Mobile App Testing: What Makes It Different
A free, hands-on mobile app testing course: emulators vs real devices, Appium vs Espresso vs XCUITest vs Maestro, app uploads, device matrices, CI and debugging failed runs on real devices.
What you will learn
- Mobile App Testing: What Makes It Different
- Emulators, Simulators and Real Devices
- Choosing a Framework: Appium, Espresso, XCUITest, Maestro
- Uploading and Managing App Builds
- Your First Appium Test on a Real Device
- Your First Maestro Flow on TestingBot
- Running Espresso and XCUITest Suites in the Cloud
- Device Matrix, Parallel Runs and CI
- Debugging Failed Mobile Test Runs
- Knowledge Check
On this page
Last reviewed November 2026 against WebdriverIO 10.2, the TestingBot CLI (@testingbot/cli) 1.5.0 and the TestingBot sample apps.
Mobile app testing is checking that a native or hybrid app installs, starts and behaves correctly on the phones and tablets your users actually carry, under the conditions they actually use them in. Most of the techniques come from web testing: locate an element, act on it, assert on the result. What changes is everything around the test. A browser test runs in a sandbox that looks the same on every machine; an app test runs on a device whose hardware, OS version, permissions, network and battery are all part of the result.
This course is the strategy layer. It explains what to test on which kind of device, how to choose between Appium, Espresso, XCUITest and Maestro, and how to get a suite running on cloud devices and in CI. Every chapter has a small, real example that ran on TestingBot while the course was written. When you want depth in one framework, the Appium with WebdriverIO course and the Maestro course take over.
Fragmentation: the matrix is the product#
On the web you test a handful of browser engines. On mobile you test combinations of:
- OS versions. Android apps commonly support several major versions at once, and each one changes permissions, background limits or notification behaviour. iOS users upgrade faster, but you still have users one or two major versions behind.
- Manufacturers and skins. Samsung, Google, Xiaomi and others ship their own launchers, keyboards, battery savers and system dialogs on top of Android. A flow that works on a Pixel can stall on a phone whose battery saver kills your background sync.
- Screen sizes and densities. Small phones, large phones, foldables and tablets. Layout bugs (a button pushed below the fold, text truncated in German) appear only on some of them.
- Hardware. Cameras, biometrics, GPS, memory. A low-memory device kills your app in the background far more often than the flagship on your desk.
You cannot test every combination, so the real skill is choosing a small matrix that covers your users. Chapter 7 shows how to build one from your analytics instead of guessing.
Permissions, interrupts and the app lifecycle#
A web page never gets a phone call. An app does, and many bugs live in the transitions:
- Runtime permissions. Camera, location, notifications and contacts are granted (or denied, or "only this time") at runtime through system dialogs your app does not own. Tests need a strategy: grant everything up front, or drive the dialog deliberately when the permission flow is what you are testing. TestingBot documents both approaches for Appium in Handling permission pop-ups.
- Interrupts. Incoming calls, notifications, the user switching apps, the screen locking. The OS may pause, background or kill your process, and the app has to restore its state when it comes back.
- Install and upgrade. A fresh install and an upgrade over last month's version start from different data. Testing app upgrades covers installing the old build, then the new one, in one session.
Networks are not your office Wi-Fi#
Users open your app in a lift, on a train and abroad. Slow and flaky connections expose missing timeouts, spinners that never stop and requests that are sent twice. Region matters too: content, pricing and consent banners can change by country. TestingBot can throttle the device's network to presets such as 3G or Edge, or simulate airplane mode, and set an IP geolocation; see Throttle Network and Geolocation Testing.
Support reports that photo uploads "sometimes vanish" for users on trains. All your tests pass on the office Wi-Fi. What is the most useful next test?
The symptom points at network conditions, so reproduce them on a device. A faster phone hides the problem, longer timeouts only wait it out, and an API test skips the app's retry and error handling, which is where the bug lives.
What you will build#
The examples all use the TestingBot sample apps, so you can run them without your own app:
-
The calculator sample app (
https://testingbot.com/appium/sample.apkfor Android,sample.zipfor the iOS Simulator andsample.ipafor physical iPhones): two number fields and a sum. Chapters 4, 5 and 7 drive it with Appium and Maestro. - The Espresso demo app: a calculator with an Espresso test suite. Its 1.0.0 release ships both APKs that an Espresso run needs. Chapter 6 runs it, and chapter 8 debugs the one test in it that fails.
Over nine chapters you will:
- Decide when an emulator or simulator is enough and when only a real device will do.
- Choose a framework with honest trade-offs.
- Upload app builds and keep them current.
- Run a first Appium test on a real Android phone.
- Run a first Maestro flow.
- Run an Espresso suite and understand how XCUITest runs work.
- Build a device matrix, respect your parallel limit and wire it into CI.
- Debug failed runs from videos and device logs.
- Check what you learned.
Prerequisites#
- Node.js 22 or newer for the Appium examples (WebdriverIO 10 requires it) and the TestingBot CLI.
-
A TestingBot account for the cloud runs. Your key and secret are in the account settings; the examples read them from
TB_KEYandTB_SECRET. - No Android Studio or Xcode is needed to follow along: every run in this course uses prebuilt sample apps on cloud devices. You need them later to build your own app and native test suites.
Basic JavaScript helps for chapters 4 and 7. Everything else is YAML, shell commands and reading Java test code.