testing · qa craft
Visual regression — Figma states vs rendered output
Every named Figma state has a corresponding screenshot in the visual regression suite. When the pipeline runs, it renders the output and compares it to the stored baseline. If the per-student-marking state looks different from the Figma frame, the test fails. This closes the loop between design and implementation — without it, design drift accumulates silently.
QA's execution craft
QA is not a gate the code passes through at the end. QA is a discipline that runs alongside development — starting before the PR is merged and continuing through the release.
Before the PR merges
QA's first verification happens on the branch, before the PR is approved. This is not a full regression — it is a focused check that the scenarios from the amigos session work as specified. The developer deploys their branch to a preview environment or walks QA through the local build. QA clicks through the prototype, checks each named scenario, and either confirms or raises a gap. A gap raised here costs minutes. A gap raised after merge costs hours.
Exploratory testing — beyond the scenarios
Scenarios cover the cases the team thought of. Exploratory testing covers the cases they didn't. After the scenarios pass, QA spends time using the feature the way a real person would — not following the happy path, but trying the unexpected. What happens if Gal marks half the students and closes the browser? What if she opens the same exam in two tabs? Exploratory testing is a craft, not a random walk — experienced QA testers follow heuristics, document what they tried, and report what surprised them.
The QA report
Before the release gate, QA produces a short report for the slice:
- What was tested (scenario by scenario)
- What was explored (the heuristics and areas covered)
- What wasn't tested (the known gaps — devices not tested, edge cases deferred)
- What surprised (anything that worked differently than expected, even if it's not a bug)
The QA report is a release-gate condition. Without it, the release gate cannot pass.
Accessibility testing
If the ilities in Volume III named accessibility as a requirement for this Epic, the execution discipline is here. Keyboard navigation works. Screen readers announce the right content in the right order. Colour contrast meets WCAG thresholds. Focus states are visible. The automated accessibility checks run in the pipeline; the manual checks — can a real person navigate this with a keyboard? — are part of QA's craft.