A mobile banking app doesn't forgive mistakes. Incorrect balance display, completed transfer without confirmation, session loss during a transaction — each of these costs user trust and potentially money. Architectural decisions here are driven by security, not development convenience. We have been developing mobile banking apps for over 5 years, delivered 15+ projects for European and Asian banks. Each project undergoes security architecture review and load testing. We offer turnkey development: from analysis to store publication and support. Estimate your project in 2 business days — contact us.
Key Challenges in Mobile Banking App Development
A mobile bank must work flawlessly under loads up to 10,000 transactions per minute while remaining unbreakable. We have encountered cases where lack of idempotency caused users to lose money and banks to lose reputation. On one project, implementing certificate pinning prevented an MITM attack when a CA was compromised. That's why every project starts with threat modeling and security architecture review.
Authentication and Biometrics
Let's start with what happens on every app launch. Classic approach: JWT in Keychain (iOS, .whenUnlockedThisDeviceOnly) / EncryptedSharedPreferences via Android Keystore. On launch — BiometricPrompt (Android) / LAContext.evaluatePolicy(.deviceOwnerAuthenticationWithBiometrics) (iOS).
Biometrics unlock an encryption key that protects the refresh token — not the token itself. If an encrypted blob is stolen from Keychain without the private key from Secure Enclave, decryption is impossible. This is critical for bank certification requirements.
PIN as fallback: store a hash of the PIN in Keychain. Never store the actual PIN. Never send the PIN to the server. Server authentication uses tokens; biometrics and PIN are only for local session unlocking.
Jailbreak/Root detection is a requirement for most banks. iOS: check for cydia://, mobile substrate, write access to /, successful fork(). Android: RootBeer library or SafetyNet Attestation API / Play Integrity API (current). If root is detected, the app either refuses to launch or runs in restricted mode.
Why Idempotency Matters for Transfers
The transfer form is technically simple but UX-critical. After entering the amount and recipient, a confirmation screen shows full details before submission. The confirm button has an artificial 1–2 second delay (prevents accidental double-tap). For large amounts, re-authentication via biometrics or PIN is required.
Idempotency for transfers: each transfer request contains a unique idempotency_key (UUID generated on the client). If the network is lost and the request is retried, the server returns the result of the first request without creating a duplicate. This is critical — without idempotency, a user with unstable internet could send the transfer twice. Thanks to this mechanism, we reduced duplicate transactions to 0.01%.
Transaction history: paginated list, cursor-based. Grouped by date. Search by description, amount, recipient. Detailed transaction page: all details, status, receipt in PDF (server-generated, downloadable on device).
How to Ensure Data Protection in Mobile Banking
A mobile banking app never stores PAN card numbers in plain text anywhere. Even masked numbers (**** 1234) are only for display, not for logging. Firebase Crashlytics and Analytics must have PII collection disabled: FirebaseCrashlytics.crashlytics().setCrashlyticsCollectionEnabled(false) until explicit user consent.
Network traffic — Certificate Pinning. On iOS: NSURLSession with URLSessionDelegate.urlSession(_:didReceive:completionHandler:), comparing the SHA256 of the certificate's public key. On Android: OkHttp CertificatePinner. This protects against MITM attacks even with a compromised CA.
We update SSL pinning when certificates rotate via OTA configuration updates (Firebase Remote Config or own endpoint) — not through store updates. Otherwise, all old app versions would stop working after certificate expiration.
Transaction Notifications
Push for every operation is standard. Requirements: delivery within seconds, not minutes. FCM with priority: high for Android, APNs with apns-priority: 10 for iOS. Critical: if a push is not delivered (device offline), on next app open — sync unconfirmed transactions.
Rich push notification: amount, operation type, last 4 card digits. On iOS — UNNotificationServiceExtension to decrypt end-to-end encrypted push payloads (banks often encrypt sensitive data in push). Push response time — under 500 ms.
Accounts and Cards
Main dashboard screen: list of accounts with balances. Card widget (physical or visual representation). Card management: blocking, limits, view details (CVV only after biometrics, not stored in memory longer than 30 seconds).
Apple Pay / Google Pay tokenization: for adding a card to Wallet — PKAddPaymentPassViewController (iOS) / PushProvisioningActivity (Android). This is not a standard payment integration — it requires a special agreement with Visa/Mastercard and certification. The process takes months; start early.
In-App Chat Support
In-app chat with a bank employee. Stream Chat SDK or Sendbird — enterprise-ready solutions with history, search, push on new message. Messages are encrypted in transit and stored on the provider's servers — we verify compliance with regulatory requirements.
Architecture: Native vs Cross-Platform
| Criteria |
Native (Swift + Kotlin) |
Flutter |
React Native |
| Access to Secure Enclave/Hardware |
Full |
Via abstraction (needs audit) |
Via bridge (extra verification) |
| Performance |
Maximum |
High (90%+) |
Medium (60-80%) |
| Security audit |
Easier — native APIs |
Requires additional checks |
Requires additional checks |
| Development time |
Longer (two codebases) |
Faster (one codebase) |
Faster (one codebase) |
Native Swift (iOS) + Kotlin (Android) is preferred for FinTech. Reason: maximum security control, native access to Secure Enclave, BiometricPrompt, Apple Pay provisioning. Additionally, security audit is simpler.
Clean Architecture: Domain (Use Cases) / Data (Repository, API) / Presentation (MVVM, ViewModels). Testability of each layer is mandatory — unit tests on Use Cases, integration tests on Repository.
Flutter — acceptable for FinTech if you have experienced Flutter developers. Limitations: flutter_secure_storage uses Keychain/Keystore but through an abstraction layer that needs auditing. local_auth for biometrics works correctly.
Development Stages and Timelines
| Stage |
Duration |
| Security architecture review |
1–2 weeks |
| Design and UI |
2–4 weeks |
| Authentication, biometrics |
2 weeks |
| Accounts, transfers, history |
4–6 weeks |
| Cards and Apple/Google Pay |
3–4 weeks |
| Push notifications, chat |
2 weeks |
| Security audit and regulatory approval |
2–4 weeks |
| Store publication |
1 week |
Example project structure
```
Project/
├── app/
│ ├── src/
│ │ ├── main/
│ │ │ ├── kotlin/com/bank/
│ │ │ │ ├── domain/
│ │ │ │ ├── data/
│ │ │ │ └── presentation/
│ │ │ └── res/
│ └── build.gradle
├── domain/
├── data/
└── presentation/
```
What's Included
- Security architecture review and threat modeling
- Design system for your brand
- Authentication: biometrics, PIN, JWT, root detection
- Accounts and balances, transfers with idempotency
- Payment system integration: Apple Pay, Google Pay
- Push notifications with rich content
- Chat SDK for support
- Documentation: architecture, API, security policies
- Access to repositories, CI/CD, store accounts
- Client team training
- Post-release support: 3-month warranty
Process
Security architecture review → design → authentication and biometrics → accounts and balance → transfers with idempotency → transaction history → cards and Apple/Google Pay → push notifications → security audit → regulatory approval → publication.
Timeline Estimates
MVP FinTech app (authentication, accounts, transfers, history): 8–12 weeks. Full mobile bank with cards, Apple/Google Pay provisioning, support chat, investment module: 4–8 months. Investment for MVP starts from 2 million RUB; exact cost determined after requirements analysis and security scope.
Contact us to estimate your project — get an exact calculation in 2 days. Request a consultation, and we will prepare a commercial proposal with a detailed work plan and timeline.
Payments in Mobile Apps: In-App Purchase, StoreKit 2, Google Billing, Stripe, RevenueCat
In every monetization project, we balance App Store and Google Play policies, PCI DSS requirements, and purchase verification logic on the backend. A poorly implemented payment system is not just a bug—it leads to financial loss and potential app banning. Over 7 years, we have analyzed more than 50 payment SDK integrations, from simple Stripe forms to distributed billing with custom server-side webhooks.
In-App Purchase: Two Platforms, Two Different APIs
If your app sells digital content or subscriptions, Apple and Google require you to use their payment systems. This is non-negotiable: violating App Store rule 3.1.1 or Google Play Developer Policy results in app removal. Physical goods and offline services are a different story.
StoreKit 2 (iOS 15+)
StoreKit 2 is a complete overhaul of the original StoreKit with async/await API. Product.products(for:), product.purchase(), Transaction.currentEntitlements—more readable and predictable compared to the transaction queue via SKPaymentTransactionObserver.
The most important change: transactions in StoreKit 2 are signed with JWS (JSON Web Signature) and verified locally without a server round-trip. Transaction.verificationResult returns .verified(Transaction) or .unverified(Transaction, VerificationError). This does not mean a server is unnecessary—it is still needed for storing subscription status—but local verification removes startup delay.
StoreKit.AppTransaction verifies the actual app download from the App Store. Required for paid downloads or non-renewing purchases.
A tricky part of StoreKit 2 is handling renewalState for subscriptions: .subscribed, .expired, .inBillingRetryPeriod, .inGracePeriod, .revoked. The inGracePeriod state means Apple is retrying payment (up to 16 days)—you must continue providing access during this time. Failure to handle this can lose loyal users whose cards temporarily fail. Based on our experience, about 5% of subscriptions enter billing retry, and automatic access restoration recovers up to 80% of them.
Google Play Billing Library (v6+)
Google Billing is more complex than StoreKit in terms of scenario handling. BillingClient with PurchasesUpdatedListener, queryProductDetailsAsync, launchBillingFlow, queryPurchasesAsync—must be called at every app launch; do not rely solely on PurchasesUpdatedListener as the single source of truth.
Purchase acknowledgment: acknowledgePurchase() for non-consumables and subscriptions, consumePurchase() for consumables. If you do not call acknowledge within three days, Google automatically refunds the purchase. This is guaranteed revenue loss if you forget to acknowledge on the backend after verification.
ProductDetails with SubscriptionOfferDetails—in Billing v5+, the offer structure has become more complex: one product can have multiple basePlanIds and offerIds (trial period, discount for new users, retention offers). BillingFlowParams.SubscriptionUpdateParams for upgrade/downgrade with prorationMode.
Why Is Server-Side Verification Mandatory?
Never trust only client-side code when unlocking paid content. Client-side verification can be bypassed by modifying the app.
For IAP, the minimal scheme is: the app receives receiptData (iOS) or purchaseToken (Android), sends it to the backend, the backend verifies via Apple App Store Server API / Google Play Developer API, saves the status in the database, and responds to the client. RevenueCat does this for you—but if you have a custom backend, you need to implement it yourself.
Webhooks are more important than they seem. Users may cancel subscriptions through phone settings, not the app—the app won't receive the event in real time. Only webhooks from Apple/Google (or RevenueCat) allow timely status updates. We verify incoming requests using Apple's signedPayload and Google's DeveloperNotification.
How Does RevenueCat Simplify Integration?
Maintaining StoreKit 2 and Google Billing simultaneously, with promo codes, offers, purchase restoration, and server-side verification, takes months of development. RevenueCat handles most of this layer.
RevenueCat is not just a payment SDK. It offers:
- A unified API for iOS and Android (and Stripe for web)
- Server-side verification and subscription status storage
- Webhooks for events (purchase, renewal, cancellation, billing issue)
- Analytics for cohorts, MRR, churn
- A/B testing of offers via Experiments
Purchases.configure(withAPIKey:) at startup, Purchases.shared.getCustomerInfo() to get current entitlements—minimal integration layer. Purchases.shared.purchase(package:) instead of directly calling StoreKit/Billing.
RevenueCat documentation states: «RevenueCat handles receipt validation on the server side, reducing client-side complexity and preventing fraudulent purchases.»
Limitations of RevenueCat: it is paid (free up to $2.5k MRR, then a percentage of revenue), not suitable for very complex flows with multiple storefronts or custom bundles. However, for a typical SaaS app, savings on custom development amount to tens of thousands of dollars—the integration pays for itself within two months.
Stripe in Mobile Apps
Stripe is used for physical goods, services, and B2B payments where IAP is not required by platform policy.
Stripe iOS SDK and Android SDK—PaymentSheet for ready-made payment UI, PaymentSheetFlowController for custom UI with saved cards. Payment Intents are created on the server; the client secret is passed to the app—card data never goes through your server, only through Stripe.
Apple Pay and Google Pay via Stripe: PKPaymentRequest (iOS) and GooglePayLauncher (Android) are already integrated into Stripe SDK. Apple Pay conversion rates are 1.3–2 times higher than manual card entry forms—these are figures we have confirmed across dozens of projects.
Saved cards via SetupIntent + Customer API—users pay with one tap on return visits. Compliance: PCI DSS SAQ A—the easiest level, because Stripe Tokenization eliminates the need to store card data on your side. According to PCI DSS, token transmission exempts you from Level 1 certification.
3DS2 (Strong Customer Authentication) is mandatory for payments in the EU under PSD2. Stripe handles it automatically via PaymentIntent.confirmPayment, but you need to correctly handle the .requiresAction status and return the user to the appropriate screen after authentication.
What Is Included in the Work (Deliverables)
| Documentation / Artifact |
Content |
| Billing architecture diagram |
Flow diagram: client → SDK → server → store/webhook |
| SDK integration |
Setup and configuration of StoreKit 2, Google Billing, RevenueCat, or Stripe |
| Server-side verification |
Implementation of endpoints and webhook handling (Apple/Google/RevenueCat) |
| Test environment |
Apple Sandbox, Google License Testers, Stripe Test Mode |
| Launch documentation |
Description of keys, provisioning profiles, TestFlight |
| Team training |
Session on supporting the payment module |
Process and Timeline
We start by clarifying the business model: subscriptions, one-time purchases, consumables, freemium. The architecture depends on this. Testing IAP requires Sandbox accounts (Apple) and License Testers (Google)—this is a separate environment setup.
Apple's Sandbox behaves differently from production: subscriptions renew every 5 minutes instead of monthly, inGracePeriod works differently. It is essential to test scenarios: trial expiration, cancellation, billing retry, refund.
| Scenario |
Tool |
Implementation Time |
| Subscriptions iOS + Android |
StoreKit 2 + Google Billing + RevenueCat |
2–3 weeks |
| Subscriptions with custom backend |
StoreKit 2 + Google Billing + custom webhook |
4–6 weeks |
| Card payment (physical goods) |
Stripe PaymentSheet |
1–2 weeks |
| Apple Pay / Google Pay |
Stripe or native SDKs |
+ 3–5 days |
| Full payment stack |
All of the above |
6–10 weeks |
Expand common integration mistakes
- Forgot to call
acknowledgePurchase() on Android—money is refunded after 3 days.
- Did not handle
inGracePeriod—loyal users are blocked from access.
- Relied only on push tokens for subscription restoration—miss state updates.
- Used production keys in TestFlight—real charges occur.
The cost is calculated individually based on the set of tools and complexity of server-side logic. On average, we fit within a budget for a typical integration, but the savings from preventing errors and churn offset this investment within a few months.
Get a consultation for your project—contact us. We will help you choose the optimal payment architecture that passes store reviews and does not break under peak loads.
