People do not remember an experience as an average. They remember its best or worst moment and how it ended. That is the peak-end rule. A loyalty program is a string of moments. Sign-up, a first reward, a redemption and a status change are all moments. So the rule tells a team where to spend effort. This entry covers the rule, the evidence, its limits and where a program can place a good peak and a good ending.
What is the peak-end rule?
The peak-end rule says people judge an experience mostly by its most intense moment and by its final moment. Daniel Kahneman and his colleagues built the case in 1993. They found that duration plays a small role in how people remember aversive experiences. Such evaluations are often dominated by the discomfort at the worst moment and at the end (Kahneman et al., 1993).
A companion finding is duration neglect. Fredrickson and Kahneman showed people film clips of different lengths. Memory of each clip tracked a weighted average of snapshots. It was as if length did not matter (their 1993 paper).
What did the experiments show?
The classic test used cold water. In a short trial, people held one hand in water at 14 degrees Celsius for 60 seconds. In a long trial, they held the other hand in the water for 60 seconds at 14 degrees, then 30 seconds more as it warmed to 15 degrees. A significant majority chose to repeat the long trial, apparently preferring more pain over less (Kahneman et al., 1993).
| Study | Setting | What it found |
|---|---|---|
| Kahneman et al. (1993) | Cold water, 60 s vs 90 s | Most chose to repeat the longer trial with the milder ending |
| Redelmeier and Kahneman (1996) | 154 colonoscopy and 133 lithotripsy patients | Remembered pain tracked the peak and the last 3 minutes, not length |
| Redelmeier, Katz and Kahneman (2003) | 682 colonoscopy patients, randomized | A milder added ending made the procedure rate as less unpleasant; return rates were higher |
| Alaybek meta-analysis | 85 peak studies and 95 end studies | Peak-end effect large; duration effect near zero |
| Miron-Shatz (2009) | Whole-day reports from about 2,400 participants | The average of the day’s feelings predicted the evaluation best |
Two patient studies track behavior. One logged pain in 154 patients having a scope exam and 133 having a stone treatment. Memory of total pain followed the peak and the last 3 minutes. Long procedures were not recalled as worse (Redelmeier and Kahneman, 1996).
A 2003 trial split 682 patients at random. Half got the usual ending. Half got a short extra interval with the scope held still. That group rated the last minutes as less painful, 1.7 against 2.5 out of ten. They also rated the whole exam as less unpleasant. About half of all patients came back for a repeat exam over roughly five years. The extra interval group came back more often (odds ratio 1.41) (Redelmeier et al., 2003). A better ending changed what patients did next.
Is the peak-end rule always true?
It is a strong tendency, and it has limits. Balca Alaybek’s meta-analysis covered 85 studies of peaks and 95 of endings. The peak-end effect was large (r = .57) and held in most settings. The duration effect was close to zero. The peak-end effect was also about as strong as the plain average (Alaybek, George Mason University). The figure comes from the dissertation abstract. The 2022 journal paper later had a corrigendum, so check it before quoting exact sizes.
The limit that matters most to a loyalty team is length. Talya Miron-Shatz asked about 2,400 participants in three countries to rate each episode of their previous day. The average of the feelings predicted the evaluation of the whole day best, and ends were ignored. Peaks and lows counted more when people compared experiences than when they rated one alone (Miron-Shatz, 2009). A member’s life with a program runs for months, so it is closer to a day than to a three-minute procedure. Expect peaks to help and the baseline to matter.
Waiting shows another limit. As Wikipedia sums up a study by Kahneman and Ziv Carmon, some people felt unhappy for most of a wait. The line sped up near the end. They rated the whole wait as satisfying. The final moment predicted their verdict. For a program, a quick last step can lift a slow journey, but only when it beats what the member expected.
How does the peak-end rule show up in loyalty programs?
Programs create peaks on purpose. Starbucks gives each member a Birthday Reward of one free handcrafted drink, food item or bottled beverage. Gold members get seven days to redeem it, and Reserve members get thirty (Starbucks Rewards Terms). Reserve members are also invited to request a personalized card. Those are peak moments that scale with status.
Endings need the same care. A demotion, a lapse and a redemption each close an episode. Delta offers members who lose a tier after a life event the previous Medallion tier back for 90 days (Delta Reclaim My Status). That turns the end of a bad stretch into a second chance.
Where are the peaks and endings in a member journey?
Candidate moments are easy to list. Each is a hypothesis to test with a short survey and repeat-earn data.
- Sign-up, where a welcome reward can be an early peak.
- The first reward earned and redeemed, a likely peak.
- Each redemption, which ends an earn and burn cycle.
- A status change, up or down.
- A lapse and the win-back offer that follows it.
- An exit, which is the final impression.
How can a loyalty team use the peak-end rule?
Start by listing the moments in the member journey and marking the best and worst. Then work on three things. Raise the peak, usually the first reward or a surprise. Fix the worst moment, often a failed redemption or an unclear expiry. The studies say the worst moment weighs heavily in memory, so a fix there can beat a small perk. And redesign the ending of each episode, such as the last screen of a redemption.
Pair a moment rating with behavior. A rating captures memory. Behavior shows what the memory did. In the scope trial, a better memory raised the chance that patients came back.
For illustration, a program changes only the last step of redemption. Group A sees the usual confirmation. Group B sees a thank-you with the new balance and the next reward within reach. If group B rates the redemption 8.3 out of 10 against 7.8, and 41% earn again within 30 days against 38%, the ending is paying. Judge it on the repeat earn rate, since the rating is only the memory.
How GRAVTY handles the peak-end rule
Peaks and endings are things a team sets. A reward or a message can change in minutes. The GRAVTY technology page says business teams set rewards for any segment without code. Our guide to loyalty personalization covers how to tailor each moment. Agentic AI Compass then compares results by segment, so a team can see which moment moved repeat earning.
Frequently asked questions
What is an example of the peak-end rule?
The best-known example is a trial of 682 patients. Half got the usual ending to a medical exam. Half got a short, milder one. The milder ending rated as less unpleasant, and those patients came back for repeat exams more often (Redelmeier et al., 2003).
Who discovered the peak-end rule?
Daniel Kahneman and colleagues built the evidence in the early 1990s. In the 1993 cold-water study, people chose to repeat a longer trial that ended less painfully (the 1993 study). Fredrickson and Kahneman added duration neglect the same year (their paper).
What is duration neglect?
Duration neglect is when the length of an experience has little effect on how people remember it. People who watched film clips of different lengths rated them by snapshots of how they felt (1993 study). Length added little. A large review of studies found the same (the review).
Does the peak-end rule apply to customer experience?
It applies to short, discrete moments better than to long relationships. For whole-day evaluations, the average of the feelings predicted the verdict best and ends were ignored (Miron-Shatz, 2009). A loyalty program spans months, so teams should fix the baseline and also build peaks and endings.
How do loyalty programs use the peak-end rule?
They place surprise rewards at peaks and repair moments at endings. Starbucks gives a Birthday Reward, and the redemption window grows with status (Starbucks Rewards Terms). Delta returns a prior tier for 90 days after a status loss tied to a life event (Delta Reclaim My Status).
