Jun 20, 2019

The Peak-End Rule: Why Endings Matter More Than Experience

Behavioral Economics Psychology UX Design Customer Experience

The Memory Paradox

Imagine two versions of a colonoscopy procedure:

Version	Duration	Pain Level	Ending
A	8 minutes	High pain throughout	Ends abruptly at peak pain
B	24 minutes	Same pain + 16 extra minutes of mild discomfort	Ends gradually, low pain

Which procedure would patients rate as worse?

Counterintuitively: Patients rated Version A as worse, despite Version B having three times the duration and more total pain.

This is the Peak-End Rule—and it fundamentally changes how we should design experiences.

Kahneman’s Cold Water Experiment

Daniel Kahneman demonstrated this phenomenon with a simple experiment:

The Setup

Participants immersed their hand in painfully cold water (14°C) in two trials:

Trial	Duration	Temperature Profile
Short Trial	60 seconds	Constant 14°C
Long Trial	90 seconds	60s at 14°C, then 30s gradually warming to 15°C

The Surprising Result

When asked which trial they’d repeat, 80% chose the Long Trial—even though it had:

More total pain (30 extra seconds)
More cold exposure
Longer duration

Why? The ending was slightly less painful.

The Peak-End Rule Explained

Definition: Our evaluation of a past experience is dominated by two moments: the peak (most intense point) and the end (final moment).

What Our Brains Ignore

Factor	Weight in Memory
Peak intensity	High
Final moment	High
Total duration	Almost zero (Duration Neglect)
Average experience	Low

The Implications

This means:

A 2-week vacation isn’t remembered as “better” than a 1-week vacation
A 3-hour movie isn’t rated higher than a 90-minute one (if peaks/ends equal)
A painful process with a gentle ending is remembered more favorably

Duration Neglect: Time Is (Almost) Irrelevant

Duration Neglect is the companion finding: we barely factor duration into our evaluations.

Example: Vacation Memories

Vacation A	Vacation B
1 week	2 weeks
Amazing peak moment (snorkeling)	Same amazing moment
Relaxing final day	Rushed airport ending

Remembered as: Vacation A was better.

Despite B having:

Twice the duration
More total enjoyable moments
Same peak experience

The rushed ending of B dominates the memory.

Business Applications: Designing for Memory

The IKEA Model

IKEA’s experience seems designed to torture customers:

Maze-like layout (no shortcuts)
Exhausting walk through entire store
Self-service warehouse
DIY furniture assembly

Yet customers love IKEA. Why?

Pain Point	Peak-End Strategy
Long walk	Peak: Surprisingly cheap items discovered (“Only $5!”)
Exhaustion	Peak: Inspiring showrooms mid-journey
Checkout frustration	End: $1 hot dog &$ 0.50 soft-serve ice cream

The final memory: “Great deals, delicious cheap food.”

Disney’s Approach

Disney designs experiences with explicit peak-end management:

Element	Purpose
Castle at center	Create anticipation (sets up peak)
Fireworks at closing	Guarantee powerful ending
Character meet at exit	Sweet final interaction
”Have a magical day!”	Verbal ending ritual

Apple Store Experience

Moment	Design Choice
Entry	Open, inviting, no pressure
Peak	Hands-on product trial (emotional high)
End	Genius Bar resolution or purchase celebration
Exit	Staff says goodbye by name

Apple invests heavily in the leaving experience, not just the buying.

Service Design Framework

The Peak-End Design Process

1. Map the customer journey
2. Identify potential peak moments (positive or negative)
3. Amplify positive peaks
4. Mitigate negative peaks
5. Engineer a strong positive ending
6. Don't worry about duration (within reason)

Creating Peaks Through Surprise

Peaks don’t just happen—they’re engineered. The most reliable way to create a peak is through unexpected positive experiences:

Strategy	Example
Unexpected upgrade	Airlines upgrading to business class
Hidden delights	Mailchimp’s high-five animation after sending
Over-delivery	Amazon arriving a day early
Personal touch	Handwritten thank-you note in package
Celebration moments	Duolingo’s streak celebrations

The Surprise Formula: Peak = Expectation Gap × Emotional Intensity
The bigger the positive gap between what was expected and what was delivered, the stronger the peak.

Practical Checklist

Question	Action
Where is the emotional high point?	Amplify it—make it memorable
Where can we exceed expectations?	Add a surprise moment
Where is the pain point?	Either fix it or ensure it’s not the ending
How does the experience end?	Design a positive closing ritual
Are you optimizing the middle?	Stop. Invest in peaks and ends instead.

Medical Applications

The colonoscopy study led to actual changes in medical procedure:

Before (Traditional)

Procedure ends immediately when complete
Last moment = peak discomfort

After (Peak-End Optimized)

Extra minute of scope stationary (less painful)
Last moment = reduced discomfort
Patients more likely to return for follow-ups

Impact: Potentially life-saving, as patients are more willing to undergo preventive screenings.

Digital Product Design

Onboarding Flows

Poor Design	Peak-End Optimized
Ends with “Setup complete”	Ends with accomplishment (“You just created your first X!”)
Long form, same intensity	Front-load hard parts, end with easy wins
No memorable moment	Include a “wow” moment mid-flow

Error Handling

Poor Design	Peak-End Optimized
Generic error message	Recovery + small positive (discount code)
Session ends on error	Graceful save + friendly message
User left frustrated	Transform negative peak into resolution story

Subscription Cancellation

Poor Design	Peak-End Optimized
Guilt trip, friction	Easy process, genuine thank you
Ends with “You’re unsubscribed"	"We’ll miss you. Here’s a gift for when you return.”
Burns the bridge	Leaves door open positively

Detecting Peak-End Effects in Data

import pandas as pd
import numpy as np

def analyze_journey_peaks(session_df, retention_df=None):
    """
    Analyze customer journey for peak-end effects.
    
    Expects columns:
    - session_id: unique session identifier
    - timestamp: event time
    - satisfaction_score: momentary satisfaction (1-10)
    
    Optional: retention_df with session_id and returned (bool)
    """
    
    results = []
    
    for session_id, group in session_df.groupby('session_id'):
        group = group.sort_values('timestamp')
        
        # Calculate metrics
        peak_score = group['satisfaction_score'].max()
        end_score = group['satisfaction_score'].iloc[-1]
        avg_score = group['satisfaction_score'].mean()
        duration = len(group)
        
        # Peak-End prediction
        peak_end_prediction = (peak_score + end_score) / 2
        
        results.append({
            'session_id': session_id,
            'peak': peak_score,
            'end': end_score,
            'average': avg_score,
            'duration': duration,
            'peak_end_score': peak_end_prediction
        })
    
    results_df = pd.DataFrame(results)
    
    # If we have retention data, this is the key test:
    # Does Peak-End Score predict FUTURE BEHAVIOR better than Average?
    if retention_df is not None:
        results_df = results_df.merge(retention_df, on='session_id')
        print("Correlation with FUTURE RETENTION (the real test):")
        print(f"  Peak-End Score ↔ Retention: {results_df['peak_end_score'].corr(results_df['returned']):.3f}")
        print(f"  Average Score ↔ Retention:  {results_df['average'].corr(results_df['returned']):.3f}")
        print(f"  Duration ↔ Retention:       {results_df['duration'].corr(results_df['returned']):.3f}")
    else:
        # Fallback: correlate with average (less meaningful)
        print("Correlation analysis (connect to retention/NPS for real insights):")
        print(f"  Peak-End Score: {results_df['peak_end_score'].corr(results_df['average']):.3f}")
        print(f"  Duration:       {results_df['duration'].corr(results_df['average']):.3f}")
    
    return results_df

What to Measure

Metric	What It Reveals
Peak moment identification	Which touchpoint drives memory
End moment score	How journey closes
Peak-End vs. Average correlation	Which better predicts retention
Duration vs. satisfaction	Likely weak correlation

The Two Selves: Experiencing vs. Remembering

Kahneman distinguishes between two selves:

Self	Focus	Question It Answers
Experiencing Self	Moment-to-moment feelings	”How do you feel right now?”
Remembering Self	Retrospective evaluation	”How was your vacation?”

The Conflict

Experiencing Self	Remembering Self
Cares about total duration	Ignores duration
Values every moment equally	Values peaks and ends
Lives in the present	Creates the narrative

Which Self Matters for Business?

Usually the Remembering Self, because:

Word-of-mouth recommendations come from memory
Repeat purchases are based on remembered experience
Ratings and reviews reflect remembered, not experienced, satisfaction

The Psychology of the “End”

“A bad ending can ruin a great movie, but a great ending can save a mediocre one.”

The ending has disproportionate power because it’s what we carry forward:

Experience	Ending	Memory
Perfect date, argument at the end	😡	“It was a disaster”
Mediocre date, sweet goodbye	😊	“It was nice”
Great vacation, lost luggage on return	😤	“Never flying with them again”
Average vacation, surprise upgrade home	✈️	”What a trip!”

The Final Moment Checklist

✅ Is the last touchpoint positive?
✅ Does it leave them wanting more?
✅ Is there a clear “closing ritual”?
❌ Does it end on administrative tasks? (Bad)
❌ Does it end with payment? (Move payment earlier)

Summary

Concept	Implication
Peak-End Rule	$\text{Memory} = \frac{\text{Peak} + \text{End}}{2}$
Duration Neglect	Length barely matters within reason
Experiencing vs. Remembering	Design for the remembering self
Surprise = Peaks	Exceed expectations strategically
Endings are Fragile	A bad ending can erase everything

⚠️ Caveat on Duration Neglect: While duration is almost irrelevant, there’s a threshold. Disney’s fireworks can’t save a 5-hour queue. If duration crosses into “unreasonable” territory, no ending can rescue the experience. The rule works best within “acceptable” duration ranges.

Design Principles

Create at least one memorable peak (through surprise)
Never end on a low note (endings are fragile)
Don’t obsess over duration (but stay within reason)
Front-load pain, back-load pleasure
Design closing rituals