App Churn Rate: How to Find Where Users Leave and Win Them Back

App churn rate is the percentage of users who stop using or stop paying for an app within a defined period. It is the inverse of retention. If 100 users install your app on Monday and 25 are still opening it on the following Monday, your day 7 retention is 25% and your day 7 churn rate is 75%. For subscription apps the more important version is paying churn: the share of active subscribers who cancel, lapse, or fail to renew in a given month or billing cycle.

Most teams treat churn as a product problem and stop there. Admiral Media treats it as a marketing economics problem, because churn is the single variable that decides what you are allowed to pay for a user. Every cost per install target, every target ROAS setting, and every scaling decision is downstream of how long users stay. This guide explains where app churn actually happens, how to measure each stage separately, and which levers move which stage, using real numbers from Admiral Media client work and from Apple and Google platform documentation.

Admiral Media has managed more than €500M in ad spend across 150+ mobile brands and holds a 5.0 rating on Clutch. The patterns below come from running acquisition, creative, and monetisation programmes for subscription apps in health and fitness, dating, edtech, fintech, and AI.

What app churn rate actually measures

App churn rate measures user loss over a period, but the word covers four different things that behave differently and need different fixes: install churn, activation churn, voluntary paying churn, and involuntary paying churn. Install churn is enormous and mostly unavoidable. Involuntary paying churn is small and almost entirely fixable. Treating them with the same tactics wastes budget on the wrong end of the funnel, and confusing them is the most common reason churn work fails.

Install churn: the number everyone quotes

Install churn is the share of installers who stop opening the app. It is brutal across the whole market. According to Business of Apps data published in 2026, iOS apps see a 75% churn rate after day one, rising to 95% by day 30, while Android apps see 80% after day one and 96% by day 30.

These numbers are useful as context and useless as a target. A 95% day 30 install churn rate is normal, so beating it by two points says little about business health. Admiral Media does not optimise against install churn directly. It optimises against what install churn hides: which channels, creatives, and geos deliver the small minority who stay.

Activation churn: the stage you can actually move

Activation churn is the share of users who open the app but never reach the moment that makes the product worth paying for. This is where the largest recoverable losses sit. A user who never completes onboarding cannot convert to trial, so every euro spent acquiring them is wasted regardless of how good the paywall is.

Acquisition feeds this rung directly. In Admiral Media’s work with NeuroNation, the German brain training app, the client noted that alongside doubling user acquisition efficiency they got useful feedback to optimise store listing pages and onboarding flows. The creative that wins the auction sets the expectation the onboarding has to meet, so a mismatch between the two shows up as activation churn rather than as a creative problem.

Voluntary paying churn: the cancel decision

Voluntary paying churn is a subscriber actively turning off auto renewal. It is a judgement about value received versus price paid, and it clusters around two moments: shortly after the trial converts, and shortly before a large renewal charge. Billing period is the strongest single predictor. Business of Apps reports that over 98% of weekly subscriptions churn before one year, dropping to 92% for monthly subscriptions and 73% for annual subscriptions. Annual plans are not simply better products. They compress the number of cancel decisions a user has to make from 52 to one.

Involuntary paying churn: the leak nobody watches

Involuntary paying churn is a subscriber who wanted to stay but whose payment failed. Expired cards, insufficient funds, and issuer declines end subscriptions that had no intention of ending, which makes this the cheapest churn to recover.

Both platforms ship recovery machinery. Apple states that when a subscription renewal fails, Apple attempts to recover it for 60 days, and that enabling Billing Grace Period in App Store Connect lets Apple address the billing issue while the subscriber retains access, with a configurable duration of 3, 16, or 28 days. On the Google side, Google reported in 2020 that account hold helped developers achieve 8% lower involuntary churn and a 35% higher payment decline recovery rate compared to developers without account hold. From 1 December 2025, Google Play’s default account hold duration is automatically calculated as 60 days minus any grace period duration.

These are settings, not strategies. Admiral Media checks them in the first week of every subscription app engagement, because a misconfigured grace period silently taxes every acquisition euro the account spends.

Why churn is a media buying problem, not just a product problem

Churn sets the ceiling on your acquisition budget. Lifetime value is a function of how long a paying user stays, and lifetime value divided by your payback target is the most you can bid. Cut churn and you can outbid rivals in the same auction, on the same creative, with the same product.

Most teams get this backwards. They fix a weak ROAS by lowering cost per install, which means buying lower quality traffic, which raises churn, lowers lifetime value, and makes ROAS worse. Admiral Media calls this the cheap install trap, and it is the most common pattern the Admiral Media team finds when auditing an underperforming account.

Admiral Media’s work with Inshallah, a dating app for the Muslim community, shows the correct diagnosis. Low cost Android acquisition helped community building, but analysis showed high churn by day 5 from those users with revenue below targets, while iOS users generated significantly higher revenue and nearly twice as high retention rates. Identifying that early let the team narrow ad spend allocation accordingly. Admiral Media reallocated Inshallah’s spend towards iOS and purchase optimised campaigns, and United States iOS revenue increased 1,253% and United States iOS subscriptions increased 824% since Admiral Media took over. The full breakdown is in the Inshallah case study.

The lesson generalises. Cheap installs that churn by day 5 are not cheap. They are a negative return on capital that surfaces two months later in a cohort report. This is why Admiral Media anchors scaling decisions to CAC payback period and blended ROAS and MER rather than install cost alone.

Churn changes what your bidding algorithm learns

Campaign types optimise towards the signal you send them. Send install events and Google App Campaigns and Meta Advantage+ App Campaigns will find people likely to install, including people who install and leave. Send purchase or value events and the algorithm optimises for people who pay, which correlates far better with people who stay.

The mechanism is auction level. Bidding systems build a predicted conversion probability for each impression and bid your target against it. Feed them a shallow event and that prediction is trained on a population full of high churn users. Feed them a revenue weighted event and it shifts towards the population that renews. This is why moving from cost per install optimisation to target ROAS optimisation typically reduces measured churn in the acquired cohort with no product change at all.

ChatPDF, an AI powered document interaction tool, is a worked example. Admiral Media restructured ChatPDF’s accounts, moved to value based bidding with value rules and an LTV signal, and ran weekly creative concept tests, delivering +320% ROAS, +156% subscriptions, and a 42% CAC reduction on a blended basis. By channel, Google delivered 320% ROAS year on year growth, +142% subscriptions, and a 38% CAC reduction, while Meta delivered 280% ROAS growth, +171% subscriptions, and a 45% CAC reduction. Numbers use an index baseline comparing year one against year two year to date, per the ChatPDF case study.

The Admiral Media Churn Ladder

Churn is not one event. It is six separate drop off points, each with a different owner, lever, and measurement window. The Admiral Media Churn Ladder is the diagnostic sequence the Admiral Media team runs before recommending any tactic, because fixing rung four while rung two is broken produces no measurable gain.

The Admiral Media Churn Ladder Framework

1. Rung 1: Source quality. Measure churn by acquisition source before anything else. Break day 1, day 7, and day 30 retention out by channel, campaign, creative concept, geo, and platform. If one source shows materially worse day 5 behaviour, that is not a product problem and no onboarding change will fix it. This is the rung where Admiral Media found Inshallah’s high day 5 Android churn.
2. Rung 2: Activation. Define the single action that predicts payment in your category, then measure the share of installs reaching it within the first session and within 24 hours. For a fasting app that might be logging a first fast; for a dating app, completing a profile. Every point of activation lift compounds through every rung above it.
3. Rung 3: Trial or paywall conversion. Measure install to trial and trial to paid as separate rates. A weak install to trial rate is a paywall placement and offer problem; a weak trial to paid rate is a value delivery problem inside the trial window. One blended number hides which is broken.
4. Rung 4: First renewal. The highest risk moment in the subscription lifecycle, because it is the first time the user pays without the novelty that drove the original decision. Measure it as its own cohort metric, separate from ongoing renewal.
5. Rung 5: Involuntary loss. Split failed renewals from cancelled renewals in reporting. Without that split you cannot know whether your retention problem is a value problem or a payments configuration problem. Check grace period and account hold settings before running any product experiment.
6. Rung 6: Win back. Treat lapsed subscribers as a distinct, high intent audience with their own offer, creative, and economics. A former subscriber has already proven willingness to pay, which makes them the cheapest revenue in the account.

Work the ladder from rung 1 upward. The order is deliberate: source quality problems contaminate every metric above them, so diagnosing rung 3 with a polluted cohort produces a confident wrong answer.

Which rung owns which metric

Each rung has a metric that moves and a team that owns it. Mapping them explicitly prevents the most common failure mode: marketing and product each assuming the other is handling churn.

How to reduce app churn at the acquisition layer

The fastest churn reduction available to most apps is not a product change. It is changing what you buy and what signal you optimise towards, because acquisition sets the composition of every cohort you later try to retain.

Optimise on the deepest event you can feed reliably

Target ROAS bidding needs enough conversion volume to exit the learning phase and produce stable predictions. Where an account cannot supply that volume, Admiral Media typically uses a proxy event that correlates with retention, such as activation or trial start, rather than falling back to install optimisation. The proxy keeps the algorithm pointed at durable users while the account accumulates the purchase volume needed for full value bidding.

On iOS the constraint is tighter, because SKAdNetwork and AdAttributionKit deliver coarse, delayed, privacy thresholded signal. Admiral Media’s approach for Inshallah included refining the SKAN conversion value schema, shifting from CPI to ROAS as the main KPI, and moving to purchase optimised campaigns in the United States in September 2023. If your schema encodes only installs and registrations, no amount of bid tuning will teach the algorithm what a retained subscriber looks like. Admiral Media covers the mechanics in its guide to SKAdNetwork conversion values.

Buy the platform that keeps users, not the platform that is cheapest

Cost per install differences between platforms are visible immediately. Retention differences are visible six weeks later. Teams optimise what they can see, which biases budget towards the cheaper, worse cohort. The correction is to hold platform allocation open until cohort data arrives, then reallocate on retained value rather than install cost, as Admiral Media did for Inshallah.

Cheap does not always mean low quality, and the test is worth running properly. For PURE, a dating app focused on honest connections, Admiral Media tested a demand side platform against an established self attributing network on United States Android, with separate budgets and tailored ads for each. The test delivered a CPI of $2.44 against the self attributing network’s $9.43, a 74% CPI reduction, while exceeding D7 ROAS goals and enabling new market entries. The D7 ROAS condition is the point: cheap installs were only accepted because seven day revenue held. Details are in the PURE case study.

Match creative promise to product reality

Creative that overpromises buys installs and manufactures churn. The user arrives expecting something the onboarding does not deliver and leaves inside 24 hours. This cause sits entirely inside the media team’s control and almost never appears in a retention post mortem. The counterweight is creative volume with fast feedback, so winners are identified on downstream performance rather than on click through rate. In Admiral Media’s work with Fastic, the world’s number one fasting app, the Admiral Media team built the AI Creative Factory: generative AI produces hundreds of on brand variants, live performance data keeps the winners and drops the rest, and brand rules keep every ad consistent. The system cut Fastic’s cost per result by 70% and won Silver in the AI category at The Drum Awards for Marketing EMEA 2026, within a wider Admiral Media programme that delivered +639% installs, +1,655% purchases, +439% revenue, and +952% monthly active users. The full account is in the Fastic case study.

Purchases growing faster than installs is the signal that matters: it indicates the creative selected for people who convert rather than people who tap.

How to reduce churn at the trial and paywall layer

Trial and paywall churn is the conversion of interest into payment, and it is the rung where pricing structure does more work than persuasion. The decisions made here also set how many future cancel decisions the user will face.

Three structural choices dominate. Billing period: annual plans face one renewal decision per year against 52 for weekly plans, which is the mechanical reason annual churn is lower. Trial length: long enough for the user to experience the value that justifies the price and no longer, because every extra day loosens the link between charge and benefit. Paywall timing: showing the offer before the user has felt any value converts a smaller share and produces worse retained cohorts.

Admiral Media treats the paywall as a media asset rather than a product screen, because the price and offer on it define the bid ceiling for the whole account. The method is in Admiral Media’s guide to app paywall optimization.

Use platform offer mechanics rather than discounting blindly

Both stores provide native offer types that reduce churn without permanently repricing the product. Apple documents introductory offers, promotional offers, offer codes, and win back offers, notes you can have up to 10 offers for each subscription, and describes a save offer that can be shown to a subscriber who has turned off auto renew but whose subscription has not yet expired. That is a targeted intervention at exactly the moment voluntary churn is decided. Google’s equivalent set includes pause for up to 3 months with automatic resumption, resubscribe within 12 months of expiry from the Play subscriptions centre, and a cancellation module that reminds users of up to 4 subscriber benefits.

Pause deserves specific attention. A paused subscriber is a retained subscriber with a delayed renewal. A cancelled subscriber is a re acquisition cost. Offering pause converts some share of the second group into the first at zero media cost.

How to stop involuntary churn

Involuntary churn is the cheapest churn to eliminate because it requires configuration rather than persuasion. The subscriber already wants the product; the payment simply failed. The platform mechanics are documented and specific, and the table below summarises the recovery settings both stores expose.

Sources: Apple, Auto-renewable Subscriptions, Google Play Console Help on subscription payment recovery periods, and the Android Developers Blog on subscriber retention features.

The revenue rate argument for keeping subscribers alive

There is a commercial reason to fight involuntary churn beyond the lost subscription itself. Apple states that during a subscriber’s first year of service you receive 70% of the subscription price at each billing cycle minus applicable taxes, and that after a subscriber accumulates one year of paid service your net revenue increases to 85%. Apple further states that if a subscription expires due to a cancellation or billing issue the days of paid service stop accumulating, and resume from the recovery date if the subscription is renewed within 60 days.

The implication is precise. A subscriber lost at month 11 and re acquired at month 14 does not resume where they left off. Involuntary churn therefore costs more than the missed renewals: it resets the revenue rate on that relationship. Apple notes that developers enrolled in the App Store Small Business Program receive 85% regardless of accumulated paid service, so the size of this effect depends on enrolment status.

How to win back churned users profitably

Lapsed subscribers are the highest intent audience an app has, and most accounts underinvest in them. They have installed, activated, converted, and paid. The only missing input is a reason to return.

Win back works on three surfaces at once, and running only one leaves most of the value on the table. The store surfaces work automatically for eligible users: Apple notes that a win back offer sheet automatically appears to eligible customers within your app or game with no additional work required, and gives the example of a pay up front offer at $9.99 for six months against a standard renewal price of $39.99 per year. The CRM surface handles users you can still reach by email or push. The paid surface handles everyone else through retargeting.

The paid surface holds most of the addressable volume. Admiral Media budgets win back retargeting against a separate payback target from prospecting, because a returning subscriber typically converts on a shorter consideration cycle and at a lower cost than a cold user. Admiral Media’s approach to segmenting and budgeting these audiences is covered in its guide to app retargeting.

One caution: win back campaigns always report conversions, because some lapsed users would have returned anyway. Hold out a randomised portion of the audience and compare, or you are measuring reappearance rather than incrementality.

How to measure app churn rate correctly

Most churn reporting is wrong in a way that flatters the account. The two most common errors are mixing cohorts and mixing billing periods, and both make churn look lower than it is.

Measure by cohort, not by period snapshot

A period snapshot divides cancellations in a month by subscribers at the start of that month, blending users at very different lifecycle points. A month with heavy new acquisition therefore looks better than it is, because new subscribers have not yet reached their first renewal. Cohort measurement groups users by the month they started paying and follows each group forward. It is the only view that shows whether the users you are buying this quarter are better than the ones you bought last quarter.

Admiral Media reports cohort revenue alongside acquisition metrics for this reason. In Admiral Media’s first 15 months with NeuroNation, using data from January to August 2019, the programme delivered +66% installs, +32% purchases, +42% net cohort revenue, a 39% CPI reduction, and a 117% ROAS increase. Net cohort revenue is the figure that proves the extra installs were worth buying. Full detail is in the NeuroNation case study.

Split voluntary from involuntary in every report

If your reporting cannot distinguish a cancelled subscription from a failed payment, every retention conclusion you draw is unreliable. Apple exposes subscription state through its Get All Subscription Statuses endpoint, which determines whether a subscription is active, expired, in billing retry, or in grace period. Google directs developers to the SubscriptionPurchaseV2 endpoint for current subscription state including account hold and expired states, and warns that assuming a static value in your app, backend, or CRM may result in a confusing user experience. Wire these in before you run experiments.

Source level retention also needs a permanent place in reporting rather than a one off audit, because channel quality drifts as auctions, audiences, and creative age. Admiral Media builds this into standing reporting, an approach it calls retention first user acquisition.

Churn reduction priorities by app stage

The right first move depends on your stage, because the same tactic returns very differently at different scales. Below is the prioritisation Admiral Media uses when scoping a new subscription app engagement.

Two rules cut across every stage. Never run a paywall test on a cohort contaminated by a bad traffic source. Never discount your way out of a retention problem, because a discount lowers the lifetime value that sets your bid ceiling and shrinks the audience you can afford to buy.

Where churn work usually goes wrong

Churn programmes fail for a few repeated reasons, and most are organisational rather than technical. Teams optimise the wrong rung, reaching for an onboarding redesign when one traffic source is delivering users who churn by day 5. They measure churn without cohorts. They ignore involuntary churn because it belongs to nobody. And they scale before the ladder is stable, which widens targeting, lowers user quality, and turns a churn problem into a bigger one. Admiral Media establishes cohort stability at the current spend level before adding budget, the same discipline behind its work with Clark, the German insurance management app, where moving optimisation to custom lower funnel events delivered a 41% conversion rate improvement, a 29% CPI reduction, and a 50% reduction in cost per lead comparing month 3 against month 1.

Frequently Asked Questions

What is a good app churn rate?

There is no single good number, because churn depends on billing period, category, and platform. For install level churn, Business of Apps data published in 2026 reports 95% churn by day 30 on iOS and 96% on Android, so a day 30 install churn rate near those levels is normal rather than alarming. For paying subscribers, the same source reports that over 98% of weekly subscriptions churn before one year, against 92% of monthly and 73% of annual subscriptions. Compare yourself against your own prior cohorts, not a market average.

How do I calculate app churn rate?

For install churn, divide the users from a cohort who have stopped opening the app by the size of that cohort at install, measured at a fixed day such as day 1, day 7, or day 30. For subscriber churn, divide the subscribers in a cohort who cancelled or lapsed during a period by the cohort size at the start of that period. Always group users by the date they installed or first paid, and never blend billing periods into a single figure, because plan mix changes move the blended number without any change in user behaviour.

What is the difference between voluntary and involuntary churn?

Voluntary churn is a subscriber deliberately turning off auto renewal because they no longer see enough value for the price. Involuntary churn is a subscriber losing their subscription because a payment failed, typically an expired card, insufficient funds, or an issuer decline. They need different fixes: voluntary churn responds to product value, pricing, and offers, while involuntary churn responds to billing configuration such as grace periods, account hold, and payment update prompts. Involuntary churn is usually cheaper to fix because the user already wants to stay.

Does reducing churn let me spend more on user acquisition?

Yes, and this is the main commercial reason to work on it. Lifetime value is a function of how long paying users stay, and your maximum sustainable cost per acquisition is lifetime value divided by your payback requirement. Lower churn raises lifetime value, which raises the bid you can afford, which wins auctions and audiences that were previously unaffordable. Admiral Media routinely finds a retention improvement unlocks more incremental spend capacity than any bid or creative optimisation available at the same time.

Why do cheap installs often churn faster?

Cost per install reflects competition for a given user, and expensive users are usually expensive because advertisers with good data are bidding for them. Very cheap inventory often reaches users with low intent for your category, who install out of curiosity and leave within days. Admiral Media saw this in its Inshallah work, where low cost Android acquisition showed high churn by day 5 with revenue below targets, while iOS users generated significantly higher revenue and nearly twice as high retention. Judge sources on retained cohort value, not install cost.

How long should I wait before judging a cohort’s churn?

Long enough to cover at least one renewal event for that billing period, plus the platform recovery window. For a monthly plan that means roughly 60 days, since Apple attempts to recover failed renewals for 60 days and Google Play’s default account hold from December 2025 is calculated as 60 days minus any grace period. Judging a monthly cohort at 30 days overstates churn, because subscriptions still in grace period or account hold have not actually been lost.

Who should own churn reduction in an app team?

No single team can own it, which is why churn work stalls so often. Source quality churn belongs to media buying, activation to product, paywall and trial conversion to monetisation, first renewal to CRM and product together, involuntary churn to engineering and finance, and win back across media buying and CRM. The Admiral Media Churn Ladder assigns each rung an owner and a single metric precisely so the work does not fall into the gaps between functions.