Skip to content
😊 ProductFeeling

Peak–End Rule

People remember an experience by its most intense moment and by how it ends—not by its average quality or duration.

Definition

Section titled “Definition”

The peak–end rule (Kahneman, Fredrickson, and colleagues) is the finding that retrospective evaluation of an episode is disproportionately determined by two samples: the emotional peak (best or worst moment) and the end. Duration barely registers (“duration neglect”).

flowchart LR
  subgraph experience [Lived episode — many moments]
    A[Step] --- B[Step] --- P((Peak)) --- C[Step] --- E((End))
  end
  experience --> M[Remembered story]
  P -.-> M
  E -.-> M

The user who rates your product later is not replaying the session; they are recalling a compressed story built from its extremes and its finish.

Why it matters

Section titled “Why it matters”

Funnels treat every step as equal; memory does not. A session with nine smooth steps and one moment of panic is remembered as panic. A cancellation flow is a few seconds of a multi-year relationship, yet it can dominate what the customer says about you forever, because it is both a potential negative peak and the ending. The rule tells you where designed feeling pays compound interest: peaks and endings first, averages later.

Deep dive

Section titled “Deep dive”

Three practical consequences:

  1. Engineer the peaks you want. First value, a finished project, a recovered mistake—these are candidate positive peaks. Success Moments exists to mark them so they register; Time to Value pulls the first peak earlier.
  2. Defuse the negative peaks you already have. Errors, scary waits, and payment failures are your current peaks whether you designed them or not—negativity bias makes them count double. Graceful Recovery, Loading Feedback, and Fail Safe are peak-defusal tools.
  3. Design endings on purpose. Every scale of the product has an ending: the end of a task, a session, a subscription. Most products lavish design on beginnings and let endings happen by accident. Graceful Exit applies the rule at the largest scale: the final impression is the one that decides whether a churned user returns or warns others away.

One caution: the rule describes memory, not ethics. Using a manufactured emotional spike to paper over a genuinely poor experience is manipulation, and users eventually reconcile the story with reality. The honest application is to make real value memorable and real failures survivable—not to fake peaks.

For engineers and agents

Section titled “For engineers and agents”
  • Peaks are enumerable in code: the first-success event, every user-visible error path, payment results, and final screens of flows. Grep for them—throw sites that surface to the UI, catch blocks that render fallbacks, terminal states of multi-step forms—and you have the list of moments that will dominate user memory.
  • Allocate code-review rigour by emotional stakes, not code size. A two-line change to an error message or a cancellation screen deserves more scrutiny than two hundred lines of internal refactor, because it edits a probable peak or ending.
  • Endings are states you already have: the last screen of every flow, session end, subscription end. Check each terminal state for what it leaves behind—confirmed work, a clear next step, or a shrug. A flow that simply stops rendering has an ending; it’s just an accidental one.
  • Instrument peaks directly: error-path telemetry with user-visible outcome (“saw fallback, retried, succeeded” vs. “saw fallback, left”), time-to-first-success for new users, and completion screens reached. Session-average satisfaction hides exactly what the rule says matters.
  • For agents: when summarising a flow audit, order findings by peak/end proximity—an issue on the first-value moment or the final screen outranks the same issue mid-flow.

Where it shows up

Section titled “Where it shows up”

Further reading

Section titled “Further reading”