The cleanest demonstration of this is almost absurd. In a 1993 experiment, Daniel Kahneman and colleagues had people hold a hand in painfully cold water for 60 seconds. Then they did it again — same 60 seconds of cold, followed by 30 extra seconds during which the water was warmed slightly, to a temperature that was still unpleasant but less so.
By any rational measure, the second trial contained more total discomfort. It was strictly worse. Yet when participants were asked which one they'd rather repeat, most chose the longer trial. The gentler ending rewrote the memory of the whole thing.
The same pattern showed up in a far more consequential setting. In studies of patients undergoing genuinely uncomfortable medical procedures, how patients remembered the experience — and whether they were willing to come back — was predicted far better by the peak and final moments than by how long the discomfort lasted. The duration barely registered. The worst moment and the last moment did almost all the work.
This phenomenon has a name too: duration neglect. We are remarkably insensitive to how long an experience lasts, and remarkably sensitive to its sharpest and final notes.
Most organisations distribute their improvement effort evenly. Every touchpoint gets a little better. Average handle time drops by a few seconds across the board. The thinking is intuitive: a better experience everywhere should mean a better experience overall.
The science says otherwise. If memory is built from peaks and endings, then a flat, evenly-good experience is a strategic mistake. You will have spent real money smoothing moments that the customer was never going to remember anyway — and you may have left the two moments that actually get encoded exactly as average as everything else.
The implication is sharper than it first appears. It means two experiences with identical average quality can produce completely different memories, loyalty, and word of mouth, depending entirely on how they peak and how they close. It means a service interaction that goes slightly wrong in the middle but ends well can be remembered more fondly than one that was competent throughout and ended with a flat, transactional "is there anything else?"
So the question stops being "how do we make everything a bit better?" and becomes "where is the peak, where is the end, and are we treating them as the load-bearing moments they actually are?"
Here is where we part company with the breezier version of this idea that circulates on conference stages. The peak-end rule is real, but it is not a universal law you can apply mechanically to an eighteen-month B2B relationship the way you'd apply it to a 90-second phone call.
When researchers tested the rule on longer, more complex, more varied experiences — the kind with many distinct episodes rather than one continuous sensation — its predictive power weakened. For rich, heterogeneous journeys, the average of the experience and the overall emotional intensity started to matter again, sometimes more than the peak and end alone.
This matters enormously for customer experience, because a real customer journey is almost never a single sensation. It is dozens of episodes stretched across months and channels. Treat that like the cold-water experiment and you'll engineer a beautiful final invoice while a dozen quietly bad moments in the middle erode the relationship.
This is exactly why, in our own teaching, we don't stop at "peak and end." We work with peak, trend, and end — because in a long journey the direction of travel is itself a remembered signal. Is the relationship getting better or worse over time? That trend is the third moment customers encode, and it's the one a naïve peak-end model misses entirely.
Three practical moves follow from the evidence.
Find your real peak — and decide whether it's the one you want. Every journey has a most-intense moment. Often it's not the one you designed; it's a moment of friction, confusion, or relief. Map where the emotional intensity actually spikes, then ask whether that peak is working for you or against you.
Design the ending deliberately. The end of an interaction is disproportionately powerful and almost always under-managed. The close of a support call, the final screen of an onboarding flow, the last email of a project — these are not housekeeping. They are the moment that gets filed away as "what it was like to deal with this company."
Watch the trend, not just the touchpoints. For any relationship that unfolds over time, track whether the experience is improving or degrading across the arc. A journey that ends well but has been quietly deteriorating is a relationship in trouble, whatever your last survey says.
The uncomfortable truth at the start of this piece is also a gift. You don't have to make everything perfect. You have to know which moments your customers will actually keep — and refuse to let those moments be average.
Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). When More Pain Is Preferred to Less: Adding a Better End. Psychological Science, 4(6), 401–405.
Redelmeier, D. A., & Kahneman, D. (1996). Patients' memories of painful medical treatments. Pain, 66(1), 3–8.
Redelmeier, D. A., Katz, J., & Kahneman, D. (2003). Memories of colonoscopy: a randomized trial. Pain, 104(1–2), 187–194.
Strijbosch, W., Mitas, O., van Gisbergen, M., Doicaru, M., Gelissen, J., & Bastiaansen, M. (2019). From Experience to Memory: On the Robustness of the Peak-and-End-Rule for Complex, Heterogeneous Experiences. Frontiers in Psychology, 10, 1705. (Finds that for complex, multi-episode experiences, average valence and arousal predict remembered experience better than peak and end alone.)
