Most apps don’t have a push notification problem. They have a permission problem, then a relevance problem, and only after that a volume problem. A push notification strategy is the set of rules an app uses to decide who can be messaged, which moments earn a notification, when it lands, and how many a user gets before the app starts losing them. Get those rules in the right order and push is the cheapest retention channel you own.
Table of Contents
Teams usually attack the problems in reverse: they tune send times and emoji for an audience that never opted in, then decide the channel is dead. This guide is how Admiral Media builds the rules for subscription, dating and habit apps: the opt-in numbers that changed after Android 13, a four-gate test every message has to pass, the platform rules that catch teams out, and a calculator that shows how many of your installs you can reach at all.
What does a push notification strategy actually decide?
It decides four things, in this order: whether you’re allowed to message a user at all, which moments earn a notification, when that notification arrives, and how many the user gets before they stop listening. Copy, emoji and rich media are optimisation on top of those four decisions.
The order matters because each decision caps the next one. A perfect message sent to 40% of your installs reaches fewer people than a decent one sent to 60%. And a well-timed message nobody asked for still teaches users to swipe you away without reading.
How many of your users can you actually reach with push?
Fewer than most retention plans assume. Airship’s 2026 benchmark report puts the median opt-in rate at 53.3% on Android and 48.85% on iOS for 2025, which means roughly half of a typical app’s users can’t be reached by push at all.
Android opt-in has fallen to meet iOS
Median across Airship customer apps, percent of users opted in
The story in that chart is Android. Until Android 13, apps could send notifications without asking, so Android opt-in sat far above iOS. Android 13 added a runtime permission, POST_NOTIFICATIONS, and on new installs notifications are now off until the user allows them. Airship’s median Android rate dropped from 71.3% in 2023 to 59.5% in 2024 and 53.3% in 2025. iOS barely moved.
Two caveats before you turn these into targets. They’re medians across Airship’s customers, who skew toward larger consumer brands, and the 2025 edition covers more verticals than the earlier one. Use them for direction. Your own rate, split by platform and install cohort, is the number that matters.
Where this bites is forecasting. We still come across retention models built in 2021 or 2022 that treat most Android users as reachable by push. If your reactivation plan, your LTV model or your trial win-back flow leans on push, rerun it with your current opt-in rate by platform. The Android side of those models is often the most optimistic line in the spreadsheet.
Try it
Push reach calculator
How many of the users you install each month can you actually message? Swap the default opt-in rates for your own.
- users you can reach by push each month
- of new installs reachable
- users you can only reach in-app, by email or with paid retargeting
- extra reachable users per month from 5 more points of iOS opt-in
How this is calculated
Reachable users = monthly installs × platform share × opt-in rate, summed across iOS and Android. It counts users who opt in during their install month, so treat it as an upper bound for later months, when some users revoke permission. The default opt-in rates are Airship medians for 2025 data; your own rates by install cohort are more useful.
When should you ask for notification permission?
Ask after the user has seen the thing notifications will protect, not on first launch. Apple’s guidance is to request authorization in a context that makes the value obvious, such as right after someone schedules their first task. Airship’s data backs this up: apps running onboarding campaigns that explain the value of notifications see opt-in rates as high as 40% above their category average.
On iOS the system prompt is effectively a one-shot request. A user who taps Don’t Allow has to find your app in Settings to change their mind, and very few do. Android 13 now gives apps targeting API level 33 or higher the same control over timing that iOS teams have always had. Use it at a moment of value rather than at app start.
| Moment | Why it works | What it costs | Best for |
|---|---|---|---|
| System prompt on first launch | No design or engineering work | The user has no reason to say yes yet, and you spend your one clean ask | Almost no one |
| Soft prompt, then system prompt | You explain the value in your words; users who aren’t ready can skip without triggering the system dialog | One more screen in onboarding | Most apps |
| At the first moment of value | The benefit is concrete: “tell me when she replies” | Needs event logic and a primer per trigger | Dating, social, marketplaces, productivity |
| iOS provisional authorization | Notifications arrive quietly in Notification Center, with Keep and Turn Off options | Quiet delivery means little visibility until the user keeps them | Content-heavy apps that can earn the opt-in by showing value |
| Later, through an in-app message | Airship measured a 14% average opt-in increase among opted-out users | Must deep-link to system settings | Recovering users who declined early |
How do you decide whether a push should be sent at all?
Run every candidate message through four gates before it goes out: did the user cause or ask for it, would they be worse off missing it, is this the right moment for them, and do they have budget left this week. A message that fails a gate isn’t deleted. It gets rerouted to a quieter channel.
Admiral Media framework
The Admiral Media Send-or-Hold Test
Four gates, in order. Any “no” sends the message somewhere quieter instead of to the lock screen.
Trigger. Did this user do something, or ask for something, that this message responds to?
If noHold it for the in-app inbox or the next session. Scheduled blasts to everyone rarely clear this gate.
Miss test. Would the user be annoyed to learn later that you knew and didn’t tell them?
If noMove it to email or an in-app card. It’s content, not news.
Moment. Is it inside this user’s active hours, in their own time zone, and outside quiet hours?
If noQueue it for their next active window, unless the user would call it urgent.
Budget. Does this user’s lifecycle stage have push budget left this week?
If noDrop it or fold it into the next message. One strong push beats three average ones.
Passes all four: send it, with a deep link to the exact screen it mentions.
We use gates rather than a weighted score on purpose. With a score, a weak trigger can buy its way through with a great send time or a clever headline, and that’s exactly how apps drift into daily “We miss you” pushes. Gates force the uncomfortable question first: did anyone ask for this?
The miss test does most of the work in practice. A reply from a match, a streak about to break, a price drop on a saved item and a trial that ends tomorrow all pass it easily. A new feature announcement, a blog post and a generic discount usually don’t. The same logic applies to Apple’s interruption levels: time-sensitive notifications can break through Focus, so reserve them for things the user would call urgent, not things your roadmap calls urgent.
The budget gate is what keeps a good program good as it grows. Every team that adds a new trigger believes theirs is the important one. A weekly ceiling per user makes triggers compete, and the weakest ones fall back to email or the in-app inbox where they belong.
What should a push notification actually say?
Say the specific thing that changed and what the user can do about it, in words they’d use themselves. “Sara replied to your message” beats “You have a new notification”, and “20 minutes left in your fast” beats “Stay on track today!”
| App type | Vague version | Specific version | Why it’s better |
|---|---|---|---|
| Dating | Someone new is waiting for you! | Lena liked your profile. See her photos. | Names the event and the next action |
| Fasting | Don’t forget your goals today | 20 minutes left in your 16:8 fast. Log it to keep your 12-day streak. | Time-bound, with something to lose |
| Language learning | We miss you! | 14 review cards are due. Five minutes clears them. | Makes the task small and concrete |
| Subscription | Special offer inside | Your free trial ends tomorrow. Here’s what you used this week. | States the deadline and the value in one line |
Specific copy also performs better in the data. Airship’s 2025 benchmark report found that tailored messages raised open rates by 37% on average, and that rich push with images or media lifted direct opens by 22% on average over plain notifications. Neither number will rescue a message that fails the trigger gate, though.
One rule catches a lot of iOS teams. App Store Review Guideline 4.5.4 says push notifications shouldn’t be used for promotions or direct marketing unless the user has explicitly opted in through consent language shown in your app, with a way to opt out inside the app. In practice that’s two permissions: the system permission to show notifications, and your own consent for marketing messages. Store them separately. It keeps you compliant and gives you a cleaner segment for promotional sends.
Is triggered messaging really better than scheduled campaigns?
Yes, and by a wide margin. Batch’s 2025 push benchmark measured a 14.4% open rate for triggered, contextual pushes against 4.19% for generic sends, and OneSignal’s 2026 engagement report puts behaviour-triggered messages at 4 to 9 times the click-through rate of standard sends.
Triggered pushes open at more than three times the rate of generic ones
Open rate, percent of pushes sent
Segmentation is the middle ground when you can’t trigger everything yet. Airship reports that segmented pushes lift direct open rates by between 18% and 65% over unsegmented sends. Segment by lifecycle stage and by what the user actually does in the app, not by demographics you inferred at install.
How many push notifications is too many?
There’s no universal number, and anyone who gives you one is guessing. The honest answer is a weekly budget per lifecycle stage plus close monitoring of opt-outs by message type, because the public data on frequency is thinner than most guides admit.
Airship’s 2026 report shows the heaviest senders, apps at the 90th percentile, now sending close to 180 to 190 pushes per user per month, while also naming “too frequent” as the top reason people opt out. The most-quoted frequency study, VWO Engage data compiled by Business of Apps, found unsubscribes stayed under 1% up to five pushes a day and reached about 7% at 16 to 20 a day. That data is several years old, so we use it as a sanity check, not a target.
What we do instead:
- Give each lifecycle stage a weekly ceiling (the table below has our starting points) and make triggers compete for those slots.
- Track opt-out and uninstall rates per message type, not per app. One bad campaign type can hide inside a healthy average.
- Set a stop-loss. If a campaign’s opt-out rate after the first send is clearly above your app’s baseline, pause it before the second send.
- Keep a no-push holdout group, so you can show the channel adds retention rather than just collecting credit for opens. Our guide to incrementality testing covers the setup.
How do iOS and Android push rules differ?
They differ on permission, on how urgency works and on marketing consent. The table below covers the rules that most often change a plan.
| Dimension | iOS | Android 13 and later |
|---|---|---|
| Permission | Explicit opt-in through the system prompt | Runtime POST_NOTIFICATIONS permission; off by default on new installs |
| Existing users on upgrade | Not applicable: always opt-in | Apps that already had a notification channel generally keep permission when the device upgrades, unless the user had turned notifications off |
| Quiet trial delivery | Provisional authorization delivers to Notification Center with Keep and Turn Off options | No direct equivalent |
| Urgency controls | Interruption levels: passive, active, time-sensitive and critical (critical needs an Apple entitlement) | Five channel importance levels, from IMPORTANCE_NONE to IMPORTANCE_HIGH; users can change a channel’s importance after you create it, you can’t |
| Marketing pushes | Guideline 4.5.4: explicit opt-in consent in your app plus an in-app way to opt out | Give marketing its own channel so users can mute it without losing transactional alerts; Play policy and local law still apply |
| Median opt-in (Airship, 2025 data) | 48.85% | 53.3% |
Which push belongs at which stage of the user lifecycle?
Match the message to the stage. A nudge to finish setup helps a new user and irritates someone who’s been subscribed for two years. The map below is how we start; it connects directly to the work in our guides on onboarding optimization and app churn.
| Stage | What earns a push | Weekly ceiling to start with | Metric to watch | Fallback channel |
|---|---|---|---|---|
| New (days 0 to 3) | Finishing setup, the first value moment: first match, first plan, first lesson | One a day, tied to onboarding steps | Activation rate | In-app checklist |
| Activating (days 4 to 14) | Streaks, reminders the user set, replies | Three to five | Day 7 and day 14 retention | Email recap |
| Engaged | Social activity, changes to saved items, new content in followed topics | Up to one a day, event-driven only | Sessions per week, conversion | In-app inbox |
| At risk | The feature they used most, with one clear reason to come back | Two or three, tapering off | Reactivation rate | Email, then paid retargeting |
| Lapsed or cancelled | What changed since they left, a win-back offer | One or two a month | Reactivation and resubscription | Email and paid retargeting |
Treat those ceilings as starting points for testing, not benchmarks. They move with your category, and your own opt-out data should move them.
How do you measure whether push is working?
Measure incremental retention, not opens. The only clean way to know push adds value is to hold a random slice of opted-in users out of non-transactional pushes, then compare their retention and revenue with everyone else’s over the same period.
Opens still have a place as a diagnostic. Airship’s 2025 report puts the median direct open rate at 3.4% on Android and 3.1% on iOS for 2024. It also reports that opted-in customers make 13% more purchases than opted-out ones. That’s a correlation, not proof: people who opt in were already more engaged. The holdout is how you separate the two.
The scorecard we set up for clients has four lines:
- Opt-in rate by platform and install cohort, so onboarding changes show up within weeks.
- Direct and influenced opens. A direct open is a tap on the notification. An influenced open is a user who opens the app on their own shortly after receiving one.
- Opt-out and uninstall rate per message type.
- Incremental retention and revenue against the holdout group, reviewed monthly.
Where does push fit next to paid user acquisition?
Push decides how much of the lifetime value you paid for actually shows up. When you buy a cohort, part of its value depends on your ability to bring those users back, so a lower opt-in rate shortens the tail of every cohort you acquire. That’s why Admiral Media reviews CRM and push as part of a user acquisition audit, not as another team’s problem.
When Admiral Media took over user acquisition for Inshallah, a dating app for the Muslim community, the work started with an audit of UA, onboarding, CRM communications and paywalls, including research into how top competitors handled onboarding, paywall tests and CRM ideas for email and push. The team also found that iOS users retained at nearly twice the rate of Android users, which moved budget toward iOS. Since then, US iOS subscriptions grew 824% and US iOS revenue grew 1,253%. Those results come from the whole program, not push alone; the Inshallah case study has the detail.
- +824%US iOS subscriptions, Inshallah
- +1,253%US iOS revenue, Inshallah
- +42%Net cohort revenue, NeuroNation
- +117%ROAS, NeuroNation
NeuroNation, a brain-training app, is the other side of the same idea. Over the first 15 months, Admiral Media combined creative and channel testing with feedback on store listing pages and onboarding flows. NeuroNation’s co-founder said the team “doubled our user acquisition efficiency”. In the January to August 2019 data, installs rose 66%, net cohort revenue 42% and ROAS 117%, with CPI down 39%. The NeuroNation case study has the full breakdown.
There’s a practical overlap too. The in-app events that power good push triggers, a completed lesson, a match, a trial start, are the same events you want to send to ad platforms for value-based bidding. Instrument them once and both channels get better, which is the thinking behind predictive LTV bidding.
Frequently asked questions
What is a good push notification opt-in rate in 2026?
Airship’s benchmarks put the 2025 median at 48.85% on iOS and 53.3% on Android, so a rate near 50% is typical rather than good. Category matters a lot, and so does when you ask, so compare yourself with apps in your own vertical. Track the rate by platform and by install cohort, because an average across all users hides whether recent onboarding changes are working.
Should you use a soft prompt before the system permission dialog?
In most apps, yes. A soft prompt is an in-app screen that explains what notifications will do for the user before the operating system asks. Users who aren’t ready can skip it without triggering the system dialog, which on iOS you effectively only get to show once. Airship recommends this approach and reports that onboarding campaigns explaining the value of notifications lift opt-in as high as 40% above category average.
Do push notifications actually improve retention?
They can, but most published numbers mix cause and effect. Airship reports that opted-in customers make 13% more purchases, yet people who opt in were more engaged to begin with. The reliable test is a holdout: keep a random share of opted-in users out of non-transactional pushes and compare their retention and revenue with everyone else. If the gap is small, your triggers need work before your volume does.
What is the difference between a push notification and an in-app message?
A push notification reaches the user outside your app and needs operating system permission. An in-app message only appears while the user is in a session and needs no permission. In-app messages are the best tool for earning push permission later: Airship measured a 14% average increase in opt-in among previously opted-out users who were shown in-app messages.
Are marketing push notifications allowed on iOS?
Yes, with conditions. Apple’s App Store Review Guideline 4.5.4 says push notifications should not be used for promotions or direct marketing unless the user has explicitly opted in through consent language in your app, and you must give them a way to opt out inside the app. That consent is separate from the system notification permission, so keep the two in separate fields in your user data.
What is Android 13’s notification permission and who does it affect?
Android 13 introduced a runtime permission called POST_NOTIFICATIONS. On new installs, notifications are off until the user allows them, and apps that target Android 13 or higher choose when to show the request. Apps that already had a notification channel generally keep permission when an existing device upgrades, unless the user had turned notifications off. The effect is that Android opt-in rates are converging with iOS.
Sources and methodology
Every statistic in this guide links to where it came from. Benchmarks from push platforms describe their own customers, so we use them for direction and label the data year. Admiral Media client results come from the published case studies linked above. The framework, the lifecycle ceilings and the copy examples are Admiral Media’s own working practice.
- Airship, Mobile App Push Notification Benchmarks for 2025: 2023 and 2024 opt-in medians, direct open rates, personalization, rich push, segmentation and opt-in lift figures.
- Airship, Mobile App Push Notification Benchmarks 2026: 2025 opt-in medians, send volumes and top opt-out reason.
- Batch, The Great Push Notifications Benchmark 2025: triggered vs generic open rates.
- OneSignal, 2026 State of Customer Engagement report: click-through for behaviour-triggered messages.
- Business of Apps, Push Notifications Statistics: VWO Engage frequency and unsubscribe data.
- Android Developers, Notification runtime permission and notification channels.
- Apple, Asking permission to use notifications, UNNotificationInterruptionLevel and App Store Review Guidelines, 4.5.4.
Last reviewed: 23 September 2026.
