Visual Responsive Regression Testing

A visual regression is a change nobody intended. Catch it by comparing screenshots automatically — not by manually checking every screen after every change.

1. Define the screenshot inventory

List every unique screen/view that needs visual regression coverage. For each view:

• Route/URL
• Authentication state required
• Data state (empty, populated, error state)
• Breakpoints to capture: all breakpoints from responsive-breakpoint-strategy

Priority: cover critical path screens first, then secondary screens.

2. Capture baseline screenshots

For the first run, capture baseline screenshots at every defined breakpoint for every screen. Store these as the "approved" reference images. These are the ground truth.

3. Generate Playwright visual regression scaffolds

Generate test files using Playwright's screenshot comparison:

// Run: npx playwright test visual-regression/
// First run: npx playwright test --update-snapshots (captures baseline)
test('Homepage at mobile breakpoint', async ({ page }) => {
  await page.setViewportSize({ width: 375, height: 812 });
  await page.goto('/');
  await expect(page).toHaveScreenshot('homepage-mobile.png', {
    maxDiffPixels: 100 // allow minor anti-aliasing differences
  });
});

4. Run comparison against baseline

After any code change, run the visual regression suite. The tool compares current screenshots pixel-by-pixel against the baseline. Flag any difference above the threshold.

5. Classify differences

Not all visual differences are bugs:

• Intentional change: design was updated intentionally → update baseline with --update-snapshots
• Regression: unintentional change that broke existing behavior → must be fixed
• Environment noise: font rendering differences, anti-aliasing → adjust threshold or ignore

For each flagged difference, classify it before acting.

6. Audit responsive transformations

Beyond pixel comparison, verify the transformation table from adaptive-component-behavior:

• At bp-sm (mobile): is the sidebar a bottom-nav? Is the table a card-list? Is the modal fullscreen?
• At bp-md (tablet): are intermediate transformations correct?

This cannot be done by pixel comparison alone — it requires semantic assertions (check that the element with role navigation has the correct structure at each breakpoint).

7. Report and escalate

For each regression:

• Screenshot before and after (diff highlighted)
• Breakpoint affected
• Route/component affected
• PRD requirement ID if the regression violates a specified behavior
• Recommended fix

Completion Criteria

• Baseline screenshots captured for all critical screens at all breakpoints
• Visual regression suite runs in CI (or on-demand) after every change
• All regressions classified (intentional vs bug vs noise)
• Responsive transformation audit passed at each breakpoint
• Zero unresolved regressions at ship time

Output

• Playwright visual regression test files (visual-regression/*.spec.js)
• Baseline screenshot library
• Regression report (diff images + classification + PRD reference)

Anti-patterns

• Setting the diff threshold to 0 (every font-rendering difference becomes a false positive)
• Setting the threshold too high (real regressions get masked)
• Only testing at one or two breakpoints
• Not updating baselines after intentional design changes (leads to permanent failures)
• Relying only on pixel diff — missing the semantic transformation validation entirely
• Not running regression tests after every deploy (defeats the purpose of automated regression)