Quality at Speed — Testing, Accessibility, and Mobile

Speed Without Sacrifice

February shifted focus from building features to making sure the features we'd built actually worked — for real users, on real devices, with real accessibility needs. Over 300 commits went into quality: mobile responsiveness, accessibility compliance, and a testing strategy that let us keep shipping with confidence.

WCAG 2.1 AA Compliance

We closed every item in our accessibility audit — 33 issues across the platform:

• ARIA labels on all form inputs and interactive elements
• Keyboard navigation support with visible focus indicators
• Mobile navigation patterns that work with screen readers
• Color contrast ratios meeting AA thresholds across all UI elements
• Focus management in modals and dynamic content
• Proper form field associations for assistive technology

Accessibility compliance improved the experience for all users. Better labels, clearer navigation, consistent interaction patterns make the platform easier for everyone.

Mobile-First Rebuild

Early development prioritized desktop. February fixed that across every page: responsive layouts, touch-friendly controls, navigation optimized for small screens. For a platform serving tradespeople who primarily use phones, mobile isn't a secondary viewport. It's the primary one.

Testing Strategy

We built a multi-layer testing approach:

• Unit tests — individual functions and components tested in isolation
• Integration tests — API endpoints tested against real database state
• End-to-end tests — full user flows validated against the staging environment
• Data consistency tests — automated checks that the same data displays identically across every surface where it appears

The data consistency tests were the highest-value investment of the month. When the same information appears on multiple pages, those pages will drift over time. Automated checks that compare output across surfaces catch drift before users notice it.

The Race Condition That Taught Us a Pattern

One of our worker profile pages loaded data from three API calls in parallel: profile details, certifications, and recent activity. When all three returned in the expected order, everything worked. When the activity call returned before the profile call — which happened in roughly one in twenty page loads — the page rendered with the activity tied to the wrong worker briefly, then corrected itself a half-second later.

The fast fix would have been a setTimeout or a spinner that hides the flicker. We didn't do that. We rebuilt the page's data flow so the component doesn't render until all its dependencies are available, regardless of which call finishes first.

The lesson: timeout-based workarounds are tech debt with a delay timer. Every one of them ages into a bug somebody else has to fix.

Usability-Driven Fixes

Structured usability sessions generated more actionable fixes than weeks of internal testing. Each session produced a concrete list of issues — not opinions, but observable problems where users got stuck, confused, or gave up. Every issue was reproduced, fixed, and covered by a test.

The pattern: usability sessions identify the problem, engineering investigation reveals the root cause, the fix addresses the root cause, and a test prevents regression. No band-aids.

CI Integration

All tests run in CI on every push. The API server starts in the CI environment so integration tests can run against real database connections. Tests that require infrastructure unavailable in CI are clearly marked and skipped with documentation explaining why.

A test suite that doesn't run automatically isn't a test suite — it's a suggestion. Every test either runs in CI or has a documented reason why it can't.

The Result

By the end of February, the platform was accessible, mobile-responsive, and covered by a test suite that ran on every commit. This quality foundation made the security hardening and production push in March possible. You can't harden a platform you can't test.