---
title: "The Peak-End Rule: Why People Mostly Remember the Peak and the End"
url: "https://insideuxdesign.com/en/blog/the-peak-end-rule-why-people-mostly-remember-the-peak-and-the-end"
language: "en"
description: "People mostly remember the peak and the end of an experience. Learn how to use the peak-end rule to design the moments that stick in users’ memory."
published: "2024-05-29T09:00:00+02:00"
updated: "2026-10-15T00:34:45+02:00"
---

# The Peak-End Rule: Why People Mostly Remember the Peak and the End

People don’t judge an experience by its average, but mostly by its most intense moment and by how it ends. What the research says and how to design the moments that stick.

Picture someone reporting a parking dent through their car insurance app. The first steps go smoothly: pick the policy, enter the date, briefly describe what happened. Then it’s time to upload photos. The upload fails twice, and the error message just says “Something went wrong.” On the third try it works. At the end, a plain screen appears: “Your report has been submitted.” No claim number, no word on what happens next.

Objectively, the process wasn’t that bad. Eight out of ten steps went without a hitch, and the whole report took less than ten minutes. Still, that evening the person says: “That app is awful.” What stuck was the photo upload and the unsatisfying ending. Everything else has faded from memory.

This is exactly the pattern the peak-end rule describes. It explains why people don’t judge an experience by its average, but mostly by its most intense moment and by how it ends. For UX design, that’s a useful insight: it shows you which moments of a product deserve the most care. The insurance app serves as a running example throughout this article. It’s made up, but typical.

## What the peak-end rule says

The peak-end rule goes back to psychologist Barbara Fredrickson and psychologist Daniel Kahneman, who later won the Nobel Memorial Prize in Economic Sciences. It describes a cognitive bias, a systematic error in thinking: when people look back on how pleasant or unpleasant an experience was, they rely mainly on two moments. The peak, meaning the most intense moment, positive or negative. And the end.

Fredrickson and Kahneman described this as a “snapshot model.” Memory doesn’t store an experience as a complete film, but as a few representative frames. Later, the overall judgment is assembled from those frames.

Closely related is a second effect called duration neglect. How long an experience lasted matters surprisingly little for how it’s remembered. A twenty-minute process with a good ending can be remembered more fondly than a five-minute one that ends in frustration.

In his book “Thinking, Fast and Slow,” Kahneman later framed this difference with two terms: the experiencing self, which lives through each moment, and the remembering self, which forms a judgment afterwards. What matters for design is this: whether someone opens an app again, recommends it or cancels is usually decided by the remembering self.

## The studies behind the rule

### Cold water and a slightly better ending

The best-known experiment on the peak-end rule was published in 1993. Daniel Kahneman, Barbara Fredrickson, Charles Schreiber and Donald Redelmeier asked 32 students to hold one hand in water at 14 °C (about 57 °F). That’s unpleasant but harmless. Each person went through two trials:

- **Short trial:** 60 seconds in water at 14 °C.
- **Long trial:** the same 60 seconds at 14 °C, followed by another 30 seconds while the temperature slowly rose to about 15 °C.

So the long trial contained everything unpleasant about the short one, plus a bit more. Yet when asked which trial they would repeat, 22 of the 32 participants, about 69 percent, chose the long one. The slightly milder ending had improved the memory of the whole experience. The title of the study sums it up: “When more pain is preferred to less: Adding a better end.”

### Memories of real medical procedures

Donald Redelmeier and Daniel Kahneman showed in 1996 that the effect isn’t limited to the lab. They studied 154 people undergoing a colonoscopy and 133 people undergoing lithotripsy, a procedure to break up kidney stones. During the procedures, patients continuously reported how much pain they felt. Afterwards, they rated the procedure as a whole.

The result: retrospective ratings were closely tied to the worst pain and to the pain in the last three minutes. How long the procedure lasted hardly mattered. Long procedures were not remembered as particularly bad.

A randomized trial with 682 people followed in 2003. For half of them, the colonoscopy was extended by a short, less unpleasant interval at the end. This group rated the procedure as less unpleasant overall, and over the following years slightly more of them came back for their next screening. A better ending had influenced not only memory but also later decisions.

### How robust is the rule?

A 2022 meta-analysis, a study that combines the results of many individual studies, by Balca Alaybek and colleagues pooled 174 effect sizes. It found a strong link between the peak and end of an experience and how it was rated afterwards. The duration of the experience had practically no effect. The beginning of an experience and its low point carried less weight than the peak and the end.

Still, it isn’t quite that simple. The same meta-analysis shows that the average of all moments predicts ratings about as well as the peak and the end. For longer, multi-part experiences the rule is less clear-cut. In 2009, Talya Miron-Shatz found no clear support for it when people looked back on entire days. In 2008, Simon Kemp and his team did find support when people rated longer events from their own lives, but the moment people found most memorable predicted their overall recollection better than the peak or the end.

In practice, that means the peak-end rule is not a formula for calculating memories. It is, however, well-supported evidence that not all moments count equally. It fits short, clearly defined experiences especially well. And that’s exactly what many digital products are made of: an order, a sign-up, an insurance claim, a support chat.

## What this means for digital products

Applying the rule to UX design raises three questions. Where are the peaks of an experience, positive and negative? What do people experience at the very end? And which moments in between can you deliberately keep calm, because they’ll barely be remembered anyway?

Perspective matters here. What a team considers the end of a process often isn’t the end for the people using it. In the insurance app, the technical process ends when the form is submitted. For the person reporting the damage, the matter is only settled once they know their report has arrived, what happens next and when they can expect the money. So the perceived ending comes later, perhaps with the confirmation email or the call from the claims handler.

The same goes for peaks. Negative peaks almost always occur where people get stuck: an error message with no way forward, a form that wipes all input, a wait without any feedback. Positive peaks occur where people reach a goal or something turns out easier than expected.

## Finding peaks and endings

Before you design, you need to know where the decisive moments are. Methods many teams already use work well for this.

**Journey map with an emotional curve:** A [journey map](https://insideuxdesign.com/en/glossar/journey-map) shows a person’s path through a product or service step by step. Many teams add a line showing how the person feels at each step. The highest and lowest points on that line are your candidates for peaks. The last point shows how the ending is doing. It’s important that the curve is based on observations and interviews, not on assumptions within the team.

**Usability tests with a look back:** Have participants complete a task and then ask them what stuck with them. The moments they mention first are often the peaks in the sense of the rule. Compare that with what you observed during the test. Sometimes people remember a brief irritation more strongly than a longer search they barely noticed themselves.

**Support requests and reviews:** What people write to support or mention in app store reviews is already a selection from their memory. If the same step keeps coming up in reviews, that’s a strong sign of a negative peak.

For the insurance app, a clear picture emerges: the photo upload is the negative peak, the confirmation screen is a weak ending, and the weeks until payout are a blind spot nobody in the app takes care of.

## Five ways to put the rule into practice

### 1. Soften negative peaks

The most effective lever is often not creating new positive moments, but taking the edge off the worst ones. A single frustrating step can define an otherwise good experience. So start by checking where people fail, and give those spots priority.

Errors are especially critical. Nielsen Norman Group recommends that, when errors or outages happen, you explain what happened, give a realistic resolution or time estimate and don’t blame the user. For the photo upload, that means:

- **Before:** “Something went wrong.” All selected photos are gone.
- **After:** “This photo is too large (12 MB). We’re resizing it automatically and uploading it again.” Photos that were already uploaded are kept, and a progress bar shows the status of each image.

Even better is preventing the error in the first place. If the app resizes photos itself before uploading, the negative peak disappears entirely.

### 2. Create positive peaks on purpose

Every product has moments where people reach a goal. You can amplify these moments so they’re remembered as positive peaks. Nielsen Norman Group cites Duolingo as an example, where a character celebrates correct answers and reinforces the sense of progress.

It doesn’t have to be fireworks. Small, well-crafted [microinteractions](https://insideuxdesign.com/en/glossar/microinteractions), short animations or feedback on a single action, are often enough. In the insurance app, this could be the moment all photos have uploaded successfully: a quick checkmark on each image and the line “All photos are in. That lets the claims team assess the damage right away.” This not only confirms success, it also explains why the step mattered.

A positive peak can also be something turning out surprisingly easy. If the app already knows the license plate and policy details and doesn’t ask for them again, people notice. Effort they’re spared is a small peak of its own.

### 3. Design the ending deliberately

The ending is the moment with the longest afterglow, and at the same time the one many projects pay the least attention to. Confirmation screens, wrap-up emails and sign-out screens are often built last, with little time and boilerplate copy.

A good ending answers three questions: Did it work? What happens now? Do I need to do anything else? Nielsen Norman Group describes TurboTax, a US tax software, as an example: it celebrates the completed tax return with an illustration instead of using that moment to sell more products.

For the insurance app, it looks like this:

- **Before:** “Your report has been submitted.” Nothing else.
- **After:** “Done. Your claim has been received (claim number 2026-48213). A claims handler will get in touch within two business days. You don’t need to do anything until then. You can check the status anytime under ‘My claims’.” Plus a confirmation email with the same details.

Keep the perceived ending in mind. If the matter is only truly settled weeks later, the notifications up to that point are part of the ending too. A short message like “Your claim has been settled and the payment is on its way” can noticeably improve how the whole process is remembered.

### 4. Don’t interrupt at the end

Because the ending carries so much weight, it’s also especially fragile. Many products use exactly that last moment for things people don’t want right then: a rating request, an upsell, a ten-question survey or an exit pop-up offering a discount to make you stay. Nielsen Norman Group explicitly advises against these “please don’t go” pop-ups, because they annoy people and can leave a negative last impression.

That doesn’t mean an ending can never offer a next step. A subtle pointer to a meaningful follow-up action is fine if it fits the situation. But someone who has just reported damage to their car doesn’t need a pitch for dental insurance. For every element on a confirmation screen, ask yourself: does this help the person, or only us?

Cancellation flows are particularly delicate. When someone ends a subscription, the farewell is their last impression. If that path is deliberately convoluted, that’s exactly what sticks. Such patterns are [dark patterns](https://insideuxdesign.com/en/blog/spotting-and-avoiding-dark-patterns), and they damage not only memories but potentially your reputation and legal standing as well. A clear, respectful goodbye, on the other hand, leaves the door open for a return.

### 5. Design waiting and progress

Waiting times are classic candidates for negative peaks, especially when it’s unclear how long they’ll last. A 2007 study by Chris Harrison and his team showed that people perceive the duration of a process differently depending on how the progress bar behaves. Nielsen Norman Group sums this up by noting that progress indicators that speed up toward the end are perceived as faster than ones moving at a constant pace.

That doesn’t mean you should fake progress. It does mean that a process that stalls at the end is remembered especially negatively. If a step takes longer, such as checking the photos, it belongs not at the end but ideally in the middle or in the background. In the insurance app, photos can upload while the person is still typing the description. The final submit then takes a second, and the process ends without waiting.

## Limits and responsibility

The peak-end rule invites a misunderstanding: if only the peak and the end count, you could neglect everything else. That would be a mistake. First, research shows that the average plays a role too. Second, people live through every minute, even if they don’t remember it later. A sluggish process costs time, patience and possibly drop-offs before there’s even an ending. The rule helps you set priorities. It doesn’t replace good usability along the entire path.

The rule also carries an ethical tension. In the 2003 colonoscopy study, the procedure was extended so that it would be remembered more favorably. Applied to design, this means you can improve memories without improving the experience itself. A friendly ending can gloss over poor service, at least for a while. That’s not the goal.

A good yardstick is whether the nice ending is honest. A confirmation screen that clearly says what happens next is honest. A shower of confetti after an order whose delivery time was deliberately hidden is not. The peak-end rule works best when it helps you genuinely improve the most important moments, not just make them look better.

For a compact overview of the peak-end rule and other psychological principles, see the resource [Laws of UX](https://insideuxdesign.com/en/ressourcen/laws-of-ux).

## Conclusion

People don’t remember every moment of an experience. They mostly remember the most intense moment and the end. The studies by Kahneman, Fredrickson and Redelmeier show this vividly, and a large meta-analysis confirms the effect. For longer, multi-part experiences it’s less clear-cut, and the average counts as well.

For design, this suggests a clear order. First, find the negative peaks and soften them. Then design the ending so people know everything worked and what happens next. Create positive peaks where people reach a goal. And keep the final moment free of anything that only benefits you.

For the insurance app in our example, this doesn’t mean a complete rebuild. A robust photo upload, a confirmation with a claim number and next steps, and a message when the money is on its way are often enough for a different sentence that evening: “That was actually pretty easy.”

## Checklist

- Key flows are captured in a journey map with an emotional curve, based on observations.
- The negative peaks are known and get priority for improvements.
- Error messages explain what happened and show a way to fix it.
- Input isn’t lost when errors occur.
- Moments where people reach a goal are visibly confirmed.
- Every confirmation screen answers: Did it work? What happens now? Do I need to do anything else?
- The perceived ending is taken into account, even if it comes after the technical one.
- No pop-ups, surveys or unrelated offers interrupt the ending.
- Long waits don’t sit at the end of a process.
- Canceling and signing out are as clearly designed as signing up.
- Every positive ending is honest and doesn’t hide problems.

## Sources

- [Daniel Kahneman, Barbara L. Fredrickson, Charles A. Schreiber and Donald A. Redelmeier (1993): When more pain is preferred to less: Adding a better end. Psychological Science, 4(6), 401 to 405](https://doi.org/10.1111/j.1467-9280.1993.tb00589.x)
- [Barbara L. Fredrickson and Daniel Kahneman (1993): Duration neglect in retrospective evaluations of affective episodes. Journal of Personality and Social Psychology, 65(1), 45 to 55](https://doi.org/10.1037/0022-3514.65.1.45)
- [Donald A. Redelmeier and Daniel Kahneman (1996): Patients’ memories of painful medical treatments: real-time and retrospective evaluations of two minimally invasive procedures. Pain, 66(1), 3 to 8](https://doi.org/10.1016/0304-3959(96)02994-6)
- [Donald A. Redelmeier, Joel Katz and Daniel Kahneman (2003): Memories of colonoscopy: a randomized trial. Pain, 104, 187 to 194](https://pubmed.ncbi.nlm.nih.gov/12855328/)
- [Balca Alaybek et al. (2022): All’s well that ends (and peaks) well? A meta-analysis of the peak-end rule and duration neglect. Organizational Behavior and Human Decision Processes, 170, 104149](https://doi.org/10.1016/j.obhdp.2022.104149)
- [Talya Miron-Shatz (2009): Evaluating multiepisode events: Boundary conditions for the peak-end rule. Emotion, 9(2), 206 to 213](https://doi.org/10.1037/a0015295)
- [Simon Kemp, Christopher D. B. Burt and Laura Furneaux (2008): A test of the peak-end rule with extended autobiographical events. Memory & Cognition, 36(1), 132 to 138](https://doi.org/10.3758/MC.36.1.132)
- [Chris Harrison, Brian Amento, Stacey Kuznetsov and Robert Bell (2007): Rethinking the progress bar. Proceedings of UIST ’07, 115 to 118](https://www.chrisharrison.net/index.php/Research/ProgressBars)
- [Lexie Kane, Nielsen Norman Group (2018): The Peak-End Rule: How Impressions Become Memories](https://www.nngroup.com/articles/peak-end-rule/)
- [Jon Yablonski, Laws of UX: Peak-End Rule](https://lawsofux.com/peak-end-rule/)
- Daniel Kahneman (2011): Thinking, Fast and Slow. Farrar, Straus and Giroux