Working · 2020-08SecondaryRetentionAnalytics

Read cohort retention curves for a plateau

Plot the share of each signup cohort still active over time and look for where the curve flattens.

Also called: retention curve, N-day retention

Claim and evidence

The claimthe average app loses 77% of its DAUs within the first 3 days after the install.

Mobile apps (Quettra data) · andrewchen.com · undated

What the numbers show

77%

of daily actives the average mobile app loses within days of install (Quettra data, via Andrew Chen, 2015)

andrewchen.com

Why it works

Diagram, not data
Group users bysignup periodPlot the sharestill activeFind where thecurve flattensCompare cohortsafter changes
Group users by signup periodPlot the share still activeFind where the curve flattensCompare cohorts after changes

A flattening curve means a stable core uses the product; one that decays to zero signals no product-market fit. Early decay is steep for most apps.

How to run it

  1. Build weekly or monthly signup cohorts
  2. Plot % active by period
  3. Locate the plateau level
  4. Compare cohorts after each change

Ways to run it

Design options to test, not promised results.

  1. Option AWeekly cohorts

    Use weekly cohorts for fast-moving consumer products where the drop happens in days.

  2. Option BMonthly cohorts

    Use monthly cohorts for B2B or slower-use products and read the plateau over several months.

  3. Option CBefore and after a change

    Overlay cohorts from before and after a product change to see whether the plateau moved.

Timeline

  1. 12015 Mobile apps (Quettra data) — Andrew Chen with Quettra's founder showed typical mobile decay.
  2. 22020-08 status checked: Working

Dated examples

  1. 2015

    Mobile apps (Quettra data)

    Andrew Chen with Quettra's founder showed typical mobile decay.

    Average app lost most daily users in days.

    “the average app loses 77% of its DAUs within the first 3 days after the install.”
    ✓ quote checked word-for-word on the page

    andrewchen.com · undated

Does it still work?

Workingas of 2020-08

The 2015-era stat is dated; method remains standard. source

Risks

  • Old benchmark figures
  • Blended cohorts hide segment differences

Sources

  1. lennysnewsletter.com · Lenny's Newsletter · Lenny Rachitsky · 2020-08-25
    “In this case, the curve flattens at about 10%, which means only 10% of your users stick around long-term.”
    ✓ quote checked word-for-word on the page
  2. andrewchen.com · Andrew Chen blog · Andrew Chen · undated
    “the average app loses 77% of its DAUs within the first 3 days after the install.”
    ✓ quote checked word-for-word on the page

Change log

  • 2026-11-20 — harvest v0.1 (retention)