Mobile App Retention Benchmarks 2026

§Mobile app retention in 2026: fintech collapsed, subscription rewrote the rules

The widely-cited "fintech retains best" narrative is stale. Adjust Mobile App Trends 2026 reports finance vertical D30 at 2%, down from 3% the year prior: a 33% YoY drop, the steepest in any tracked category (Adjust 2026). The 30/17.6/11.6 D1/D7/D30 figure still circulating across SEO listicles in 2026 is the AppsFlyer 2023 subscription-cohort number, recycled for three years. It was an apples-to-oranges comparison then; in 2026 it's wrong.

What is true in 2026: business model dominates category. AppsFlyer's State of Subscriptions 2026 reports cross-category D30 of 14% for subscription apps versus 5.4% for ad-supported, a 2.5x premium (AppsFlyer). A subscription gaming app retains better than an ad-supported fintech app. The business model is the benchmark; the vertical is a tiebreaker.

Long tail is shortening. Sensor Tower's State of Mobile 2026 puts top-25 midcore games at 4-5% D365 retention, and declining (Sensor Tower). For the median game, the number is materially lower and decays faster. Most public reports stop at D30, which hides where the user economy actually breaks.

The LTV equation is unchanged; the inputs moved. The calculator above models the new shape: pick category, platform, business model. See where your D30 falls in the segment distribution (P10 to P95). See your retention curve plotted against segment median and top quartile from D1 through D365.

§What good actually looks like in 2026

Cross-category D1 / D7 / D30 medians from Adjust 2026 + AppsFlyer + Sensor Tower + UXCam aggregator:

Two structural patterns stand out.

Subscription verticals (Streaming Video, parts of Productivity) plateau, not decay. Notice OTT D30 = 14, the same as D7. That is the signature of a contract-based business model: users who survive the first month largely keep the subscription. Ad-supported categories drop ~30-50% between D7 and D30.

Habit-trigger categories outperform their nominal Day 1. News & Media D1 is 26 (middle of pack) but D30 is 9, the highest of any non-subscription vertical. The pattern: external events (news cycles, social proof, scheduled releases) pull users back regardless of product quality.

§The platform premium is real but smaller than people claim

Adjust 2026 aggregate iOS vs Android D30 ratio is 1.33x, not the 1.5-2x figure recycled from older AppsFlyer subscription cuts (Adjust 2026 via UXCam aggregator).

The premium grows with day index because iOS users skew higher commitment throughout the cohort, not because Day 1 sees a different population. Per-vertical iOS/Android splits require AppsFlyer's gated benchmark portal. The calculator models them by applying the aggregate 1.33x premium to the category baseline, which is roughly correct for non-subscription verticals and conservative for subscription (where the gap is wider).

§Business model determines the curve shape, not category

AppsFlyer State of Subscriptions 2026 reports subscription apps retain 2.5x cross-category median at D30: 14% vs 5.4% baseline. The premium is concentrated between D7 and D30 because that's where subscription commitment manifests behaviorally (recurring charge after trial, opt-out window closing).

Working approximation used in the calculator, anchored to AppsFlyer 2026 data:

• Ad-supported (free): D1 ×1.00, D7 ×1.00, D30 ×1.00. Baseline.
• Freemium IAP: D1 ×1.05, D7 ×1.15, D30 ×1.40. Modest commitment via IAP funnel.
• Subscription: D1 ×1.05, D7 ×1.35, D30 ×2.50. The big premium.
• Paid upfront: D1 ×1.10, D7 ×1.25, D30 ×1.55. Selection bias of paying customer.

These multipliers stack on top of the category baseline. A Gaming Subscription app at iOS expects D30 = 5 × 2.50 × 1.20 = roughly 15%, close to the OTT subscription benchmark. The math compounds the same way for any category × platform × model combination the calculator runs.

§The long tail: modeled D60-D365, because Tier 1 stops at D30

No Tier 1 source publishes D60-D365 medians by vertical. Sensor Tower State of Mobile 2026 publishes a single long-tail anchor: top-25 midcore games at 4-5% D365 (Sensor Tower). Industry consensus heuristics fit a smooth decay from there.

The calculator models the long tail from your D30 input:

• D60 = 0.75 × D30 (the immediate post-month-1 stabilization phase)
• D90 = 0.60 × D30 (long-time-horizon return rate)
• D180 = 0.40 × D30 (the half-year mark; most users gone)
• D365 = 0.25 × D30 (Sensor Tower midcore-anchored decay rate)

These factors are modeled, not measured. They are explicitly heuristic. If your specific app has six months of cohort data, plug your actual numbers in and the calculator stops extrapolating. If you don't, the model is closer to reality than the alternatives (Tier 1 reports leaving you to invent the curve, or industry articles that quote D30 once and stop).

The most useful read: the D365 modeled number is what determines whether you have a real product or a leaky bucket. A Gaming app with D30 = 5% and D365 = 1.25% (modeled) is fundamentally a single-month acquisition funnel. A Subscription Streaming app with D30 = 14% and D365 = 3.5% (modeled) is still a leaky bucket but is shipping real long-tail revenue.

§The $1.24 global CPI is the most misleading number in mobile

The 2026 BusinessOfApps CPI Research report puts the cross-segment global average at $1.24 (BusinessOfApps). That number is technically correct and operationally useless.

It hides a roughly 50x spread between top and bottom of the segment distribution. Same report publishes the underlying bands:

• Hyper-casual gaming, Tier 3 geos (India, Indonesia, MENA), Android: $0.20-0.40 typical CPI.
• North America Android, mid-vertical aggregate: $1.20-3.40.
• North America iOS, mid-vertical aggregate: $2.50-5.00.
• High-CAC finance verticals (banking, brokerage, BNPL) on iOS in NA/EU: $8-15+.

The $1.24 mean sits in a distribution where one app's CPI is fifty times another's, and the actual unit-economics question is whether your CPI in your segment beats your segment LTV. Benchmarking against the global mean returns "you are within range of normal" when you are bleeding cash relative to your peer set.

The right read of the $1.24 number is as evidence that no single CPI benchmark applies. It exists because Android Tier 3 gaming installs at $0.20 numerically dominate the install-weighted average even though their LTV is sub-dollar. Removing those installs from the dataset would raise the global mean to $3-4. The cheap-install majority is dragging the global mean down to where it looks reachable for any app, and that is the deception.

iOS CPI being 2-3x Android is not about iOS being worse value. Advertisers bid higher because iOS users monetize better. The right question is not "which platform pays more" but "which platform × business model combination fits your retention curve at a healthy breakeven CPI."

Breakeven CPI in this calculator = 90-day LTV ÷ 3, computed against your segment's actual retention curve. The 3 covers infrastructure drag plus reasonable acquisition margin. Single-purchase model targets 1.5; subscription with low churn can run 2:1. The output is segment-specific because the inputs are segment-specific.

The mistake most apps make is to optimize CPI down to a number smaller than $1.24, declare victory, and not notice they are still above breakeven for their segment. Lowering CPI 20% on a segment where your breakeven is half your current CPI fixes nothing. Retention is what determines the breakeven; CPI optimization is a smaller lever further down the chain.

§How to read the calculator

The calculator runs three selectors plus five inputs, then produces a percentile placement plus LTV math at 90 and 365 days.

Selectors:

• Category. 13 verticals with Tier 1 + composite + flagged-estimate sourcing. Source confidence shown per row above.
• Platform. Cross-platform / iOS / Android. Multipliers apply on top of category baseline.
• Business model. Ad-supported / Freemium IAP / Subscription / Paid upfront. The biggest determinant of D30+.

Inputs:

• CPI ($). Your acquired install cost from ad reports.
• ARPU/day ($). Daily revenue per active user from your analytics.
• D1 / D7 / D30 retention (%). From your cohort view. Defaults seeded to category medians.

Results:

• Hero. Your D30 percentile in the segment (P10 to P95). Top quartile starts at P75.
• Stats. LTV:CPI ratio, 90-day LTV, 365-day modeled LTV, D1 grade band.
• Line chart. Your D1-D365 curve vs segment median vs top 25%. The gap at each day is your fix surface.
• Bar chart. D30 median across all 13 categories at your platform + business model. See which categories beat yours structurally.

Compare mode (button above the calculator): side-by-side two segments or two app configs. Useful for "my iOS Subscription Gaming vs my Android Freemium Gaming" decisions.

For first-session retention diagnosis, three structural causes recur in post-mortems:

• FTUE over 60 seconds. Pre-screens, registration walls, permission prompts gating value.
• No Day 1 push trigger tied to onboarding intent. Generic blasts don't count.
• Wrong-source users. Incentivized networks deliver cheap installs and terrible cohort retention. Segment retention by acquisition source via cohort analysis.

The 90-day LTV is approximated by user-days across three retention buckets: D1 covers days 2-7, D7 covers days 8-30, D30 covers days 31-90. Multiplied by daily ARPU.