QA Risk & Release Readiness Audit · Performance & Scalability Assessment · Accessibility
Release decision: from 14 defects to a clear Go/No-Go
Browser, accessibility and performance tests, combined with the API results, turned into a one-page release decision: v1.0 No-Go, v1.1 Go, with the reasons in business terms.
Demonstration on synthetic data in my own demo project, not client work.
Scenario
The same shop, the day before launch. The business is not asking "how many tests passed?" but "can we release, and what is the risk if we do?" The answer has to come from the test results, in a form a non-technical owner can sign off.
Risks
- Phone shoppers unable to complete checkout
- Pages too slow during a campaign spike
- Screen-reader and low-vision users unable to sign in or read prices (WCAG 2.2)
- Malicious input running in other shoppers' browsers (stored XSS)
- Tax under-collected by a cent on many orders
Tests
- 12 Playwright tests on desktop and a phone screen size, with axe-core WCAG 2.2 scans
- k6 load test: 50 simultaneous users, with the limit agreed up front (95% of requests under 800 ms)
- 18 requirements traced to their tests, and 7 business risks rated by impact × likelihood
- Verdict rules agreed before testing: any open Critical or High defect means No-Go
Defects found
| ID | Defect | Type |
|---|---|---|
| D14 | Customer name runs as code in the browser (stored XSS) | High |
| D11 | "Place order" off-screen on phones | Medium |
| D12 | Unlabelled fields and low-contrast prices | Medium |
| D10 | Product list slows to ~1.5 s at 50 users | Medium |
| D04 | Tax truncated instead of rounded | Medium |
Evidence
- The Go/No-Go report explained — Verdict rules, risks and traceability
- v1.0 report: No-Go (PDF) — Generated from the test results, never edited by hand
- v1.1 report: Go (PDF) — The same report after the fixes
Fix
Fixes verified in v1.1, then the report rebuilt from the new results. CI builds both reports on every run and fails unless v1.0 is No-Go and v1.1 is Go.
Result
- v1.0
- No-Go: 14 open defects (6 Critical or High), 3 of 18 requirements pass
- v1.1
- Go: 18 of 18 requirements pass, all 14 defects verified fixed
- Speed at 50 users (95th percentile)
- About 1.5 s on v1.0, about 3 ms on v1.1
Lessons
- Agree the verdict rules before testing starts, so the decision follows from the results and is not negotiated afterwards.
- Phone layout defects pass every desktop test: the hidden "Place order" button only appears at phone width.
- A speed number means little without a limit. 1.5 s only became a finding once 800 ms had been agreed as the target.
Need this for your product?
Get a suggested scope and starting price in four questions, or tell me about your project.
Next case study: AI support features: a chatbot, a help-centre assistant and an order agent →