Mobile App Acquisition and Retention
Review store-view-to-install conversion, Apple Ads costs, retention, permission consent, paywalls, and new-app revenue distributions with explicit denominators.
Written and reviewed by Eric Han · Co-founder & CTO, reopt
Key takeaways
- Store “conversion” has two denominators. The same US App Store dataset reports 3.8% per impression and 8.56% per page view; an earlier page-view figure was 25%. Never use a value without its denominator.
- Median US Apple Ads search CPI ranges from $1.33 for shopping to $26.70 for sports. Startups with little branded search are better represented by Generic at $5.31 and Competitor at $6.78.
- US app retention medians, for apps with at least 1,000 D30 downloads, are D1 27.3%, D7 9.2%, D30 3.9% (method undisclosed, D). Amplitude's 7% is a top-quartile threshold for products across industries, a different population.
- ATT consent is 38% among users shown the prompt in Q1 2026, versus 13.85% among first-launch devices in Q2 2024.
- New subscription apps' median monthly revenue one year after launch is about $72; 17.3% reach $1K monthly within two years. These apps have recorded subscription revenue, likely making the figures higher than for all new apps (author judgment).
The Full Impression-to-Paid Funnel Is Not Connected
We found no large public dataset connecting every stage. Within-source segments include:
| Source | Connected stages | Figures |
|---|---|---|
| AppTweak, Korean Apple Ads | Impression → tap → install | 6.30% × 51% ≈ 3.2%, calculated; CPI $1.84 |
| SplitMetrics | Impression → tap → download | 9.7% × 66.2% ≈ 6.4%, calculated; CPA $3.76 |
| Adapty State of Paywalls | Paywall view → paid | US 2.75%; trial → paid 26.2% |
| RevenueCat | Download → trial → paid | See App Subscription Conversion |
Products of means or medians are approximations. Download-to-paywall-view rates are undisclosed, leaving a gap between the segments.
Store View to Install
We found no overall benchmark directly published by Apple or Google. Apple provides peer-group percentiles inside App Store Connect.
| Value | Denominator | Scope | Evidence |
|---|---|---|---|
| 3.8% | First downloads ÷ impressions | US App Store overall, 2025 | AppTweak, C |
| 8.56% | First downloads ÷ page views | Same | Same source, C |
| 25% | Downloads ÷ page views | First half 2024, secondary citation | Adapty, D |
Do not use AppTweak's 3.8% or 8.56% store conversion as inputs
The 2025 value is about one-third of the first-half-2024 figure, and traces of older values remain on the page, suggesting changed aggregation (author judgment). Do not interpret this as an actual decline.
AppFollow's public table reports US App Store category medians for August 2026 (C). Search and Browse use impressions, while Referrer paths use product-page views; do not compare those groups.
| Path | Install conversion | Denominator |
|---|---|---|
| Search | 2.6–12.4% | Impressions |
| Browse | 0.1–2.5% | Impressions |
| App Referrer | 4.5–53.0% | Product-page views |
| Web Referrer | 4.0–75.1% | Product-page views |
About 65% of App Store downloads occur directly after search (Apple Ads, 2022, platform figure).
We found no average store A/B effect or win rate. The 5–50% in Google's Tapps case describes selected successes; the +3% and +8% in Apple help are illustrative examples.
Apple Ads Costs
- Sports$26.70
- Finance$9.65
- Business$6.86
- Entertainment$4.99
- Games$4.63
- Health / fitness$3.77
- Travel$3.70
- Productivity$3.58
- Education$2.91
- Photo / video$2.79
- Lifestyle$2.62
- Utilities$2.25
- Books$1.53
- Social$1.45
- Shopping$1.33
| Campaign type, US apps | Median CPI |
|---|---|
| Brand | $2.84 |
| Discovery | $3.10 |
| Generic | $5.31 |
| Competitor | $6.78 |
Tap-to-install medians by placement are 56% for search results, 15% for Search Tab / Product Pages, and 7.7% for Today Tab (about 3,500 apps, 38 countries). The high rates of 56–66% apply only to people who have already tapped an ad, and branded campaigns push them upward. Generic/Competitor figures are more relevant to startups with little branded demand (author judgment). One solo habit app's observed CPI around $8.85 points in the same direction but is only one case.
Custom product pages show 58.7% versus 55.5%, a self-selected comparison. We found no official Google App Campaign or Meta app-ad benchmarks.
Post-Install Retention
| Scope | D1 | D7 | D30 | Evidence |
|---|---|---|---|---|
| US apps, at least 1,000 D30 downloads | 27.3% | 9.2% | 3.9% | MWM, Q3 2025, methodology undisclosed, D |
| Social | 31.9% | 12.3% | 5.9% | Same source |
| Lifestyle | 23.6% | 9.6% | 4.8% | Same source |
| Productivity | 23.0% | 8.9% | 4.5% | Same source |
| Education | 24.9% | 8.6% | 3.6% | Same source |
| Games | 36.6% | 9.3% | 2.9% | Same source |
| Mobile games overall | About 22% | About 4% | About 0.7% | GameAnalytics 2026 summary, D |
For games, MWM's D7/D30 values of 9.3%/2.9% differ from GameAnalytics' roughly 4%/0.7%. Use only for orders of magnitude.
- Amplitude's 7% rule is a top-quartile threshold across more than 2,600 companies, not evidence that 7% causes growth. The median product also loses more than 98% of new users by week two.
- GameAnalytics' 2025 report, using 2024 data (C), gives median game D7 retention of 3.42–3.94%, with the upper quartile at 7–8%. A tool widely used by indie and early teams may be more representative of new studios (author judgment).
Do not rely on MWM's country table
D1 ranges only from 27.1% to 27.5% across eight countries. Compared with GameAnalytics' regional differences, this is too flat to support “Korea is the same as the US” (author judgment).
Permission Consent
| Metric | Value | Denominator | Evidence |
|---|---|---|---|
| ATT consent | 38%, games 39%, Q1 2026 | Users shown the prompt | Adjust press release, C |
| ATT-authorized devices | 13.85%, Q2 2024 | First-launch devices | Singular, C |
| Median push opt-in | Android 53.3%, iOS 48.85% | Monthly average across all app users, not first-prompt acceptance | Airship 2026, C |
We found no primary onboarding-stage abandonment benchmark. Vendor blog figures conflict.
Paywalls and New-App Outcomes
Adapty State of Paywalls covers App Store data from more than 30,000 apps, October 2025–September 2026 (paywall vendor, B). Median 30-day view-to-paid conversion is 2.75% in the US and 1.38% in Europe. Of trials, 32% cancel within one hour and 44% within 24 hours.
New-app distributions from RevenueCat 2026, grade B:
| Metric | Value |
|---|---|
| Median monthly revenue one year after launch | About $72; top 25% $429; top 10% $2,574 |
| Reaching $1K monthly within two years | 17.3%; Business 14.7% |
| Reaching $10K monthly within two years | 4.6%; Business 1.6% |
| Median time to milestone | $1K: 58 days; $10K: 109 days, only among apps reaching it |
| Median revenue per install, D14 and D60 | $0.23, $0.34 |
| Revenue share by launch period | Before 2020: 69%; 2025 onward: 3% |
Comparing $0.34 revenue per install with CPI above illustrates how difficult it can be to recover paid acquisition costs. Because sources differ, use it only as an order-of-magnitude check.
We found no recent direct measurement of “most apps earn no revenue” across all apps. Sensor Tower's Q3 2019 top-1% concentration estimate—$20.5 billion of $22 billion—uses publishers, not apps.
One public example, solo habit app Habit Doom, reports 341 first downloads and $59 revenue in 90 days. It is a self-disclosed case, not a distribution.
Randomized Experiments
| Experiment | Context | Result | Grade |
|---|---|---|---|
| Cao et al., JMR | About two million existing free users of a Chinese education app | Hard conversion increased subscriptions 124% and revenue 102%; not new installs | A (summary) |
| Yancey and Settles, KDD 2020 | Duolingo push-copy optimization, two weeks | New-user D7 +2.0%, DAU +0.5%; company-affiliated authors, relative gains at a large app | A |
We found no peer-reviewed randomized study of onboarding or pricing in newly launched apps.
Korea
Few usable figures are public.
| Metric | Value | Evidence |
|---|---|---|
| Apple Ads tap rate; tap → install | 6.30%, 51% | AppTweak, C |
| Apple Ads CPI; CPM | $1.84, $48.11 | Same source, C |
| Game D1; IPM | 20%, 6.13 | Adjust coverage, C via press |
Millie's Library filings report free-subscription-to-paid conversion of 34.7–37.3% (Korean Cases, B). We found no recent Korean ATT or push-consent figures. In 2021, AppsFlyer reported ATT 37% during the two weeks after iOS 14.5 (C via press), while Singular reported about 24% in May (C). Mobile Index and Wiseapp public reports provide rankings and spending estimates, not conversion rates.
Simulation Inputs
| Stage | Input | Evidence and conditions |
|---|---|---|
| US CPI | Generic $5.31; Competitor $6.78; US search in the country table $4.06 | AppTweak, C; little branded search |
| Korean CPI | $1.84 | Same country table; includes brand campaigns. Different aggregation from US Generic; do not compare directly |
| Download → paid within 35 days | Mean 3.1%, | Fitted to RevenueCat median 2.0%, B; mid-price upper quartile used as a proxy |
| First-year revenue per paying customer | $23 overall, $32 North America | RevenueCat median first-year realized value per payer, B |
Do not use AppTweak's 3.8%/8.56% store conversion, Adapty's 25%, MWM's country table, or Amplitude's 7% as inputs.
Paid app-acquisition intervals
Choose CPI to estimate installs and paid conversion. CPI and conversion come from different sources, so this is an order-of-magnitude check. The conversion mean is raised so the distribution's median matches the source median.
8,000 runs. The 90% interval holds only rate uncertainty around the expected value; the actual count adds chance variation in counts. Vertical lines mark the 5th percentile, median (darker), and 95th. The last bin collects values above the 99th percentile.
Sales-Led Conversion
Review lead-to-opportunity-to-win rates, lead-source differences, and email, phone, and LinkedIn outreach metrics while excluding figures with unclear origins.
App Subscription Conversion
Review download-to-trial-to-paid conversion by category, region, and price, plus hard paywalls, AI-app retention, and Korean trial-rule effects using transaction logs.