Working · 2026-11PrimaryRetentionLifecycle

Protect the notification channel

Optimize timing and copy of push notifications but cap volume, and kill winning notifications that are too spammy.

Also called: notification budget, don't kill the goose

Claim and evidence

The claimwe established one foundational rule: protect the channel.

Duolingo · lennysnewsletter.com · 2023-02-28

What the numbers show

No published number for this tactic has passed our word-for-word check yet, so there is no chart.

Why it works

Diagram, not data
A notification testwinsKeep it if opt-outshold steadyRetire it if opt-outsriseThe channel staysusable
A notification test winsKeep it ifopt-outs holdsteadyRetire it ifopt-outs riseThe channel stays usable

Every extra push raises the chance a user opts out, and opt-out is permanent loss of a free channel. Treating channel health as a constraint stops short-term lifts from burning long-term reach.

How to run it

  1. Set a rule: tests can change timing and copy, not volume
  2. Require a high bar before adding a new notification
  3. Track opt-out rate alongside the lift
  4. Retire winners that damage opt-outs

Ways to run it

Design options to test, not promised results.

  1. Option ATiming and copy only

    Tests may change when and what you send, never how much, as in Duolingo's rule.

  2. Option BHigh bar for new notifications

    Require a strong case before any new notification type is added.

  3. Option COpt-out guardrail

    Report the opt-out rate next to the lift for every notification test.

  4. Option DRetire spammy winners

    Switch off notifications that win tests but feel too spammy, as Duolingo has done.

Timeline

  1. 12023 Duolingo — Team could tune push timing and copy but not send more.
  2. 22024 Duolingo — Engineering VP says some notifications that won tests were killed as too spammy.
  3. 32026 Airship benchmark — Vendor survey cites over-frequency as top opt-out reason.
  4. 42026-11 status checked: Working

Dated examples

  1. 2026

    Airship benchmark

    Vendor survey cites over-frequency as top opt-out reason.

    “the top reason users opt out is because messages get sent “too frequently.””
    ✓ quote checked word-for-word on the page

    airship.com · undated

  2. 2024

    Duolingo

    Engineering VP says some notifications that won tests were killed as too spammy.

    “We've killed some notifications that were huge wins”
    ✓ quote checked word-for-word on the page

    lennysnewsletter.com · 2024-04-02

  3. 2023

    Duolingo

    Team could tune push timing and copy but not send more.

    Rule stated by former CPO.

    “we established one foundational rule: protect the channel.”
    ✓ quote checked word-for-word on the page

    lennysnewsletter.com · 2023-02-28

Does it still work?

Workingas of 2026-11

Airship's 2025-data benchmark shows median iOS direct open rate down 13.71% and Android volume down 15%, consistent with fatigue. source

Risks

  • Hard to see opt-out cost in a single test window

Sources

  1. lennysnewsletter.com · Lenny's Newsletter · Jorge Mazal · 2023-02-28
    “we established one foundational rule: protect the channel.”
    ✓ quote checked word-for-word on the page
  2. lennysnewsletter.com · Lenny's Newsletter · Sean Colombo · 2024-04-02
    “We've killed some notifications that were huge wins”
    ✓ quote checked word-for-word on the page
  3. airship.com · Airship · Airship · undated
    “the top reason users opt out is because messages get sent “too frequently.””
    ✓ quote checked word-for-word on the page

Change log

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