QA and software testing

QA testing services that hand your developers bug reports they can act on

We write the test plan, run manual and automated tests on real devices, and send back reports that tell a developer how to reproduce each problem.

Get your testing scoped in writing
Trusted by the world's fastest-growing brands

What QA is

Testing that tells your team what to fix next

QA testing services put a piece of software through the situations its users will create and report what breaks. The work covers what to test and in what order, the tests themselves, and the reports that let a developer fix a problem without a meeting.

Bug log1 high · 3 medium · 1 lowTB-212Approved swap stays pendingHighFound in automated run · webStepsExpectedActualRecordingTB-213Push alert opens home screenMediumTB-214Approve button cut offMediumTB-216Swap buttons under 24 × 24MediumTB-215Shift list stuttersLow

A plan that follows the risk

The test plan ranks each part of the product by what a failure would cost, so sign-in, shift swaps and payroll export get the deepest coverage and a settings screen gets a lighter pass.

Tests that run again

Checks you will repeat on each release become automated suites in your repository. Anything that needs a human eye, such as a layout on a small phone, stays a manual test.

Reports a developer can use

A bug report lists the steps, the expected result, the actual result, the device and a recording, so the person fixing it can see the problem on their own screen.

Two ways to test

QA from a separate team or testing by the build team

The two differ in who knows the code and who is looking at it fresh.

Build team tests its own workDevelopers or in-house testersSeparate QA teamA second set of eyes on the product
Who knows the codeThe people who wrote the code know where it is fragile and can test those parts directlyTesters start from the product’s behavior and the test plan, and read code when a report needs it
What gets foundLogic errors found while writing the code, through unit tests and code reviewProblems a user would hit when following paths the authors did not think of
Whose code it can beThe team’s own codeYours, another vendor’s or ours, since the testers have no stake in the code
StaffingPermanent staff who learn the product in depth over the yearsBrought in for a release, a project or a handover, then scaled down
Best whenA mature team has the time and the tooling to cover testing itselfA release is close, a vendor is handing over software, or no one on staff owns testing
Where independent QA fits

You may already have developers who write unit tests. Independent QA sits above that layer, with the browser, the phone and the user’s path through the product. If another vendor built your software, a written test report is also a record of what you accepted on the day you took it over.

Rules testing works inside

Six facts that decide what gets tested

App store requirements, accessibility standards and the tools themselves. Sources under each.

  1. Apple reviewApp Store

    Apple rejects app bundles that crash or show obvious technical problems.

    Guideline 2.1(a) asks developers to test an app on-device for bugs and stability before submitting it, and to provide working demo account details if the app has a login. A release plan therefore includes a pass on physical iPhones and a demo account that works.

  2. Android vitalsGoogle Play

    Google Play sets bad-behavior thresholds of 1.09% for crashes and 0.47% for unresponsive screens.

    Google asks developers to keep the user-perceived crash rate under 1.09% and the ANR rate under 0.47% overall to maximize a title’s visibility. The limit for a single phone model is 8% on both, so one weak model can matter even when the overall rate looks healthy.

  3. WCAG 2.2W3C

    The W3C advises using WCAG 2.2, which also meets 2.1.

    WCAG 2.2 adds success criteria to 2.1, and content that conforms to 2.2 also conforms to 2.1. Our accessibility pass tests against 2.2 level AA, which includes criterion 2.5.8, a minimum target size of 24 by 24 CSS pixels, with exceptions.

  4. ADA guidanceDepartment of Justice

    The Justice Department says it has no regulation setting detailed web standards and points to WCAG as helpful guidance.

    The Department’s web guidance says the ADA applies to services offered on the web. It names WCAG and the Section 508 Standards as existing technical standards that provide helpful guidance, and says the guidance itself has no legally binding effect. Our report records results against WCAG 2.2 and carries no legal opinion.

  5. Common failuresWebAIM

    WebAIM’s 2026 report found an average of 56.1 detectable accessibility errors on a home page.

    The WAVE engine scanned one million home pages, and low-contrast text appeared on 83.9% of them. WebAIM adds that no automated tool detects all conformance failures. For that reason a scan is one part of our accessibility pass, alongside keyboard and screen reader testing.

  6. Browser coveragePlaywright

    Playwright runs one test suite in Chromium, WebKit and Firefox.

    Playwright Test supports all three engines on Windows, Linux and macOS, locally or in CI, with mobile emulation for Chrome on Android and Mobile Safari. Playwright’s WebKit approximates Safari, and emulation approximates a phone, so the physical-device pass runs on real hardware in BrowserStack.

What we test

Six kinds of testing, and what each produces

A scope picks the ones your release needs. The test plan says which, and in what order.

Test plan, release 4.2AgreedSign-in and resetHigh14Shift swapHigh22Payroll exportHigh18Push notificationsMedium9Offline scheduleMedium6Settings and profileLow757 automated19 manual76 cases

Test strategy and plans

A written plan that lists the parts of the product, ranks each by what a failure would cost, names the method for each and sets the conditions for calling a release ready. You receive: the test plan and the case list.

Risk rankingScopeExit criteria
9:41Swap shiftSat, Oct 114–10 p.m.OfflineSwap queuedRequest swapSession noteCharterLose signal mid-swapSwap queued offlineSent on reconnectNo duplicate sent

Manual and exploratory testing

A tester works through the product as a new customer would, then goes off the script with a question in mind, such as what happens when a phone loses signal halfway through a shift swap. You receive: the session notes and a bug report for each defect.

Pull request checks2 failingPlaywright · web UITB-21229 / 30swap-approval.spec.tsTraceScreenshotPlaywright · API18 / 18Appium · mobileTB-2138 / 955 of 57 passed2 failed

Automated tests and CI

Browser and API tests written in Playwright or Cypress, and Appium tests for phones, committed to your repository and wired into your build service so a failing check shows on the pull request. You receive: the suites and the pipeline configuration.

Device matrixReal devicesiPhone 15iOS 18PassiPhone SE (3rd gen)iOS 17TB-214Galaxy S24Android 15PassGalaxy A15Android 14TB-215

Real-device coverage

Tests on physical phones and tablets in BrowserStack, chosen from your own analytics so the matrix reflects the devices your users hold. You receive: the device matrix, screen recordings of each failure and the OS versions covered.

Accessibility passWCAG 2.2 AASwap request · webSat, Oct 11 · 4–10 p.m.24 × 24 target2.5.8 Target sizeSwap buttons · webTB-2161.4.3 ContrastSchedule gridPass

Accessibility testing

Automated scans, keyboard-only walkthroughs and screen reader sessions with VoiceOver and TalkBack, each result tied to a WCAG 2.2 success criterion. You receive: the accessibility report, with the criterion, the screen and a suggested fix for each failure.

Load test resultsTest environmentResponse timeError rateWhere it slowedDatabase firstSimulated users, rising

Performance and load testing

Scripts that simulate a stated number of users signing in, opening schedules and requesting swaps, run against a test environment, with response times and error rates recorded as the load rises. You receive: the load script and a results sheet with the point where it slowed.

How it runs

Plan, automate, cover, report

No stage goes live until you have approved it.

test-plan.csv

Test plan, release 4.2

Sample · Sample Tallyboard shift-scheduling app
AreaRiskMethodCases
Sign-in and password resetHighAutomated browser tests14
Shift swap and manager approvalHighAutomated browser tests, exploratory22
Payroll exportHighAutomated API tests18
Push notificationsMediumAppium on real devices9
Offline schedule viewMediumExploratory6
Settings and profileLowManual checklist7
76 cases: 57 automated (14 sign-in, 16 shift swap, 18 payroll export, 9 push) and 19 manual (6 shift swap, 6 offline view, 7 settings). Low-risk areas get a manual pass only.

Our stack

The tools behind your test runs

What the tests are written in, run on and read with. Pick one for the detail.

BrowserStack
Why BrowserStack

BrowserStack rents physical phones, tablets and browsers over the internet, so a test can run on hardware you do not own.

How we use it

The device matrix is built from your own analytics, and failures are recorded on the real phone.

Device rangeReal devicesCloud
BrowserStackYes
Firebase Test LabYes
Sauce LabsYes
Local onlyNo

Questions

QA testing services, answered

Start here

Get your testing scoped in writing

Tell us what is being tested and who built it. We’ll send back a written scope.