Cognitive Biases · CB-33
People judge an entire experience mainly by its most intense moment and how it ended — the duration and the average quality of the experience barely register at all.
A memory heuristic in which people judge the overall quality of a past experience largely based on its peak (most intense point, positive or negative) and its ending, rather than the total sum or average of every moment — a phenomenon that also produces the counterintuitive finding that extending an unpleasant experience with a milder ending can make the overall memory of it less negative than a shorter version without that ending.
Established by Daniel Kahneman, Donald Redelmeier, and Barbara Fredrickson through a series of studies in the 1990s, most famously involving patients undergoing colonoscopy procedures, comparing remembered pain against real-time pain reports.
The Mechanism
Adding a milder ending can improve the memory of a worse experience
Kahneman and Redelmeier's actual colonoscopy patients rated a longer procedure (with more total pain, but a milder ending) as less unpleasant in memory than a shorter procedure that ended abruptly at a high-pain moment — objectively more total suffering was remembered as better, purely because of how the experience ended.
01 · TOTAL DURATION AND AVERAGE INTENSITY ARE LARGELY IGNORED BY MEMORY
This is the counterintuitive core of the finding
Classical utility theory would predict that a longer painful experience should be remembered as worse than a shorter one — the peak-end rule shows memory instead discounts total duration almost entirely, weighting mainly the single most intense moment and the final moments before the experience ends.
02 · IT MEANS 'REMEMBERED SELF' AND 'EXPERIENCING SELF' CAN DISAGREE SHARPLY
Kahneman's broader distinction between how you feel in the moment and how you remember it later
The peak-end rule is a specific instance of Kahneman's broader finding that the 'experiencing self' (moment-to-moment feeling) and the 'remembering self' (how an experience is recalled and evaluated afterward) can diverge substantially — decisions about future experiences are typically made by the remembering self, which systematically discounts duration.
03 · IT'S DELIBERATELY EXPLOITED IN SERVICE DESIGN, HOSPITALITY, AND MEDICAL PRACTICE
Ending an experience well pays disproportionate dividends in customer and patient memory
Businesses and clinicians aware of the peak-end rule deliberately design experiences to manage the peak (minimizing the worst moment where possible) and especially the ending (ensuring a positive or at least gentle final impression) — since customers' and patients' remembered satisfaction, which drives future behavior, is disproportionately shaped by these two points rather than the full experience.
Where It Fails / Inversion
Where it fails / inversion
Not every decision should be optimized purely for remembered evaluation — if the actual, moment-to-moment experienced wellbeing matters in its own right (not just how it's later recalled), designing purely around the peak-end rule to improve memory while ignoring real-time suffering during the bulk of the experience is a genuine ethical shortcut, not a neutral design choice.
How To Use It
Worked example · designing the end of a customer service interaction
A customer service process that resolves the core problem competently but ends the interaction abruptly, without a warm closing, will likely be remembered less favorably than an equally competent resolution that ends with genuine acknowledgment and a positive final note — even if the total time and effort involved were identical, the ending disproportionately shapes what the customer actually remembers and reports afterward.
How to use it
When designing or evaluating any extended experience — a customer interaction, a project, an event — pay disproportionate attention to its most intense moment and, especially, how it ends, since that's what will actually be remembered and will disproportionately shape future decisions, regardless of how the full experience averaged out.
See Also