Habit loop

The four-stage descriptive framework for how habits operate: cue → craving → response → reward. Each stage answers a different question about why a habit fires.

StageWhat it isStreak example
CueThe trigger that initiates anticipation — a context, a time, a notificationThe streak number on the home-screen widget
CravingThe motivational state the cue elicits — the want for the predicted rewardAnticipation of the Duolingo jingle / “well done” beat
ResponseThe behavior performed in response to the craving — made easy by designTap into the app, autoprogress through one lesson
RewardThe payoff that satisfies the craving and reinforces the loopThe completion animation; +1 on the count; the chime

The framework is canonically James Clear’s reformulation (Atomic Habits, 2018) of B.F. Skinner’s operant-conditioning vocabulary, drawing on Charles Duhigg’s three-stage cue/routine/reward version (The Power of Habit, 2012) and extending it with the craving stage that explicitly hooks habit formation onto the dopamine-as-anticipation finding from Wolfram Schultz and Kent Berridge.

Why Streaks Work (It’s Not Discipline) (video) uses the four-stage form without crediting Clear, treating it as common ground.

The cue does most of the work

The single most operationally important finding the source surfaces:

Studies found that when people are reminded of their streak — just seeing the number — they’re more likely to keep going. We haven’t even opened the app yet, but the cue already triggered the loop.

Implication: in any product running on a habit loop, the cue is where most of the engineering should go. The streak number on a widget. The notification with the streak count. The lock-screen badge. The mascot’s emoji state. All of these fire the loop before the user has touched the product. Once anticipatory dopamine (Reward prediction error) is running, the craving → response stages are downhill.

This also explains why streak engagement does not depend on the lesson itself feeling rewarding. The source’s gloss:

A streak doesn’t need to feel satisfying every single time. It just needs to make our brain feel like something satisfying is about to happen.

The bridge to the dopamine substrate

The habit loop is the descriptive frame; Reward prediction error is the neuroscience underneath. Mapping:

Habit loop stageDopamine substrate
CueAnticipatory dopamine spike — RPE migrates backward onto the predictor
CravingThe motivational pull anticipatory dopamine produces (Berridge’s “wanting”)
ResponseThe behavior the wanting motivates
RewardLiking / consummatory satisfaction — quieter than the anticipation peak

Because the dopamine peak is at the cue (anticipation), not at the reward (receipt), it follows that the cue is where engagement is created — not the reward. This inverts the naive intuition that products should focus on making the reward great. Cue engineering compounds; reward engineering doesn’t.

The habit-vs-surprise overlay

A pure habit loop with a fully predictable reward eventually flattens — the brain catches up to the prediction and the dopamine signal weakens. The Habit-vs-surprise dilemma is what this concept becomes when you try to keep the loop firing over months. The standard fix: keep the cue and response predictable, but vary the reward magnitude. See Habit-vs-surprise dilemma for the full design tension.

Failure modes

  • Cue without craving — a notification the user ignores. The cue didn’t successfully trigger anticipatory dopamine; the user has learned the cue doesn’t predict a meaningful reward.
  • Craving without easy response — the user wants to engage but the response stage has friction (slow load, login screen, hidden CTA). The craving extinguishes.
  • Reward without anticipation — surprise rewards delivered out of any predicted context. Pleasant once, but no loop forms because there was no cue to migrate the dopamine onto.
  • Predictable reward — see above; the loop fires for a while, then flattens, and engagement drifts to obligation (see Streak).

Sources