- QA is the difference between software that works in a demo and software that works for customers.
- Here is what testing actually involves and what to demand from your vendor.
Every software quote includes a line for QA, and many clients quietly wonder whether it is padding. It is not: industry studies consistently show that a bug caught in development costs a fraction of the same bug caught in production, where it arrives with customer complaints, lost transactions and emergency fixes. Here is what that QA line actually buys.
The Layers of Testing
- Unit tests: automated checks on individual functions, written by developers as they code. Thousands run in seconds on every change.
- Integration tests: verify that components work together — the checkout really talks to the payment gateway and the database.
- End-to-end tests: automated scripts drive the real application through complete user journeys.
- Manual and exploratory testing: human testers probe edge cases, odd inputs and real-device behavior that scripts miss.
- Non-functional testing: performance under load, security scanning, accessibility and cross-browser checks.
Manual vs Automated: Not a Contest
Automation excels at regression — proving that yesterday's features still work after today's changes — and pays for itself on any product with a long life. Manual testing excels at judgment: does this flow make sense, does this error message help, does the app feel right on a cheap Android phone? Professional teams do both; vendors who claim to need neither are shipping their testing to your customers.
What Good QA Looks Like From the Client Side
- A written test plan listing what will be verified before launch.
- Bug tracking you can see, with severity levels and resolution status.
- A staging environment where you perform acceptance testing yourself.
- Automated test runs on every code change — ask to see the pipeline.
- A defined device and browser matrix for anything customer-facing.
What QA Costs
Healthy projects allocate 15–25% of total effort to testing activities. Dedicated QA engineers bill below developer rates, so the line is cheaper than it looks — and dramatically cheaper than the alternative. A payment bug that survives to production can cost more in refunds and reputation in one weekend than the entire QA budget.
Questions to Ask Any Vendor
- Who tests — developers only, or dedicated QA?
- What is automated, and can we see test results per release?
- How are bugs triaged after launch, and what does the warranty period cover?
- How do you test payments, emails and third-party integrations safely?
Frequently Asked Questions
Can QA guarantee zero bugs?
No — testing reduces risk, it cannot eliminate it. The goal is that remaining bugs are rare, minor and quickly fixed, never data-losing or payment-breaking.
Is a bug-fix warranty standard?
Reputable agencies include 30–90 days of free defect fixes after launch. Confirm the period and what counts as a defect versus a change.
When should QA start?
With the first sprint, not the last. Testing bolted onto the end of a project finds problems when they are most expensive to fix.
Want software that survives contact with real users? See our development process or ask us about our QA practice.
