Table of Contents
Mobile app marketing benchmarks compare acquisition cost, revenue and retention across defined cohorts. Use them to ask whether a result deserves investigation, then set targets from your own economics. The ranges below are not accompanied here by reporting dates, sample counts and a reproducible calculation method, so they should not be treated as verified 2026 market medians or spending targets. Check the source, cohort, market, category and measurement basis before using a range in a budget decision.
A broad category average can hide differences in product, monetization, platform and country mix. Start with your own mature cohorts, then look for an external comparison that matches them. A benchmark can help identify a result worth investigating; the app’s contribution value and cash constraints determine whether a change in spend is affordable.
What mobile app marketing benchmarks actually measure
Mobile app marketing benchmarks measure the unit economics of paid user acquisition against reference performance across comparable apps. The core metrics are CPI (the cost to acquire an install), ROAS at a specific time window (revenue generated against spend over D1, D7, D30, or D90), retention rates (percentage of users who return on a given day after install), ARPU and LTV (average revenue per user over time), and conversion rates through the funnel from install to registration to monetizing event.
Read CPI together with mature user value and the costs that value must cover. A $0.50 CPI and $0.80 ARPU do not by themselves establish a losing trade; the revenue horizon and contribution costs are missing. Likewise, a $12 CPI against $95 revenue LTV is not automatically profitable. Keep revenue per user, contribution value and collected cash distinct, then agree the payback period the business can support.
Check the data period before comparing years
Check a report’s data period before treating its publication year as the benchmark year. Adjust’s 2026 update states that its original 2026 edition covered data through December 2025 and that the update uses H1 2026 activity. That distinction matters when comparing markets or seasons. Privacy-framework versions, campaign measurement and production methods can also change the comparison, but they do not establish a universal CPI increase or a required tenfold rise in creative volume.
Revisit a benchmark when its market, measurement setup or observation period no longer matches your decision. An older report may still provide useful historical context, while a newer headline can use an earlier data period. Record the data cutoff and method rather than assuming every reference becomes invalid after ninety days.
The Admiral Media Benchmark Reality Check Framework
Use these five checks before comparing an app with a published range. Record any mismatch rather than treating a familiar category label as proof that the reference fits your acquisition plan.
A practical comparison checklist
- Match the sub-vertical, not just the category. Compare apps with a similar product, monetization model and maturity. A broad gaming average may combine very different businesses; the size of the difference needs evidence rather than an assumed multiple.
- Separate platform from product. Review iOS and Android results separately when measurement or user economics differ. Also match product, country, acquisition source and cohort age; platform alone does not make two samples comparable.
- Normalize for geography. Match the actual countries and currencies, rather than assuming a tier label represents the same market mix in every report. State how country results are weighted before combining them.
- Match the observation window and monetization model. Compare revenue at the same cohort age and document subscription, purchase and advertising revenue separately. Choose the decision horizon from the app’s actual payment and renewal behavior, not a universal day range for the category.
- Use benchmarks as comparison points, not automatic targets. A percentile needs a defined sample and a clear direction: lower cost can be desirable, while higher revenue or retention can be desirable. A position in either distribution does not replace a contribution-value and payback calculation.
Write down the matching criteria before setting a target or changing budget. If the sample or calculation method is missing, keep the range as an open question rather than making it the basis for a spending commitment.
CPI ranges by vertical: check the sample first
The CPI table presents category ranges for comparison, but does not supply a dated cohort and calculation receipt for each row. Confirm the countries, acquisition sources, app sample, currency and whether the range describes campaigns, app-level averages or another aggregation. Do not apply a fixed Android discount or market-tier discount without comparable evidence for the app and period.
| Vertical | Sub-vertical | iOS CPI range (Tier 1) | Android CPI range (Tier 1) | Primary driver |
|---|---|---|---|---|
| Gaming | Hypercasual | $0.80 to $2.50 | $0.30 to $1.20 | Volume plays, ad monetization |
| Gaming | Casual puzzle | $2.50 to $6.00 | $1.20 to $3.00 | D7 retention, ARPDAU |
| Gaming | Mid-core strategy | $8.00 to $18.00 | $4.00 to $10.00 | LTV from payers, cohort depth |
| Gaming | Slots and social casino | $20.00 to $60.00 | $10.00 to $30.00 | Whale LTV, payer concentration |
| Gaming | Idle RPG | $4.00 to $9.00 | $2.00 to $5.00 | D30 retention, creative variety |
| Finance | Neobank | $8.00 to $20.00 | $3.00 to $9.00 | KYC completion, funded accounts |
| Finance | Trading and crypto | $10.00 to $25.00 | $4.00 to $12.00 | Deposit rate, KYC, compliance |
| Finance | Payments | $2.00 to $6.00 | $0.80 to $2.50 | Transaction activation |
| Dating | Mainstream dating | $4.00 to $10.00 | $2.00 to $5.00 | Gender balance, D7 retention |
| Dating | Niche and premium dating | $8.00 to $20.00 | $3.00 to $10.00 | Subscription LTV, match quality |
| Health & Fitness | Fitness tracking | $2.00 to $6.00 | $1.00 to $3.00 | Subscription conversion, retention |
| Health & Fitness | Nutrition and fasting | $3.00 to $8.00 | $1.50 to $4.00 | D7 activation, onboarding |
| Health & Fitness | Meditation and mental wellness | $4.00 to $10.00 | $2.00 to $5.00 | Subscription conversion, LTV |
| EdTech | Language learning | $3.00 to $8.00 | $1.50 to $4.00 | Subscription conversion, D30 retention |
| EdTech | Brain training | $3.50 to $9.00 | $1.50 to $4.50 | Trial-to-paid, ROAS at D30 |
| EdTech | Kids and academic | $3.00 to $7.00 | $1.50 to $3.50 | Family subscription LTV |
| Shopping & eCommerce | Marketplace | $2.00 to $6.00 | $0.80 to $2.50 | First purchase, AOV |
| Shopping & eCommerce | DTC retailer | $3.00 to $8.00 | $1.50 to $4.00 | ROAS at D7, repeat rate |
| Travel | OTA and booking | $3.00 to $9.00 | $1.50 to $5.00 | First booking, margin per booking |
| Productivity & AI Tools | AI assistants and utilities | $2.00 to $6.00 | $0.80 to $2.50 | Subscription conversion, usage frequency |
Platform differences should be established within comparable cohorts, rather than assumed from the table. The Inshallah iOS growth case study reports nearly twice the retention and a 1,253% increase in US iOS revenue in that engagement. Those results do not establish the same relationship for another app. Compare acquisition cost with mature contribution value before labeling a platform profitable.
CPI benchmarks by platform: iOS vs Android
Compare iOS and Android on the same product outcome, market and cohort age, and document each platform’s attribution coverage. A higher CPI can be acceptable if the acquired cohort returns more contribution value within the chosen payback period. It is not justified by an assumed 2.2x revenue multiple or by the operating system alone.
ROAS ranges by vertical and cohort age
The ROAS ranges below are comparison values, not guaranteed routes to twelve-month payback. Set a return requirement from the app’s actual renewal or purchase curve, contribution costs and cash constraints. Label revenue ROAS separately from contribution return, and compare D7, D30 and D90 only for cohorts that have reached those ages.
| Vertical | Monetization | D7 ROAS range shown | D30 ROAS range shown | D90 ROAS range shown |
|---|---|---|---|---|
| Subscription apps (health, EdTech, productivity) | Weekly/monthly/annual subscriptions | 15% to 30% | 40% to 70% | 70% to 110% |
| Dating apps | Subscription + in-app purchase | 10% to 25% | 30% to 60% | 60% to 100% |
| Mid-core and strategy games | In-app purchases | 15% to 35% | 35% to 70% | 70% to 120% |
| Casual games | IAP + ads hybrid | 25% to 50% | 50% to 90% | 90% to 140% |
| Hypercasual games | Ad-based | 50% to 90% | 80% to 120% | 110% to 150% |
| Fintech (neobank, trading) | Transaction fees, trading revenue | 5% to 15% | 20% to 50% | 50% to 100% |
| eCommerce and DTC | Purchase margin | 30% to 80% | 80% to 150% | 150% to 300% |
Choose the cohort ages around the decision and the app’s observed revenue curve. D7 can support an early check, while later ages reveal payments and renewals that have not yet occurred. D90 revenue ROAS is still not a profit measure unless all relevant costs are included. Review early indicators against mature outcomes before using them to allocate budget.
Predictive ROAS for early signal
Predictive ROAS forecasts a later return from earlier observations. Validate the model on held-out mature cohorts, report its error and check whether performance changes by market and acquisition source. An early forecast is not observed revenue. A claimed D3 prediction accuracy needs a defined error measure and validation receipt before it can justify a scaling decision.
Retention ranges and measurement definitions
For retention, define the return event, denominator and whether the measure is exact-day or rolling retention. Match acquisition period, platform, market and cohort age. The table’s ranges do not include a sample receipt or a percentile calculation here, so they should not be treated as verified averages or top-quartile cutoffs. Investigate product experience and acquired-user mix together rather than assigning a fixed share of retention to marketing.
| Vertical | D1 retention | D7 retention | D30 retention | Higher D30 value shown |
|---|---|---|---|---|
| Marketplace (Amazon-style) | 30% to 38% | 14% to 20% | 7% to 12% | 12%+ |
| Banking and finance | 28% to 36% | 16% to 22% | 8% to 14% | 14%+ |
| News and media | 30% to 40% | 15% to 22% | 6% to 11% | 11%+ |
| Shopping and eCommerce | 22% to 28% | 9% to 14% | 4% to 7% | 7%+ |
| Dating (mainstream) | 22% to 30% | 10% to 16% | 4% to 8% | 8%+ |
| Health and fitness | 25% to 32% | 11% to 16% | 5% to 9% | 9%+ |
| EdTech (language, brain training) | 24% to 32% | 10% to 16% | 5% to 9% | 9%+ |
| Mid-core gaming | 25% to 35% | 10% to 16% | 4% to 8% | 8%+ |
| Hypercasual gaming | 20% to 28% | 6% to 11% | 2% to 4% | 4%+ |
| Productivity and AI tools | 24% to 32% | 10% to 16% | 5% to 10% | 10%+ |
Compare retention by acquisition source as well as product and cohort age. A change in the audience mix can coincide with a different retention curve, but that does not assign a fixed percentage of retention to marketing. On the PURE dating case study, Admiral Media reduced CPI by 74% while simultaneously exceeding D7 ROAS goals. Read that reported outcome in its own engagement; it does not establish a universal marketing contribution or prove that the cheapest install has the best downstream value.
Creative testing: distinct ideas and useful learning
Plan creative testing around distinct questions, available traffic and production capacity. Separate a new concept from a crop, translation or edit, and record total production and media cost. There is no universal monthly concept count, refresh deadline or testing-budget percentage that guarantees better acquisition. Test whether a new message improves the business outcome before turning an asset-volume goal into a scaling rule.
Count concepts and variants separately so an asset total does not stand in for the quality of the test. On the NeuroNation Google Creative Framework case study, Admiral Media executed a creative refresh strategy that reduced CPP by 34%, increased purchases by 129%, and scaled spend on video ads by 952% within a single campaign, with impressions climbing 3363% and clicks climbing 1215%. These are reported results from that campaign, not proof that an asset-volume target alone will produce the same outcome.
When more creative can answer a test question
Read creative volume alongside the quality of the ideas, the traffic available to compare them and the value of acquired users. More assets can produce more learning opportunities, but it can also fragment delivery or repeat the same hypothesis. A reliable volume-to-CPA claim needs a documented comparison that accounts for spend, targeting, seasonality and other simultaneous changes.
Compare channels using the same business outcome
Give each channel a role around a specific audience and product outcome. Google App campaigns and Performance Max are separate campaign types; App Connect can connect supported web campaigns with app measurement and deep links. Do not choose a spend split from an agency-wide portfolio average. Compare market, message, measurement and mature user value in the account you are managing.
How to evaluate Meta acquisition
For Meta, compare the same app, market, event definition and cohort age before drawing a cost or user-value conclusion against another channel. Check the available bidding strategy and valid event or value signals in the actual account. A fixed weekly conversion count is not a universal test of tROAS eligibility or stable delivery. See Meta Ads for Mobile Apps: How to Scale ROAS for the broader evaluation workflow.
How to evaluate Google App campaigns
Compare Google App campaigns with another channel using the same app, country, observation period and business event. Check whether a low install cost is accompanied by valuable users, and account for differences in measurement coverage. On the NeuroNation Google case study, Admiral Media used a structured creative framework that decreased CPP by 34%, CPI by 40%, and CPR by 40% on Google UAC, with spend scaling 952% on the refreshed creatives. Keep that result tied to the campaign rather than turning it into a universal Google-versus-Meta cost discount.
How to evaluate TikTok acquisition
For TikTok, test whether the message, format and audience fit the app’s actual market. Compare downstream user value and total test cost, not just install cost. An ad that worked on another channel is a useful hypothesis to evaluate, not evidence that it will work here. Set production and refresh cadence from observed delivery, distinct ideas and available capacity rather than a fixed monthly video count.
How to evaluate Apple Ads acquisition
For Apple Ads, separate branded and category searches and record the placement, country and new-download or redownload mix. Compare acquisition cost with the same cohort’s downstream value, and check how paid traffic relates to existing store demand. A brand search or a low install cost does not by itself establish an incremental customer, the highest-value channel or the volume available at the next budget level.
Subscription apps: measure the path to paid renewal
Track the path from install to trial, first payment and renewal. Define each denominator and give trials enough time to mature before comparing their paid conversion. Record subscription price, duration, refunds and acquisition source so a change in D30 revenue is not mistaken for a lasting improvement in contribution value. Test a paywall or onboarding change against a clear hypothesis; a quarterly redesign schedule is not evidence that it will outperform.
On the ChatPDF case study, Admiral Media managed paid acquisition for an AI subscription app and delivered +320% ROAS YoY growth, +156% subscription growth, and -42% CAC reduction. Those are reported engagement outcomes, not a forecast for another subscription app. For broader subscription app strategy, see Admiral Media’s subscription app marketing agency page.
Trial starts, paid conversion and renewal
Evaluate trial length against the time users need to experience the product and the timing of the first charge. Compare trial-start, first-payment and renewal rates for equivalent cohorts, including cancellations and refunds. A three-day versus seven-day test needs a clear question and enough mature outcomes; neither length is automatically better. Use a controlled comparison where feasible rather than treating a short-term rise in trial starts as a paid-retention result.
Gaming: compare monetization models and mature cohorts
For a game, record advertising and purchase revenue separately and inspect how value is distributed across the acquired cohort. A small group of paying users can make an average unstable, so report the denominator and the uncertainty before forecasting return. Evaluate creative tests on meaningful player behavior and mature value; neither a fixed payer-concentration threshold nor a monthly asset count establishes profitable scale.
Compare games with a similar monetization model and player journey. Measure advertising, purchase and hybrid revenue on the same basis, then choose a payback horizon from actual cohort behavior and cash constraints. Do not assume that every casual game pays back at D3, that every strategy game pays back at D30 to D90, or that a casino label establishes D180 return. For a broader planning discussion, see Admiral Media’s user acquisition for mobile games resource.
Localization and country-level economics
Compare countries individually before combining them into a market tier. Match currency, prices, acquisition source, language and measurement coverage, then evaluate mature contribution value alongside acquisition cost. Localization can be tested in the store listing and app journey, but include the work and testing costs. ASO is not automatically the lowest-cost route in every market, and a low-cost install does not identify the best country to scale.
On the NeuroNation Korea ASO case study, Admiral Media delivered a 93.10% increase in search downloads and a 43.95% increase in search impressions through ASO alone, by adapting the store listing to Korean linguistic norms. This was achieved with zero incremental paid spend. This is one reported localization outcome, not a cost ranking across markets.
Case studies: outcomes from individual engagements
The cases below report outcomes from individual Admiral Media engagements. Keep each result tied to its app, period, campaign conditions and metric definition. Without a comparable sample and percentile calculation, these results do not establish what top-quartile performance looks like across the market in 2026.
- NeuroNation (brain training subscription): +117% ROAS, -39% CPI: Admiral Media managed NeuroNation’s mobile UA across Meta and Google, achieving a 117% increase in ROAS, a 66% increase in installs, a 42% increase in net cohort revenue, and a 39% reduction in CPI over 15 months. See the NeuroNation case study.
- NeuroNation Google Creative Framework: -34% CPP, +129% purchases: Admiral Media executed a creative refresh framework on NeuroNation’s Google App Campaigns that decreased cost per purchase by 34%, increased purchases by 129%, and increased spend on top-performing video ads by 952%. See the creative framework case study.
- NeuroNation Korea ASO: +93% search downloads: Admiral Media’s ASO team adapted NeuroNation’s Korean store listing, achieving a 93.10% increase in search downloads and a 43.95% increase in search impressions. See the NeuroNation Korea case study.
- PURE Dating: -74% CPI, D7 ROAS goals exceeded: Admiral Media reduced PURE’s CPI by 74% while exceeding D7 ROAS targets by diversifying beyond walled gardens into DSPs with programmatic bidding. See the PURE case study.
- ChatPDF (AI subscription): +320% ROAS, +156% subscriptions, -42% CAC: Admiral Media scaled paid acquisition for ChatPDF with value-based bidding and structured creative testing, delivering a 320% ROAS YoY increase and a 42% CAC reduction. See the ChatPDF case study.
- Inshallah (dating): +1,253% iOS revenue, +824% iOS subscriptions: Admiral Media reallocated Inshallah’s spend from Android to higher-LTV iOS users and delivered a 1,253% increase in US iOS revenue and 824% increase in active iOS subscriptions. See the Inshallah case study.
- Clark (insurance): -50% CPL, -29% CPI: Admiral Media’s creative strategy reduced Clark’s cost per lead by 50%, cost per install by 29%, and cost per level achieved by 47%, while increasing conversion rate by 41%. See the Clark case study.
- TIER (mobility): +297% new customers, 5x budget scaling: Admiral Media scaled TIER’s UA budget 5x across multiple markets while increasing new customers by 297% and adding two new ad channels. See the TIER case study.
- Fastic (fasting app): +439% revenue growth: Admiral Media scaled Fastic from installs to subscription purchases, achieving a 439% increase in revenue. See the Fastic case study.
- Miles Mobility (car sharing): +260% conversions, -25% CPA: Admiral Media implemented Smart Bidding on web-to-app campaigns and delivered a 260% increase in conversions and a 25% reduction in CPA. See the Miles Mobility case study.
- KaufDA (deals app): +1000% growth in a day, -18% CPI, 70M impressions: Admiral Media’s creator-led campaign delivered 70 million impressions in one month, a 1000% user growth spike on a single day, 146% user activity growth, and an 18% CPI reduction. See the KaufDA case study.
How to investigate gaps before changing spend
Start with the question that changes the next investment decision. The checks below help separate a measurement problem from an audience, creative or product question. Their order can depend on the account; none establishes a guaranteed percentile improvement.
- Check measurement before interpreting a gap. Reconcile the spend and event definitions across reporting sources. Review the applicable attribution path, conversion windows and event delivery, and keep missing or modeled values visible. A saved tracking configuration is not proof that the expected outcome was recorded.
- Test distinct creative hypotheses. Separate a new concept from a minor edit and state the audience, message and outcome being compared. Set the test allocation from available traffic, reporting delay and the effect worth detecting. Review downstream value as well as CPA instead of prescribing an asset count or promising a quarterly cost reduction.
- Choose bidding for a measurable outcome. Compare the quality and availability of the event or value signals with the account’s actual bidding controls. Account for volume and reporting delay before changing strategy. A predicted LTV signal is not observed revenue, and value bidding is not automatically better than an install or event-cost approach.
- Make channel expansion a defined experiment. State what a new channel is expected to add, the budget at risk and how its outcomes will be compared with existing acquisition. Plan tests around capacity and uncertainty rather than requiring a new platform every quarter or a fixed starter-budget share.
- Investigate onboarding and payment separately. Use the observed funnel to choose a product or paywall hypothesis, keep a stable comparison where possible and wait for payments and renewals to mature. A scheduled redesign is not evidence of improvement, and changing several things at once makes the cause harder to establish.
Keep a short decision log with the hypothesis, reporting cutoff, costs and observed outcome. If the evidence is incomplete, say what remains unresolved before increasing spend or calling the change a success.
What benchmarks do not tell you
Benchmarks cannot diagnose the cause of a result on their own. Investigate acquisition mix, activation, product value and payment behavior with the app’s own evidence. Managed-spend totals and brand counts are not a substitute for a comparable benchmark sample or proof that product changes always offer the largest return.
Use the comparison to choose a question you can investigate. Is acquisition cost consistent with the value and payback the business needs? Are cohorts being measured on the same basis? Does a weak outcome coincide with a particular source, country or point in the product journey? A result inside a published range still deserves review when the app’s economics or objectives change.
External sources for benchmark validation
Use public reports as independent references, then compare their methodology with your own data. AppsFlyer’s benchmark suite, Business of Apps and Adjust’s reports can cover different periods, markets and acquisition sources. Apple’s developer documentation explains platform rules, rather than validating a campaign’s CPI or ROAS. Links to those publishers do not constitute a row-by-row validation receipt for the tables in this guide.
Frequently Asked Questions
What is a good CPI for a mobile app in 2026?
A useful CPI is one that supports the acquired cohort’s contribution value and the business’s payback requirement. Compare the same sub-vertical, platform, market and acquisition source, then confirm the benchmark sample and data period. A category range or percentile cannot establish profitability without that cohort’s downstream economics.
What D7 ROAS should a subscription app target?
Set the early return requirement from the app’s mature revenue or contribution curve and the time the business can wait for payback. A D7 revenue-ROAS range does not automatically imply twelve-month payback. Use later cohort ages to validate any early forecast and label forecast values separately from observed revenue.
How should I set a paid user acquisition test budget?
Calculate test spend from baseline volume, conversion delay, expected acquisition cost and the effect that matters commercially. Check that the creative and measurement setup can answer the test question. A fixed monthly budget, variant count or scaling percentage does not guarantee a valid benchmark comparison or profitable growth.
How does iOS privacy affect these 2026 benchmarks?
iOS measurement depends on the implemented framework, version and reporting path. SKAdNetwork 4 can report across conversion windows through Day 35, with a random delay after a window closes; Day 35 is not a universal revenue-reporting deadline. Postbacks may omit detail for privacy, and their values encode the app’s defined signals rather than a complete revenue ledger. Reconcile available attribution with product and payment records, and keep predictive values separate from observed outcomes. See Apple’s conversion-window documentation.
How often should I refresh my benchmark reference numbers?
Review a reference when its data period, market mix or measurement definition stops matching the decision. Record the cutoff date and source method, and compare meaningful updates on the same basis. A quarterly review can be an operating cadence, but it does not establish that every auction changes quarterly or that an older reference is wrong by a fixed percentage.
What retention rate should I target for my app?
Define the return event and cohort first, then compare equivalent users at the same age. A top-quartile target needs a documented distribution for that sample. Use the table as an investigation aid only after its method is confirmed, and test whether the issue lies in acquisition mix, activation or product value rather than assuming one fixed cause.
Why can benchmark reports show different ranges?
Published sources can differ because of app samples, countries, channels, dates, weighting and metric definitions. Request those details before comparing ranges. Agency-wide managed spend and brand counts are not the sample definition for each table, and a broad publisher link does not verify every value. The useful reference is the one whose method and cohort match your decision.
