In a randomized colonoscopy trial, a longer, gentler-ending procedure was remembered as less unpleasant. That is the peak-end rule: memory is built from the peak and the end, not the average. Here is how to design both on purpose.
In a randomized trial published in Pain in 2003, patients whose colonoscopy was deliberately extended with a few extra minutes of milder discomfort remembered the whole procedure as less unpleasant than patients who had a shorter, sharper one, and they were more willing to return. That is the peak-end rule: people judge an experience by its most intense moment and its final moment, not by its average or its length. For any small business, that means memory is built from two data points, not from every step you sweat over.
The finding is not a marketing anecdote. It came out of clinical research led by Donald Redelmeier, Joel Katz, and Daniel Kahneman, who later won the Nobel Prize in economics for related work on judgment.
In the 2003 trial, one group of patients had a standard colonoscopy. Another group had the scope left still at the end for a short stretch, adding time that was uncomfortable but noticeably gentler than the peak. That second group objectively endured more total discomfort.
Yet, as The Decision Lab documents, the patients with the longer procedure rated it as less bad and returned for follow-up care at higher rates. The ending rewrote the memory of everything that came before it.
I study behavioral psychology because results like this keep overturning the intuitive model of how people form opinions. We assume customers keep a running tally. They do not. They keep a highlight and a closing frame.
The mechanism underneath that result has a name of its own: duration neglect. How long an experience lasts has almost no effect on how it is remembered.
The cleanest demonstration is the 1993 cold-water study Kahneman ran, summarized in the peer-cited record of the research. Participants held a hand in painfully cold water for 60 seconds. In a second trial, they held it in the same cold water for 60 seconds, then kept it in for 30 more seconds as the temperature was raised slightly, still unpleasant but less so.
The second trial contained strictly more discomfort. When asked which one they would rather repeat, most people chose the longer trial, because its final stretch was milder and that softer ending is what they remembered.
For a business, this is oddly freeing. You do not have to make a long engagement feel short, and you cannot buy loyalty by shaving minutes off every interaction. What lands is the shape of the curve, not the area under it.
Kahneman framed the gap with two characters. The experiencing self lives every moment in real time. The remembering self writes the story afterward, and the remembering self is the one who decides whether to come back or refer you.
The remembering self is not a careful accountant. It works by snapshots. Kahneman and Barbara Fredrickson called this the snapshot model: we store a few emotionally vivid frames and reconstruct the whole from those.
This is close cousin to how cognitive load shapes decisions elsewhere. When a mind is overloaded it reaches for shortcuts, which is exactly why reducing cognitive load lifts conversions rather than adding more information. The remembering self is running the same economy: it cannot hold everything, so it keeps the peak and the end.
There is a second consequence that trips up well-meaning teams. Because the remembering self drives future behavior, satisfaction measured in the moment can diverge sharply from satisfaction reported later. A survey taken the day after delivery is a report from the remembering self, and it is anchored to whatever the peak and the end happened to be.
This is why two clients who received nearly identical service can rate you very differently. The one whose engagement trailed off into an awkward silence and a terse final bill remembers friction. The one whose work closed with a clean summary remembers competence. The middle was the same.
Once you accept that memory is compressed to two frames, the practical question changes. It stops being "how do we make everything a little better" and becomes "which single moment do we make unmistakably good, and how do we end."
A calm claim needs its boundary conditions, so it is worth asking how well this holds up. The honest answer is: well, with limits.
A 2024 review in the peer-reviewed literature, which examined 174 independent samples, found the peak-end rule robust across a wide range of conditions and better at predicting overall judgment than competing heuristics. It is one of the more replicated findings in behavioral science.
The same review is candid about where it wobbles. In some longer-timeframe settings, an average of daily experience predicted retrospective judgment better than the peak or the end. Personality traits and the emotional valence being measured also shift the effect.
So the peak-end rule is a strong default, not a law of physics. Treat it the way you would treat any cognitive bias shaping a decision: as a reliable tendency to design around, not a guarantee to exploit.
Most small businesses spread their effort evenly across the customer journey. The peak-end rule suggests concentrating it instead, and the first job is simply locating the two moments that matter.
The end is the easier one to find, and it is almost always neglected. It is the final invoice, the offboarding email, the last handoff, the moment a project closes. Teams pour energy into winning and delivering the work, then let the closing frame happen by accident.
The peak is a choice you should make on purpose. It might be the onboarding call where a client finally feels understood, the first time a dashboard shows them something they could not see before, or a delivery that arrives ahead of schedule. Pick one and invest there disproportionately.
This is the same principle behind mapping the handful of metrics that actually connect to revenue instead of tracking everything. Attention is finite. Spend it where memory is formed, not evenly across a curve nobody remembers.
The application literature reinforces this. The peak-end rule in experience design is often illustrated with a simple move: adding warmth, personality, or reassurance at a single confirmation moment, because that positive frame is what users store and recall. The same reasoning applies to a negative peak. A billing dispute or a delay is a moment of high intensity, and handling it with unusual care can convert a would-be low point into a memorable recovery.
If the end carries outsized weight, then offboarding deserves as much design as onboarding, and it rarely gets any. The final invoice is a good place to start, because it is the last thing many clients read.
An invoice that arrives with a short, warm note, a clear summary of what was delivered, and an easy path to the next step ends the relationship on a human frame. An invoice that arrives as a cold number ends it on friction, which is the snapshot the remembering self keeps.
The offboarding email is the other lever. Instead of silence when a project wraps, a brief message that recaps the outcome, hands over documentation cleanly, and leaves the door open turns the ending into a genuine peak. This matters most where trust is the thing that makes someone hire you again.
None of this requires a bigger budget. It requires deciding that the closing frame is not an afterthought. The lasting relationship a client keeps with a business is often held together at exactly these transition points, which is why our own managed operations methodology treats ongoing support and clean handoffs as first-class work rather than a wind-down.
There is an obvious temptation here, and it is worth naming plainly. The peak-end rule can be read as a recipe for gluing a pleasant ending onto a bad experience so people misremember it. That reading is both wrong and self-defeating.
The experiencing self still lives every moment, even when the remembering self forgets. A genuinely painful process leaves real costs behind, and repeated exposure erodes the trick. The design literature on the peak-end rule makes the same point: the goal is to shape real experiences, not to manufacture false memories of poor ones.
The defensible use is straightforward. Fix the experience so the middle is genuinely fine, then decide deliberately where the one peak should be and make the ending as considered as the beginning. Consistency across the middle matters less than intention at the peak and the close.
That reframe also lowers the pressure on small teams. You are not obligated to make every interaction extraordinary, which is a fast route to the kind of burnout that does not announce itself. You are obligated to get two moments right. I keep returning to this idea because so much of good operations is knowing where effort compounds, a theme that runs through much of how we write about building calm systems.
The peak-end rule is a psychological finding that people judge an experience mostly by its most intense moment (the peak) and its final moment (the end), rather than by the average of every moment or how long it lasted. It means memory is built from two data points, not a running total.
Duration neglect is the closely related finding that how long an experience lasts has very little effect on how it is remembered. In studies of both cold-water discomfort and medical procedures, a longer experience with a gentler ending was remembered more favorably than a shorter, sharper one.
It is one of the better-replicated findings in behavioral science. A 2024 review of 174 independent samples found it robust across many conditions, while noting boundary cases where an average of daily experience predicted judgment better. Treat it as a strong default rather than an absolute law.
Locate the two moments that carry the most weight: one peak worth investing in disproportionately (an onboarding call, a first result, an early delivery) and the ending (the final invoice, offboarding email, or handoff). Make the close as considered as the opening instead of letting it happen by accident.
It is only manipulative if you use it to mask a genuinely bad experience, which also fails over time because the experiencing self still bears the real costs. The honest use is to make the middle genuinely fine, then shape a real peak and a thoughtful ending rather than manufacturing false memories.
Give people a felt safety margin and many of them spend it. Risk compensation shows up in ABS brakes, MFA, and backup tools alike. The fix is to build so protection is invisible and cannot be traded away.
The whole industry optimizes to remove friction. But the case for friction is that a deliberate pause, a verification step, a rate limit, a cooling-off window, filters bots and bad actors, cuts regret and fraud, and protects the user.
Goodhart's Law says when a measure becomes a target it stops being a good measure. Here is why small teams accidentally reward the number instead of the outcome, and how to build metrics that resist gaming.
Work With Us
Kief Studio builds, protects, automates, and supports full-stack systems for businesses up to $50M ARR.
Newsletter
Strategy, psychology, AI adoption, and the patterns that actually compound. No spam, easy to leave.
Subscribe
