Framework · Measurement · Hooked
Reading retention curves
Before comparing retention, pin four choices: bounded, unbounded or bracket retention; "active" as the value event; the denominator; and the cohort grouping. Then read the shape: declining to zero (no fit yet), flattening (a core found lasting value) or a smile. Newer cohorts beating older ones is the best sign an improvement program works.
Plain label: cohort curve shapes and definitionsAlso known as: cohort retention, leaky bucket, smile curve, cohort stacking, bounded retention
Pin the definitions. N-day (bounded) retention gives low numbers; unbounded (day N or later) gives high ones; bracket retention (for example days 22-28) suits weekly products. Active means the value event, never a login. Published benchmarks use different denominators (registered users, users with a transaction, paying subscriptions), which is why they disagree by 5-10x.
Shapes. Declining to zero: fix value or activation before buying traffic. Flattening: find who is on the flat part and pull more users onto their path. Smile: confirm the returning event is real value.
Cohort stacking. Newer cohorts above older ones at the same age means the product is getting better for new users.
A dated example. Evernote's paying share rose from about 0.5% after month 1 to 26% at month 42 as habits formed (Hooked, dated 2011). HK-MEA11 (Hooked Ch.1)
Limits
What we don’t claim
- The Hook Model is a practical design tool, not a validated or complete theory of behavior.Hooked, Introduction
Sources
Hooked · Nir Eyal
Ch.1
What is good retention · Lenny's Newsletter (Lenny Rachitsky and Casey Winters), 2020
Benchmark confidence B. A reputable dataset.
Source IDs: HK-MEA11
Book content is paraphrased, never quoted at length. Last verified October 24, 2026.
Your product
Find the stage your product is stuck in
Gamify It reads your product, picks the one move that matters at its stage, and sizes the test to your traffic. Connecting is free; a Pro subscription starts it.