Mobile app testing

The testing pyramid remains the most effective model for organising tests. At the base, fast and numerous unit tests that validate isolated business logic. In the middle, integration tests that verify communication between modules. At the top, end-to-end (E2E) tests that simulate complete user flows.

The recommended proportion is roughly 70% unit, 20% integration and 10% E2E. Unit tests run in milliseconds; E2E tests can take minutes. Inverting the pyramid (many E2E, few unit tests) produces suites that are slow, brittle and costly to maintain.

Unit tests validate isolated code units: functions, classes, ViewModels. On iOS, XCTest is the native framework; on Android, JUnit + Mockito. For Flutter, the test package with mockito. For React Native, Jest is the standard.

The key is to architect the app so business logic is testable without depending on UI frameworks or the OS. Patterns like MVVM, Clean Architecture or BLoC (Flutter) separate logic from presentation, enabling fast tests without starting a simulator.

Integration tests verify that modules work correctly together: a ViewModel calling a repository that accesses a real or mocked API, or a UI component reacting to state changes. They are slower than unit tests but catch errors that isolated tests miss.

On Android, Instrumented Tests (using AndroidJUnit4) run on an emulator or real device and allow testing components that depend on the Android framework. On iOS, XCTest UI tests launch the real app and simulate interactions. In Flutter, integration_test allows testing widgets with real dependencies.

E2E tests simulate real user behaviour: opening the app, navigating, filling forms, making purchases, verifying results. They are the most valuable for detecting regressions in critical flows but also the most expensive to write and maintain.

Maestro is an emerging tool that simplifies E2E test writing with a declarative DSL that works on iOS, Android and React Native. Appium remains the most versatile option for projects needing cross-platform support with a single framework. Detox (by Wix) is optimised for React Native with automatic UI synchronisation.

Simulators and emulators are useful during development but do not replace testing on real devices. Differences in performance, gestures, sensors and system behaviour between a simulator and a physical device can hide critical bugs.

Cloud device farms allow running automated tests on hundreds of real devices without purchasing them. Firebase Test Lab (free with limits), AWS Device Farm, BrowserStack and Sauce Labs are the most established options. A practical approach is to select 10–15 devices that represent 80% of your user base (according to analytics) and run the E2E suite on them.

Before publishing a release, a beta testing programme helps catch issues that automated tests don’t: unexpected usage flows, UI confusion, problems on specific devices or under real network conditions.

TestFlight (Apple) allows distributing betas to up to 10,000 external testers and is the iOS standard. Google Play has closed and open testing tracks. Firebase App Distribution offers cross-platform distribution without going through the stores. The key is having a clear bug reporting process and a structured feedback loop.

Production crashes are inevitable. What makes the difference is how quickly you detect, diagnose and fix them. A good crash reporting system captures the stack trace, OS version, device model, app state and the steps leading up to the crash.

Firebase Crashlytics is the de facto standard: free, real-time, with intelligent crash grouping and configurable alerts. Sentry offers crash reporting alongside performance monitoring and detailed breadcrumbs. Bugsnag focuses on stability as a business metric with crash-free session dashboards.

A well-configured CI/CD pipeline runs tests automatically on every pull request, generates staging builds and automates store publishing. This eliminates human error, speeds up the release cycle and ensures every version passes the same quality checks.

Fastlane automates build, signing, screenshot and publishing tasks. GitHub Actions, Bitrise and Codemagic are the most popular mobile CI platforms. A typical pipeline includes: lint, unit tests, integration tests, build, E2E tests on a device farm, beta channel publication and, after validation, production release.