Installs are up, sessions are up, and paid acquisition is doing more of the lifting than at any point since ATT. Below is a category-by-category read of the 2026 benchmark reports from Adjust, RevenueCat, and AppsFlyer, split into two halves: what it costs to earn an install, and what happens in the hours and days after it. We then add what we see across apps running on Pantheon. The short version is that the gap between apps that grow and apps that shrink is decided less at the ad auction and more in the first session.
The market in one paragraph
2025 closed with 112.1 billion app downloads and consumer spend up 10.6% to $167 billion. Global installs grew 10% year-over-year and sessions 7%, both faster than in 2024 (Adjust). On the subscription side, RevenueCat now counts almost 15,000 new subscription apps launching every month, up from roughly 2,000 three years ago. Demand has not grown 7x. The result is a winner-take-more market: the top 25% of subscription apps grew revenue 80% YoY while the bottom 25% shrank 33%. The median app grew MRR just 5.3%.
Part 1 — Pre-install: what an install costs in 2026
| Vertical | Installs YoY | Sessions YoY | Global CPI (2025) | CPI YoY | Paid/organic ratio (2024 → 2025) | Ad partners per app |
|---|---|---|---|---|---|---|
| E-commerce | −10% | +5% | $0.98 | −2% | 0.48 → 0.54 | 6.3 |
| Finance | −4% | +21% | $1.13 | −25% | → 1.13 | 5.7 → 5.8 |
Source: Adjust, Mobile App Trends 2026 (data Jan 2024–Jan 2026).
Two patterns run across both verticals. Paid share of installs went up, and the number of ad partners per app barely moved. Budgets are concentrating on channels that can prove lifetime value, not spreading to find volume. Installs fell while sessions rose in both cases, which is the same story from the other side: fewer, better-selected users doing more.
The sub-vertical detail is where the real dispersion lives:
- E-commerce: deal-discovery apps saw CPI jump from $1.44 to $2.26 while marketplace ($0.88) and shopping ($0.91) got slightly cheaper. Regionally, North America ($2.49) and Europe ($2.25, up from $1.83) are the expensive markets; APAC nearly halved to $0.68. Paid share grew most in MENA (0.54 → 0.63), APAC (0.43 → 0.49), and North America (0.38 → 0.42).
- Finance: the only vertical where CPI fell across every region. Crypto dropped from $5.17 to $2.90, banking fell 18% to $2.09, but payments rose 29% to $1.44. Europe ($7.37 → $4.75) and North America ($7.03 → $4.13) remain the priciest. Cheaper installs pushed finance apps further into paid: LATAM's paid/organic ratio hit 3.26, APAC 1.41, and stock-trading apps nearly doubled theirs (0.34 → 0.63).
iOS opt-in is slowly improving. ATT opt-in rates rose from 35% to 38% industry-wide between Q1 2025 and Q1 2026. Entertainment, finance, lifestyle, and social all improved; publications made the biggest jump (18% → 26%). E-commerce is the only major vertical that slipped (35% → 34%).
Subscription UA has moved to Android and emerging markets. AppsFlyer's dataset of 2,900 subscription apps, 1.7 billion paid installs, and $2.1 billion in UA spend shows overall subscription app user acquisition spend grew 24% year-over-year, with Android growing at four times the rate of iOS. The Indian Subcontinent accounted for 49% of net Android paid install growth and LATAM a further 18%, while North America was essentially flat. Android paid installs grew 57% year-on-year while iOS organic installs fell 8%; Android UA spend grew 42% versus 10% on iOS. For the first time, paid installs make up the majority of all Android installs. The standout category is Short Drama, with paid installs up 155% year-over-year and over 60% of all installs now coming from paid channels.
Part 2 — Post-install: what happens after the tap
Retention and engagement
| Vertical | D1 | D7 | D30 | Average session length | Day-0 sessions per user |
|---|---|---|---|---|---|
| E-commerce | 12.6% | 6% | 3% | 9.6 min | 1.36 |
| Finance | 12% | 6% | 2% | 7.2 min | 1.48 |
Source: Adjust, Mobile App Trends 2026, global 2025 data.
Retention slipped slightly year-over-year in both verticals: e-commerce D1 went from 13.2% to 12.6%, finance from 13% to 12%, and finance D30 fell from 3% to 2%. Whatever improvements teams made in 2025 went into acquisition efficiency, not into keeping users.
The subscription funnel by category
RevenueCat's State of Subscription Apps 2026 (115,000+ apps, $16B in revenue) adds the layer that matters for subscription businesses: what share of installs actually pay.
| Category | D30 download-to-trial (median) | Trial-to-paid (median) | D35 download-to-paid (median) | D60 revenue per install (median) |
|---|---|---|---|---|
| Health & Fitness | 6.9% | 37.7% | 2.9% | $0.66 |
| Business | 9.1% | — | 2.6% | $0.50 |
| All categories | ~5–6% (non-gaming) | — | 2.0% | $0.34 |
Source: RevenueCat, State of Subscription Apps 2026.
Other category reads from the same report:
- Travel has the best trial-to-paid rate of any category (43.5%) but one of the lowest trial-start rates (4.1%). The users who try, buy.
- Education and Utilities both sit at 6.5% download-to-trial. Education monetizes early ($0.30 revenue per install by D14) but has the slowest Day-0 conversion of any category (28.5%), largely because 7-day trials are standard.
- Productivity shows the fastest Day-0 conversion: 71.9% of paid conversions happen the day of install.
- Photo & Video has the lowest trial-to-paid median at 22.2%, despite 61% of apps in the category now being AI-powered.
- Shopping converts 1.3% of downloads to paid by D35.
- Media & Entertainment shows the widest spread in the dataset: trial-to-paid ranges from 11.5% at the bottom quartile to 69.5% at the top decile.
AppsFlyer's paid-install view adds a nuance for UA teams: Education and Lifestyle convert over 40% of trialists, suggesting those apps deliver enough value within the trial to justify the cost. Health & Fitness consumers are the most decisive, with many paying upfront without a trial, particularly in Western Europe.
Day 0 decides most of it
The single most consistent finding across the reports is how compressed the decision window has become:
- 80–90% of trial starts happen on Day 0 (Business 89.9%, Health & Fitness 82.1%, Productivity 78%). After Day 3, trial starts drop below 5% in every category.
- 50.6% of all paid conversions happen on Day 0. More than 60% happen within the first week.
- 55.4% of 3-day trial cancellations happen on Day 0. 84% happen by Day 1.
Structural choices set the ceiling
A few decisions made before the first ad runs move the numbers more than most optimization work:
- Access model: hard paywalls convert 10.7% of downloads to paid by D35, freemium 2.1%. Top-decile hard-paywall apps reach 38.7%.
- Trial length: 17–32 day trials convert at 42.5%, trials of four days or less at 25.5%. Nearly half of all apps now use four-day-or-shorter trials anyway.
- Price point: high-priced apps convert downloads to trials at nearly 2x the rate of low-priced apps (8.9% vs. 4.4%) and to paid at 2x (2.8% vs. 1.4%).
- AI: AI-powered apps start trials at 8.5% vs. 5.6% and generate 41% more revenue per payer, but churn 30% faster.
- Geography: North America converts 2.8% of downloads to paid by D35, APAC 2.4%, India and Southeast Asia 0.7%.
Part 3 — Reading both halves together
Put the two halves side by side and the shape of the 2026 market is clear. Paid is a larger share of installs than a year ago. CPI is rising in the segments where competition is fiercest. Retention is flat to slightly down. And the window in which a user decides to pay is measured in hours, not weeks.
Take the medians literally for a moment (illustrative only; these come from different datasets): 1,000 installs at the finance CPI of $1.13 cost $1,130. At the global 2.0% D35 download-to-paid median, that buys 20 payers. At $0.34 D60 revenue per install, those installs have returned $340 by day 60. The median app is nowhere near payback at D60, which is why the median app grows 5%. The top quartile runs the same auction with a different funnel.
That points to a specific conclusion: every efficiency gain in the ad auction gets multiplied, or erased, by what happens in the first session. Two levers pull on both halves at once.
Part 4 — Two levers that move the whole funnel
Lever 1: test more creatives than feels comfortable
Creative is the largest controllable input to CPI, and it also determines who installs, which sets the ceiling on everything after. Adjust's analysis with Alison.Ai shows how far "one creative for everyone" is from optimal. The same elements perform differently by vertical and by operating system: on iOS, sound-off design lifted performance 24% and free-offer text shown early lifted it 30%; on Android, split-screen scenes lifted it 51% and an end card without a CTA lifted it 31%.
Same concept, different cut per OS, per market, per placement. Multiply that by the number of concepts you need to keep a challenger live against every winner, and the required creative volume is far above what most teams produce by hand. RevenueCat's report adds one more multiplier: Rise found that creatives which flop in app-install campaigns often win on web funnels, and vice versa, so running concepts across both channels roughly doubles the chance of finding a hit.
Generative AI has made this volume affordable, but the reports are blunt about execution. Adjust notes that GenAI is now judged on outcomes rather than novelty, and that the small group of organizations with disciplined deployment report efficiency gains above 20%, while most are still piloting. Access to the tools is no longer the edge. The testing cadence is.
Lever 2: make onboarding a continuation of the ad
Look again at the Day-0 numbers. A user tapped a specific promise, delivered in a specific color palette, with a specific character, hook, and wording. The moment they land in a generic onboarding flow, that thread breaks, at exactly the point where more than half of all paid decisions are made.
Creative-matched deep-link onboarding closes that gap. The mechanics are simple: the creative ID travels with the install through a deferred deep link; the first onboarding screens pick up the creative's colors, visual style, and headline; the opening screen restates the promise the user tapped; and the paywall shows the plan or offer the ad implied. The user never feels the handoff between ad and app.
Pantheon Findings: what creative-to-onboarding alignment does to install-to-subscription
What we measured. Across apps running on Pantheon that aligned their onboarding flow with the ad creative that drove the install — carrying the creative's colors, visual design, and headline through a deferred deep link into the first screens and the paywall — we compared install-to-subscription rate against each app's generic onboarding baseline.
What we found.
| Average lift across apps that implemented the practice | Best case, with the tightest creative-to-onboarding alignment |
|---|---|
| +30% | +120% |
What that means at benchmark rates. At the global 2.0% D35 download-to-paid median, a 30% lift moves an app to 2.6% of installs paying; a 120% lift moves it to 4.4%, which is above the top-quartile threshold for most categories in RevenueCat's data. Same CPI, same media budget: 30% to 120% more subscribers, and the cost per subscriber falls by the same proportion.
The spread between the average and the best case is wide because alignment is a spectrum. Matching a color palette is the floor. Matching the headline, the first-screen promise, and the offer shown on the paywall is where the upper end of the range lives.
Three things explain the lift. Continuity builds trust before the paywall appears; the expert commentary in RevenueCat's report makes the same point, that a paywall shown before context is established feels jarring, and onboarding that builds momentum first converts very differently. Second, ad platforms get a cleaner signal, because the users who convert are the users the creative selected for. Third, it turns creative testing into onboarding testing: every winning concept ships with its own onboarding variant, so the learning compounds instead of stopping at the install.
A loop you can run this quarter
- Generate net-new concepts every week, enough that each winner always has a live challenger.
- Cut each concept per OS and per market, not just per language.
- Measure hook rate, CTR, and CPI per creative, and install-to-trial and install-to-paid per creative. Kill on the second set, not the first.
- For every winner, build a matched onboarding variant behind the deep link: same colors, same headline, same offer.
- Scale the pair, not the ad.
Bottom line
The 2026 benchmarks say the same thing from three directions. Installs are more expensive to earn where it matters, harder to keep, and decided faster than ever. The apps in the top quartile are not winning the auction; they are winning the first session, and they get there by testing more creative and by refusing to let the ad and the app tell different stories. Our own data puts a number on the second half: 30% more subscribers on average, and up to 120% when the creative and the onboarding are the same story.
That loop is what Pantheon is built around: take a reference ad, generate the variations and localizations needed to test at volume, and carry each winning creative straight through to a matched onboarding flow.
Frequently asked questions
What is a good cost per install for mobile apps in 2026?
A useful CPI benchmark matches your app category, region and operating system. Adjust’s 2026 report gives global 2025 averages of $0.98 for e-commerce and $1.13 for finance, while North American averages were $2.49 and $4.13 respectively. These are reference points, not profitability targets. Compare CPI with conversion and revenue after the install. Source: Adjust.
What is a good install-to-paid conversion rate for a subscription app?
RevenueCat’s 2026 report gives a median D35 download-to-paid rate of 2.0% across categories, compared with 2.9% for Health & Fitness and 2.6% for Business. Compare your app with its category, access model and geography using the same 35-day window. A category median is a reference point rather than a universal target. Source: RevenueCat.
What is a good free-trial-to-paid conversion rate for mobile apps?
RevenueCat’s 2026 category medians range from 22.2% for Photo & Video to 37.7% for Health & Fitness and 43.5% for Travel. Use a comparable category and trial length, and measure trials that have had time to finish. Trial-to-paid measures paying trialists, not the percentage of all downloads that pay. Source: RevenueCat.
What is a good mobile app retention rate?
Retention benchmarks depend on category and the day measured. Adjust’s global 2025 data in its 2026 report shows e-commerce retention of 12.6% on Day 1, 6% on Day 7 and 3% on Day 30. Finance records 12%, 6% and 2%. Compare equivalent cohorts and acquisition sources rather than treating one retention rate as a target for every app. Source: Adjust.
How do I calculate customer acquisition cost for a subscription app?
For media cost per paying subscriber, divide CPI by the install-to-paid conversion rate expressed as a decimal. For example, a $2 CPI divided by 0.02 equals $100 per paying subscriber. Use spend and conversions from a consistent cohort and conversion window. This media-only calculation excludes production and other acquisition costs; compare it with subscriber revenue and retention.