  • Schedule unavoidable unpleasantness early. UX Tigers names identity checks and paperwork as examples, which belong before the close.

Coglode suggests mapping the customer journey and treating negative experiences as a chance to re-establish a positive peak or end. But be careful: a manufactured positive ending that does not match a real outcome is manipulation. What you design is the clarity of the close, not the illusion of success.

Illustrative example

Illustrative example: a digital banking app has a seven-step account opening flow. The tensest step is identity verification, which is slow and sometimes fails. In the previous version, verification came last, and many users abandoned right there. The team moves verification to step two, adds a message that explains how long it takes and a fallback if it fails, and ends the flow with a screen that summarizes what comes next: active account, first deposit, and where to see the balance. The flow is not shorter, but the peak is no longer a blocker and the ending leaves no doubt.

If you are designing the start of a product, read onboarding patterns. For the states that appear before there is any content, see empty states. And if the difficult moment is an error, the article on error prevention and recovery explains how to repair without punishing the user.

Common mistakes

  • Putting the best moment at the start and leaving a flat ending, so everything afterward feels like a decline.
  • Finishing a flow on an error or an ambiguous screen.
  • Using the rule to justify an artificial peak that does not match a real outcome.
  • Believing that a pretty final screen compensates for problems in the middle of the journey.
  • Treating the rule as a law and no longer measuring the whole experience.

What you can do tomorrow

  1. Walk through your main flow and mark the tensest moment and the last step.
  2. Ask whether the ending confirms the real result clearly. If not, rewrite it.
  3. Move one uncomfortable but necessary task, such as a verification, to an early stage of the flow, and measure whether abandonment changes.

Sources

Frequently asked questions

What is the peak-end rule?
It is a memory bias in which people judge an experience mainly by how they felt at its most intense point and at its end, not by the average of every moment or by its total duration.
Who proposed the peak-end rule?
Daniel Kahneman and colleagues studied it in the 1990s. A 1993 paper is the usual reference, with its cold-water experiment showing that duration has little influence on memory.
Does the peak-end rule mean duration does not matter?
Almost. Studies found the effect of duration on retrospective evaluation is very small, a pattern called duration neglect. Even so, the rule has clear limits, so treat it as a heuristic and not as a law.
How do you apply the peak-end rule in onboarding?
Design a clear, positive ending for the flow, avoid finishing on an error, and place unavoidable uncomfortable steps before the close instead of leaving them for the end.