The Hook Model

What it is

Eyal set the model out in Hooked: How to Build Habit-Forming Products (2014), written for product teams rather than for other researchers. The claim is not that any single mechanism produces addiction. It is that four stages, run in sequence and repeated often enough, turn an occasional visit into an automatic one.

The loop opens with a trigger, something that cues the person to act. Early triggers are external: a push notification, an email, an icon sitting on a home screen. The action is the smallest possible behaviour done in anticipation of a reward, a tap, a swipe, a search, and Eyal borrows the reason it has to be small from B.J. Fogg's Behaviour Model, developed at Fogg's Stanford Persuasive Technology Lab and summarised as B = MAT: a behaviour happens only when motivation, ability and a trigger converge at the same moment. Of the three, ability is the cheaper lever to pull. Raising someone's motivation to check an app is slow, expensive, and noticed as manipulation once it's tried; stripping friction out of the action, one tap instead of three, is cheap and mostly invisible. That asymmetry is why the Action stage of the model reads almost entirely as a friction-removal exercise rather than a persuasion exercise.

The third stage, Variable Reward, is the one most associated with the model in popular writing, and it is drawn directly from B.F. Skinner's operant-conditioning research: a reward delivered on a variable schedule sustains a behaviour more powerfully than a reward delivered every time. Eyal splits what a product can offer into three types, rewards of the tribe, the hunt and the self, meaning social validation, resources or information, and a sense of mastery or completion. The unpredictability of which one lands, and how large it is, does the sustaining work. The average size of the reward matters less than not knowing what's coming.

The fourth stage, Investment, is where the loop closes. The user puts something in, a photo added, a follower gained, a streak extended, and that deposit does two things at once. It raises the cost of walking away, since loss aversion makes an accumulated streak feel like something to protect rather than something to start over. And it loads the next trigger: a streak invites tomorrow's visit, a follower list invites checking who's new, a saved search invites a notification when it matches. Investment is what makes the model a loop rather than four independent steps, because it hands the start of the next cycle back to stage one.

Run the loop often enough and the external trigger becomes redundant. The person starts to feel the cue internally, boredom, loneliness, uncertainty, before reaching for a device supplies it, and Eyal is explicit that the internal triggers a well-built habit eventually runs on are usually negative emotions rather than positive ones.

In effect

B.F. Skinner's variable-ratio schedules

The Variable Reward stage is not Eyal's invention; it's Skinner's, decades earlier. In operant-conditioning experiments, animals rewarded on a variable-ratio schedule, where the number of responses needed for a payoff changes unpredictably, keep responding faster and for longer than animals rewarded every time or on a fixed schedule. The uncertainty itself sustains the behaviour. Eyal's Variable Reward stage applies the same schedule to a screen rather than a lever.

The office email check-in

Tim Wu, in The Attention Merchants (2016), traces a lineage running from Skinner's reinforcement schedules through the 1970s office worker's email check-in to the pull-to-refresh gesture on a phone (pp.186–188), describing it as one continuous line rather than three separate inventions. The habit of opening a mail client every few minutes, unprompted by any notification because notifications didn't yet exist, was already running the Hook Model's loop: a self-generated trigger, the small action of clicking in, an unpredictable mix of nothing and something worth reading, and the mild investment of having organised an inbox that then needed checking again.

Duolingo's stack

Duolingo runs the full loop inside about a minute. A morning notification is the trigger. Starting a lesson takes one tap, the action. Surprise experience-point boosts and a friend leaderboard supply the variable reward. A streak counter is the investment, and because losing a streak feels like a loss rather than a missed gain, it is the piece most responsible for a return visit the next day. The same architecture, in different proportions, runs Instagram, TikTok and Tinder.

What it does not say

The Hook Model is a design framework with case-study support, not an experimentally validated psychological theory. Nobody has run a controlled trial comparing hooked and unhooked versions of a product and measured the difference against a null hypothesis; the evidence for it is that named products built this way retained users, which is consistent with the model working but does not rule out other explanations for the same retention. The model also rests on Fogg's B=MAT, itself a heuristic account of when behaviour occurs rather than a causally tested one.

It is also worth knowing that Eyal has not disowned the framework. His later book, Indistractable (2019), is often described as a course-correction, but its introduction is titled "From Hooked to Indistractable," and Eyal reports seeing the earlier book on shelves at Facebook, Google, PayPal and Slack without disputing what it did there. He quotes Paul Virilio: "When you invent the ship, you also invent the shipwreck." Indistractable runs the same four-part structure in reverse, attacking triggers instead of installing them, rather than retracting the model that produced the shipwreck in the first place.

Who developed it

Nir Eyal, a behavioural designer and writer, set the model out in Hooked (2014), building explicitly on B.J. Fogg's Behaviour Model from Stanford's Persuasive Technology Lab.

---