Quick answer
The hook model is Nir Eyal's four-phase loop - trigger, action, variable reward, investment - that describes how products become habits. It is a lens for diagnosing repeat use, not a growth recipe: each phase has to be visible in your data before you act on it.
Formula: Trigger-to-action rate = actors ÷ users triggered; return rate = users back next period ÷ actors; investment share = investors ÷ returners; habit strength = share of a cohort active ≥ N days
What is the hook model?
The hook model is a four-phase loop — trigger, action, variable reward, investment — introduced by Nir Eyal in Hooked: How to Build Habit-Forming Products (Portfolio, 2014). It describes how repeated passes through the loop turn deliberate product use into automatic habit: users come back on their own, without costly advertising. Eyal first published the loop as a four-step article on NirAndFar.com in 2012, two years before the book.
The thesis of this article: the hook model is a framework for describing habits, not a growth recipe; to use it, you must be able to see every phase in your data. The model tells you where to look — cues, actions, payoffs, what users put in — but it does not promise that a streak counter or a push notification will manufacture retention. A phase you cannot measure is a phase you cannot fix.
Here is the whole loop on one example, a B2B reporting tool:
| Phase | The question it answers | Example (reporting SaaS) |
|---|---|---|
| Trigger | What cues the user to come? | A weekly metrics email, or the felt need to check numbers before a meeting |
| Action | What is the simplest thing they do in anticipation of a reward? | Open the app and run a report |
| Variable reward | What payoff keeps them wanting the next cycle? | An unexpected insight in this week's numbers |
| Investment | What do they put in that makes the next cycle easier or better? | Save a report template, invite a teammate |
Each pass should leave the product more valuable to that user, so the next trigger lands more easily. That makes the loop a diagnostic: when repeat use is weak, one of the four questions has a bad answer, and each answer is observable in analytics.
The four phases, with examples
Trigger. External triggers sit in the environment: an email, a push notification, a link from a colleague. Internal triggers sit in the user's head: boredom, uncertainty, a felt need. In Eyal's account, a product is habitual when an internal trigger — not your marketing calendar — sends the user back. Early on, external triggers do the work; the question is whether they hand off to an internal cue.
Action. The action is the simplest behavior done in anticipation of the reward — opening the app, running the report. Eyal builds on the Fogg behavior model, B = MAT: behavior happens only when motivation, ability and a trigger converge at the same moment. If the action is not happening, the user either wants the reward too little or the action is too hard. Making it easier — fewer steps, fewer fields, faster load — is the cheaper lever.
Variable reward. The idea traces to Skinner's operant-conditioning work: rewards on a variable schedule keep subjects wanting more than predictable rewards do. Hooked classifies three types: rewards of the tribe (social acceptance, likes, replies), the hunt (information, resources, the next interesting item) and the self (mastery, completion, a cleared inbox). The variability is the point: identical numbers every week make a report a chore; one that occasionally surfaces something unexpected is worth opening.
Investment. The user puts something in: time, data, effort or social capital — a saved template, an uploaded dataset, an invited teammate. Investment stores value that makes the product better for that user, and loads the next trigger: the teammate's comment becomes tomorrow's notification, the template makes next week's email worth opening. This closes the loop.
| Phase | Mechanism | Consumer example | What to measure |
|---|---|---|---|
| Trigger | External cue, handing off to an internal felt need | Weekly digest email; checking scores when bored | Trigger delivery, open rate, share of returns with no external cue |
| Action | B = MAT: motivation × ability × trigger, at one moment | Opening the app and refreshing the feed | Trigger-to-action rate, time to action, steps to action |
| Variable reward | Skinner's variable schedule; tribe / hunt / self | An unexpectedly good reply; a surprising video | Return rate next period; return by reward type |
| Investment | Stores value and loads the next trigger | Following accounts; saving a playlist | Investment share; return rate split by investment status |
How to measure each phase
Each phase maps to a rate computable from events. Define one core action, one investment action and one period (a week works for most products), then track one cohort through the loop:
| Phase | Metric | Formula | Events to track |
|---|---|---|---|
| Trigger → action | Trigger-to-action rate | Actors ÷ users triggered (also: actors ÷ openers) | trigger_sent, trigger_opened, core_action |
| Action | Core-action completion; time to action | Actors ÷ openers; median minutes from open to action | trigger_opened, core_action with timestamps |
| Variable reward | Return rate | Users back next period ÷ actors this period | Any qualifying activity event next period |
| Investment | Investment share | Investors ÷ returners; return rate split by investment status | investment_action (e.g. template saved, teammate invited) |
| Habit | Multi-day activity | Share of cohort active ≥ N days per period; mean active days | Daily activity flags per user |
Counting rules: same cohort and period boundaries in every ratio; the trigger-to-action denominator is users triggered, not messages sent; humans only, bots separated; logged-in users identified, so a "user" is a person, not a cookie. The habit row is stickiness in the spirit of DAU/MAU, but stricter: a monthly ratio can hide a user who binges once and vanishes; a ≥3-days-per-week share cannot. The return row is retention applied to actors, not signups; users past the multi-day bar are candidate power users. Keep first value separate: the aha moment is one good experience, a habit is it repeating without you pushing. A clean onboarding flow feeds the first pass; it is not the loop.
Worked example: a hook loop in numbers
Sample data throughout. A project tool takes 1,000 new users in week 0. Core action: create a project; external trigger: a weekly email; investment action: save a template or invite a teammate.
Week 1 — trigger and action. All 1,000 get the email; 400 open; 250 create a project. Trigger-to-action rate = 250 ÷ 1,000 = 25%; among openers, 250 ÷ 400 = 62.5%. Read together: most openers act, so targeting is fine — but three quarters of triggered users never open. The first bottleneck is the trigger, not the product.
Week 2 — return. 100 of the 250 actors come back: return rate = 100 ÷ 250 = 40%. Of those 100 returners, 60 save a template or invite a teammate: investment share = 60 ÷ 100 = 60%.
Week 3 — does investment load the next trigger? Split week-3 return by week-2 investment status: investors 45 ÷ 60 = 75%; non-investors 12 ÷ 40 = 30%; overall (45 + 12) ÷ 100 = 57 ÷ 100 = 57%. Investors return at 2.5× the non-investor rate — consistent with the model, but correlation: people who invest may simply be more motivated.
Week 4 — habit test. Count active days in week 4 for all 250 week-1 actors:
| Active days in week 4 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|
| Users | 120 | 50 | 35 | 25 | 12 | 6 | 2 | 0 |
Counts sum to 250. Share active ≥3 days = (25 + 12 + 6 + 2) ÷ 250 = 45 ÷ 250 = 18%. Mean active days = (0×120 + 1×50 + 2×35 + 3×25 + 4×12 + 5×6 + 6×2 + 7×0) ÷ 250 = 285 ÷ 250 = 1.14. Verdict: 48% (120 ÷ 250) did nothing in week 4; only 18% show multi-day use. Most actors are not habitual yet — normal, not failure: in Lally et al. (2010), automaticity rose along an asymptotic curve; time to 95% of plateau ranged from 18 to 254 days, and the BPS Research Digest summary puts the median at 66 days. Judge a hook by the ≥3-days share and mean rising across successive weekly cohorts over months — never by one spike.
One-phase fix. The weakest link is investment (60% of returners). Suppose prompting a template save right after the first project lifts investment share to 80% — 80 of 100 returners — while group week-3 return rates hold (investors 75%, non-investors 30%). Before: 60×0.75 + 40×0.30 = 45 + 12 = 57 (57%). After: 80×0.75 + 20×0.30 = 60 + 6 = 66 (66%). Effect: +9 percentage points, purely from moving returners into the higher-returning group. Caveats: sample data, an illustrative decomposition, not a causal estimate — group rates are assumed fixed, and investors self-select. Validate on successive cohorts with the same counting basis. The thesis in practice: every phase visible, the weakest located, a number to check next month.
Common mistakes and ethical limits
| Mistake | Why it fails | Fix |
|---|---|---|
| Treating the model as a growth recipe | Adding streaks and pushes without diagnosing which phase is weak | Measure all four rates first; fix the weakest link |
| Confusing a novelty spike with a habit | Launch-week activity decays; see app engagement | Track the ≥3-days share across successive cohorts for months |
| Counting triggers sent as actions | Inflates the loop's entry rate with messages nobody saw | Denominator = users triggered; report openers separately |
| Reading investment → return as causation | Investors self-select; more motivated users both invest and return | Label it correlation; compare comparable successive cohorts |
| Manufacturing compulsion | Extracts attention without delivering value; users burn out and churn | Measure to improve the reward, not to engineer dependence |
On that last row: habit is not addiction. A habit is automatic behavior that can serve the user — checking numbers before a meeting is useful; addiction is compulsive use that harms the user even when they want to stop. Eyal frames habit-forming technology as a "superpower" that should improve users' lives; the publisher's framing of Hooked is products people benefit from returning to. The boundary: deliver real value faster, never manufacture compulsion — no fake scarcity, no engineered anxiety as an internal trigger, no guilt mechanics.
The frequency caveat is the "habit zone" in Hooked: habit potential depends on perceived utility × frequency of use. A quarterly tax tool sits below the frequency threshold however useful it is; daily streaks produce annoyance, not habit. For low-frequency products, aim for top-of-mind recall at the moment of need. If-then rule: if return stays high only while you keep pushing external triggers, you have not built a habit — you are renting attention. A real hook shows returns that persist when the pushes stop.
How to see the hook model in TapCub
Run the loop from the worked example in TapCub, one measurement per phase. The point is the thesis: if a phase is not visible in the data, you cannot say the hook works.
- Define the loop's events first. In All-platform analytics › Data › Tracking plan, register
trigger_sent,trigger_opened,core_action(e.g.project_created) andinvestment_action(e.g.template_saved) with trigger timing, required properties, owner and status. Turn on 24-hour real-time validation to see arrival counts and required-property completeness before you trust any rate. Identify logged-in users withtapcub.loginso the loop is counted per person; anonymous visitors have no profile and do not enter segments. - Trigger → action. In All-platform analytics › Behavior › Behavior analysis, build the funnel
trigger_sent→trigger_opened→core_action: 250 ÷ 1,000 = 25% overall and 250 ÷ 400 = 62.5% among openers, which points at the trigger as the first bottleneck. - Return. On the same page, run weekly retention with
core_actionas the start event: period 1 is 100 ÷ 250 = 40% of week-1 actors. - Investment effect. In All-platform analytics › Users › Segments & tags, define a rule segment "did
investment_action". A segment defined once can be carried into retention as a group, so investor vs non-investor week-3 return (75% vs 30%) needs no rebuilt cohort. Label the gap as correlation, as the example does. - Habit. Back in All-platform analytics › Behavior › Behavior analysis, the stickiness view counts how many people were active on N days in each week, with a share, which is the week-4 table directly: 18% at three days or more. The lifecycle view on the same page splits each week into New / Returning / Resurrecting / Dormant; watch Resurrecting after you change a trigger.
Boundaries: TapCub does not run A/B tests or feature flags, so validate the investment fix on successive cohorts with the same counting basis. Identified bots are counted separately and cannot inflate the rates.
See also: Gamification Examples: Which Mechanics Change Behavior, and How to Check.
Sources
- Nir Eyal, The Hooked Model: How to Manufacture Desire in 4 Steps, NirAndFar.com, 2012-03-04.
- Penguin Random House, Hooked: How to Build Habit-Forming Products, ISBN 9781591847786.
- Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010), How are habits formed: Modelling habit formation in the real world, European Journal of Social Psychology, 40(6), 998–1009.
- BPS Research Digest, How to form a habit, 2010-10-06 (66-day median summary of Lally et al.).
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Adoption Consultant
Explores customer onboarding, live conversations, and self-service, turning common usage questions into practical guidance and improvements.