AI-Personalized Push Notifications for Mobile Apps

We deliver mobile push personalization using AI. A news app with 500K users initially had a push notification CTR of 2.1%. After implementing personalization based on collaborative filtering and multi-armed bandit push, it grew to 7–12%—more than triple. The financial impact: such a system can add up to $50K per month for a 500K-user app. Our team provides end-to-end AI push notification personalization: from event collection to serving layer. We process up to 200K events daily with recommendation latency under 50 ms.

We use a proven stack: Swift 5.9 with async/await for iOS, Kotlin with Coroutines for Android, Flutter 3.x for cross-platform. Server-side: Python 3.11 with Surprise (collaborative filtering push), LightGBM (CTR prediction push), and HuggingFace Transformers (text embeddings). Event storage: ClickHouse (analytics) and BigQuery (ML pipelines). For Firebase push analytics and BigQuery push events, we ensure proper logging. All solutions handle peak loads of up to 10K requests per second. Our multi-armed bandit approach is 2x more efficient than traditional A/B testing in identifying winning headlines.

What Data Is Needed for Personalization?

Without behavioral data, personalization is impossible. Minimum event set:

• notification_received — notification displayed
• notification_opened — tap on notification
• notification_dismissed — swiped away without opening
• content_viewed — content viewed
• content_shared, content_saved, content_liked

All events are logged with features: content category, time of day, day of week, device type, OS version, headline length. We use ClickHouse or BigQuery for storage—they're optimized for analytical queries. PostgreSQL won't work beyond 10M events per day.

How Does AI Choose the Right Notification for Each User?

We use collaborative filtering push, content-based push, and CTR prediction push methods.

Level 1: Collaborative filtering push. Idea: users similar to you already clicked on this. Implementation: Matrix Factorization (Surprise or implicit library). Model is trained once daily on the last 30 days of data.

Level 2: Content-based filtering push. Analyze content the user has read: extract keywords and categories via TF-IDF or sentence embeddings (all-MiniLM-L6-v2 model). When new content appears, compute cosine similarity with the user's history.

Level 3: CTR prediction push. Binary classification—will the user tap for each user-content pair. Use LightGBM or XGBoost on tabular features, CatBoost for categorical features. Inference takes tens of milliseconds.

In practice, we start with Level 1 (fast and interpretable), then move to Level 3 when 50–100K events are accumulated.

Personalizing Notification Text

One news story—different headlines for different segments. Not LLM generation on each send (too slow). Approach:

1. Editor creates 3–5 headline variants for one piece of content.
2. Multi-armed bandit push (Thompson Sampling) selects a variant for each user based on their previous CTR with similar headlines.
3. After 24 hours, analyze results and identify the winner.

An LLM (via API) can generate variants in different styles, but the editor selects them.

Why Is Suppression Push Notifications So Important?

On each send, the personalization service:

1. Gets the list of target users.
2. Requests recommendation score from the feature store (Redis with pre-computed vectors).
3. If score is below threshold—notification is not sent (suppression push notifications).
4. If above threshold—selects personalized text.
5. Logs the decision for training. The feature store in Redis is updated via nightly batch and incremental updates. Suppression push notifications reduces opt-out rate and improves relevance. In our practice, suppression reduced opt-out rate by 40% in 3 weeks.

A/B Testing Push and Metrics

Mandatory A/B testing push before global rollout: 10% of users get personalized notifications, 90% get standard ones. Metrics after 2 weeks:

• CTR—primary metric.
• Opt-out rate—did unsubscribes decrease.
• Session starts per notification.
• Revenue per notification.

Firebase A/B Testing + Remote Config covers basic scenarios.

Mobile Client: What Changes

Nothing. Push comes via standard FCM, handled as usual. All logic is server-side. Client only sends behavioral events. Encryption via UNNotificationServiceExtension on iOS.

Implementation Phases

Contact us to discuss details and get a custom plan. Order a free audit—we'll assess personalization potential for your app.

What's Included

• Audit of current notification and analytics system.
• Data pipeline design.
• Model development and training (collaborative/content-based/CTR).
• Serving layer implementation (feature store, suppression push notifications, multi-armed bandit push).
• A/B testing push and metric analysis.
• Documentation and team training.
• Post-launch support.

Our Expertise

Over 7 years of experience in mobile development, certified in ML engineering, with 50+ personalization projects. Trusted by leading apps in news and e-commerce. We guarantee a minimum 20% improvement in CTR or your money back. We comply with App Store and Google Play policies per Human Interface Guidelines. We specialize in push notification optimization. Our team delivers iOS personalized push and Android push personalization.

Assess the personalization potential for your app—order a free audit.

Push Notifications in Mobile App: APNs, FCM, Segmentation, Rich Push

We have implemented push notifications in mobile apps for 50+ projects — from startups to enterprise with audiences of 10M+ users. An irrelevant or technically broken notification is worse than none: the user disables push or deletes the app. According to a Localytics report, push permission rejection on iOS reaches 40% in the first week — the cause is almost always irrelevance, not mechanics. Within 2 weeks after implementing quality segmentation, open conversion increases by 25–30%. Contact us for an audit of your current implementation — we will evaluate the project and propose an optimal stack within one day.

How the Infrastructure Works: APNs and FCM

APNs is the only delivery channel on iOS. Everything else (OneSignal, Braze, Airship) is a wrapper on top of it. APNs accepts requests over HTTP/2, authentication via JWT token (p8 key) or certificate. JWT is preferable: one key for all apps in the account, doesn't expire annually unlike the certificate. For more details, see the official documentation.

A critical point: APNs distinguishes apns-push-type — alert, background, voip, complication, fileprovider, mdm. An incorrect type on iOS 13+ causes background notifications not to wake the app. We've seen projects where content-available: 1 was sent without apns-push-type: background — the app didn't receive silent push on some devices, and the team spent a month looking for an 'app bug'.

FCM on Android works through Google Play Services. For devices without GMS (Huawei, part of the Chinese market), Huawei Push Kit or a direct WebSocket is needed — a separate task. FCM supports data messages (handled in onMessageReceived) and notification messages (the system displays automatically if the app is in the background). Mixing them requires caution: if the notification block has a click_action but the deep link is not registered in the app, tapping the notification simply opens the main screen without navigation.

Why Segmentation Is the Foundation of Effective Push Notifications?

Sending to everyone indiscriminately quickly exhausts user loyalty. Personalized messages are clicked 3 times more often than bulk ones, and proper segmentation reduces churn by 25% (on one project it brought significant additional revenue per quarter). The cost of setting up segmentation in OneSignal or a custom backend depends on the complexity of filters.

Proper segmentation is built on several levels.

Topics are for broad categories: 'new promotions', 'order status updates'. User subscribes via FirebaseMessaging.getInstance().subscribeToTopic("orders"). Simple, but no flexible filtering.

Attribute-based segments — via OneSignal, Braze, or custom backend. We store in the user profile: language, device type, last activity, LTV segment. Notification goes only to those with last_active < 7_days and plan = premium. OneSignal allows building such filters in the interface without code.

Personalized — by specific device_token. It's important to store tokens correctly: the token updates on app reinstall, restoration from backup on a new phone, or resetting settings. On iOS, use UNUserNotificationCenter + didRegisterForRemoteNotificationsWithDeviceToken, save to backend on every launch, not just the first. Otherwise, after 3 months 30% of tokens in the database are outdated.

What Is Rich Push and How Does It Boost Conversion?

A standard notification with title and text is clicked less often than a rich push with image and action buttons — by 3 times. But implementing rich push is a separate task on each platform.

On iOS, rich content requires UNNotificationServiceExtension (to modify payload) and UNNotificationContentExtension (custom UI). The extension runs in a separate process with limited time and memory. If the extension crashes or exceeds the timeout, the system shows the original payload without media. A typical mistake is trying to load an image over HTTP (not HTTPS): ATS blocks the request, the extension silently fails, and the user sees a notification without an image.

On Android with API 26+, notifications are tied to NotificationChannel. If the channel is created with IMPORTANCE_LOW, sound and vibration are unavailable. Different notification types (transactional, marketing) should be in different channels so the user can disable marketing without losing order notifications. BigPictureStyle, MessagingStyle, InboxStyle are templates for expanded notifications. MessagingStyle with Person and avatars is the best choice for chats.

How to Track Delivery and Conversion of Push Notifications?

Sending a notification is half the work. It's important to know: was it delivered, opened, and did it lead to a target action.

FCM returns a MessageId on send, but does not guarantee a delivery callback — by design. For open tracking, custom logic is needed: on notification tap in onMessageReceived or via getInitialNotification() / onNotificationOpenedApp (OneSignal SDK), send an event to analytics with notification_id.

OneSignal provides built-in delivery and CTR analytics. For more detailed analysis — integrate with Amplitude or Mixpanel via webhook on open events. The budget for such a dashboard varies depending on event volume.

How We Implement Push Notifications: Typical Process

1. Audit current implementation — check token storage, update handling, notification types.
2. Design architecture — choose transport (FCM + APNs), segmentation layer (OneSignal/Braze/custom), personalization method.
3. Implementation — write registration code, inbound handling, rich push, deep linking.
4. Testing — send test campaigns, verify delivery on different devices, simulators, regions.
5. Monitoring and analytics — set up dashboard, open and conversion events.
6. Documentation and training — hand over operational materials to the team.

Typical stack: FCM + APNs at transport level, OneSignal or Firebase Notifications Composer for segmentation, custom backend for personalized event-based notifications. For large apps with >1M users, OneSignal has pricing limits — then we use Braze or a custom implementation on AWS SNS.

Timelines depend on complexity: basic FCM+APNs integration with transactional notifications — 1–2 weeks. A full system with segmentation, rich push, analytics, and A/B testing — 4–8 weeks. Order an audit of your current push infrastructure or get a consultation on implementing push notifications in your mobile app — we will contact you within a day and provide an accurate estimate.