QA testing runs
Compliance testing proves twelve people had your app installed. It says nothing about whether the app works. A QA run is the other thing: people using it on purpose, on devices you do not own, writing down what went wrong.
What a comprehensive run covers
A smoke test opens the app and closes it again. A comprehensive run is the opposite: every screen reached, every flow finished, and then those same flows attacked the way real users stumble into them. This is what testers work through.
- Every screen and every flow. Sign-up, sign-in, password reset, the task your app exists to do, settings, sharing, sign-out and account deletion — each completed end to end rather than glanced at.
- Regression on what already worked. Fixes break their neighbours. After each build we re-run the flows that passed last time, so a repair in checkout does not quietly take the login with it.
- Devices you do not own. Budget handsets and flagships, small screens and tall ones, Android 9 through the current release. Fragmentation is where most one-star reviews come from, and you cannot reproduce it on one phone.
- Cold start, jank and battery. How long the first screen takes on a slow device, which lists stutter as you scroll, and whether the app is still draining the battery long after it was backgrounded.
- Bad networks, not just fast ones. Offline, 2G, a connection that dies mid-upload, and airplane mode flipped at the worst possible moment. Apps that have only ever run on office wifi fail here.
- Install, upgrade and reinstall. A clean install, an upgrade over the previous version with real data already in it, and a reinstall after deletion. Upgrade paths are the least tested and the most often broken.
- Interruptions. A call arriving mid-payment, a notification over a half-filled form, rotation, split screen, and the app killed in the background then restored. This is where state quietly disappears.
- Payments and purchases. A successful payment, a declined card, a cancelled sheet and a double tap — each checked against what your backend actually recorded, because a screen that says success is not proof that anything was charged.
- Permissions, including refusals. Granted, denied, and denied now but granted later. An app that crashes when somebody says no to the camera fails Play review as surely as it fails the user.
- Accessibility. TalkBack navigation, large font scaling and colour contrast — the things that make an app unusable for some people, and that Play policy is paying steadily more attention to.
- Language and layout. Text overflowing its button in German, dates in the wrong order, and right-to-left layout in Arabic. We test in the languages we speak, which is most of the ones this site is written in.
- Crashes and ANRs, with the logs. When something dies we capture the logcat trace and attach it to the finding, so you get a stack rather than somebody's description of a stack.
What a report contains
- Steps to reproduce. Numbered, in the order they did them — not “it crashed”.
- Device and version. Model and Android release on every finding, because half of all bugs are one device deep.
- Expected against actual. What they thought would happen, and what did.
- Severity. Blocker, major, minor or cosmetic, so you can plan rather than triage.
- A screenshot. Where a picture settles it.
You can brief them
If you want the sign-up flow hammered, or the payment sheet checked in Swahili, say so when you book. Testers work to the brief and then poke around freely, which is where the surprising findings come from.
Retests are included
Ship the fix and ask for a retest on the same findings. A bug marked fixed that is not fixed is worse than one nobody found, so we check rather than take your word for it.
How deep it goes is your call
A focused run points every tester at one area — the checkout, the onboarding — and goes deep on it. A comprehensive run walks the whole app instead, several testers in parallel on different devices, and usually returns dozens of findings on a first pass. Most teams book one comprehensive run before launch and focused runs after that.