Luca Barberis

Which channels deliver retained users?

№ 136 · D1/D7/D30 retention by channel

01

Definition

MetricClassic day-N retention split by acquisition source

Unit% of an install cohort returning on day N

Bar chart of retention metric (e.g., D7) by acquisition channel. Build by joining install events with D7-active events per channel. Identifies channels bringing in committed players versus tire-kickers. Example: ASA D7 retention 32%; Meta 22%; TikTok 18% (ASA wins on quality).

D7
Day 7 retention. Share of a signup cohort returning on the seventh day after install or signup.
02

Benchmarks

Bottom 30%MedianTop 30%
D1 under 20%, D7 under 7%, D30 under 2%D1 25% to 30%, D7 10% to 15%, D30 3% to 6%D1 35% to 45%, D7 18% to 25%, D30 8% to 12% for a commercially marketed title

Commercially marketed mobile games running paid user acquisition, iOS and Android combined, 2024 to 2026. Classic retention, meaning returned on exactly day N, not rolling retention. Paid cohorts; organic and incentivised cohorts are benchmarked separately in the category table. · The largest source of disagreement between published benchmarks is population, not measurement. GameAnalytics pools 11,600 titles including a long tail of unmarketed games and lands on a median D7 of 3.42% to 3.94%. Datasets restricted to commercially scaled titles land three to four times higher, and Adjust's cross-genre D1 of 27% for 2025 sits close to the marketed-title median used above. Use the pooled figure as context and the marketed band as a target only if you are actually running paid UA at scale. Measurement is also degrading: global ATT opt-in sits near 39% among users shown the prompt, and SKAN and AdAttributionKit suppress postbacks for campaigns under roughly 20 daily installs, so channel-level retention below that volume is modelled rather than observed. The channel deltas in the category table are directional multipliers and lifts from Unity and AppsFlyer rather than a measured percentile distribution by source.

By category
CategoryBottom 30%MedianTop 30%
Organic and owned re-engagementnot publishedhighest retaining cohort in most titles; owned media remarketing through SMS, email, invites and cross-promotion outperforms paid media on retentionnot published
Paid social (Meta, TikTok)not publishedshould land within 10% to 15% of the organic baseline; a wider gap points at targeting and acquisition quality rather than productnot published
Rewarded video and offerwall networks (Unity, AppLovin)not publishedplus 4 points D7 and plus 2 points D30 versus non-rewarded; offerwall converters two to seven times organic retention, with early retention inflated by the incentive and D30 payer quality often laggingnot published
Genre-matched paid inventory (social casino example)not publishedD1 above 50% against roughly 20% organic, an inversion of the usual pattern created by high genre intent in the placementnot published
Pooled all-source median across the long tailD7 barely reaching 1.5%D7 3.42% to 3.94%D7 7% to 8% at the top quartile
03

When it looks bad

The blended curve looks acceptable while the per-source curves fan out, with one paid channel sitting far below the organic line at D7 and the gap widening rather than closing by D30.

Blended D7 of 12%, made up of organic at 19%, paid social at 14%, and a rewarded network that starts at 21% on D1 and collapses to 6% by D7 and 1% by D30.

04

What to do about it

  • Split every retention chart by media source before reading it and hold paid sources to within 10% to 15% of the organic baseline (AppAgent). A blended D7 that clears target while one source sits at half the organic rate is a budget reallocation decision, not a product decision.
  • Treat incentivised traffic as its own cohort with a D30 payer gate rather than an early retention gate. Rewarded and offerwall installs lift D1 and D7 mechanically, with offerwall converters at two to seven times retention (Unity), and can still underdeliver on payer quality by D30.
  • Restructure campaigns toward 100 to 150 installs per campaign per day. SKAN and AdAttributionKit suppress postbacks below roughly 20 daily installs, so thin campaigns return modelled retention that cannot support a channel-level decision.
  • Fix the interval that is actually broken rather than applying one retention programme. A weak D1 is an onboarding and creative-expectation problem, a weak D7 is a habit-loop problem, and a weak D30 is a content depth and live ops problem, and the three do not respond to the same lever.
05

Sources

  1. GameAnalytics2025 Mobile Gaming Benchmarks · 2025 · n=11,600 gaming applications, 1.48B monthly active users, 9 regions, 2 platforms, 16 genres, 2024 calendar year, classic retention Top quartile D1 of 26.48% to 27.69% and bottom quartile 10% to 11.5%. Median D7 of 3.42% to 3.94% with top quartile 7% to 8%. 75% of titles below 3% at D28. On iOS the top quartile D1 is 31% to 33% against 25% to 27% on Android. gameanalytics.com ↗
  2. AdjustGaming App Insights Report, 2026 Edition · 2026 · top 5,000 apps plus full Adjust dataset, January 2024 to January 2026, 23 genres, 6 regions, 21 countries D1 retention across all genres was 27% in 2025, with hypercasual and hybrid casual highest and family games up five points to 23%. Japan has the highest country-level D1 at around 25%. Median paid-to-organic install ratio rose 61% to 3.33, and casino reached 11.05. adjust.com ↗
  3. AppAgentMobile Game Retention Benchmarks: Know Your Numbers · 2026 · practitioner reference ranges drawn from commercially scaled games the agency runs and analyses; Tier 2 operator dataset Working ranges of D1 25% to 30% average, 35% to 40% good, 45% or above top tier; D7 10% to 15%, 18% to 22%, 25% or above; D30 3% to 6%, 8% to 10%, 12% or above. A healthy paid source should retain within roughly 10% to 15% of the organic baseline. appagent.com ↗
  4. Sensor Tower (via AppAgent and GameDev Reports)State of Mobile 2026 retention benchmarks · 2026 · top 25 games by revenue per segment, December 2025 snapshot Top hypercasual D1 of 38% to 40%. Midcore at 44% to 45% D1, 20% to 21% D7 and 11% to 12% D30. Top 25 casual games averaged 14.9% D7 in December 2025. Top 25 midcore D365 retention of 4% to 5% and declining. sensortower.com ↗
  5. Unity (via AppAgent)Unity rewarded format and offerwall retention data · 2026 · Unity advertising and mediation network data, sample size not disclosed in the secondary source Games using rewarded formats saw a 4 percentage point lift in D7 retention and a 2 point lift in D30. Users who convert via an offerwall show two to seven times higher retention than non-offerwall users, though incentivised installs can inflate early retention while underdelivering on D30 and payer quality. appagent.com ↗
  6. AppsFlyer (via Mistplay and MAF summaries)The 2024 State of Gaming App Marketing, retention section · 2024 · $29B in gaming install ad spend, genre and OS splits Owned media remarketing through SMS, email, user invites and cross-promotion is materially more effective for retention than paid media. Match games retain best across all retention windows, iOS retains better than Android, and Japan is the highest-retention country. appsflyer.com ↗
  7. StashGame Retention glossary and benchmark notes · 2025 · vendor reference, no sample size published; Tier 3, corroboration only Social casino titles commonly show D1 above 50% for paid cohorts against roughly 20% for organic, an inversion of the usual pattern driven by genre-intent targeting in paid inventory. stash.gg ↗
  8. MAFRewarded Ads Unpacked: Performance Insights · 2026 · internal analysis of over 500 mobile games plus Unity and ironSource data Simulation games average 2.4% D7 and 0.50% D30, puzzle 4.5% D7 and 1.20% D30. Games running an offerwall plus a rewarded campaign show materially higher retention across the D7 to D30 stretch. maf.ad ↗
  9. Sensor TowerState of Mobile 2026 · 2026 · iOS App Store and Google Play, 2014 to 2025 Global app downloads grew just 0.8% in 2025 while IAP revenue rose 10.6%, so cohorts are harder to replace and retention carries more of the growth burden than acquisition does. sensortower.com ↗
  10. AdjustMobile App Trends 2026 · 2026 · Adjust dataset across verticals and regions, 2025 calendar year Gaming installs and sessions remained broadly stable in 2025 with sessions up 1% year on year, and casual titles saw a 15% increase in session lengths to 25.92 minutes. adjust.com ↗