Sixteen accessibility problems were found. They trace back to four changes. After those four changes a re-test returned zero.
| Issue | Severity | Instances |
|---|---|---|
| Text is too faint to read (colour contrast) | Serious | 16 |
People with low vision, reduced contrast sensitivity, or anyone using a phone in daylight cannot reliably read this text. Roughly one man in twelve has some form of colour vision deficiency.
An automated scanner reports 16 problems across 6 colour combinations and stops there. Six combinations is not six decisions. Tracing each one back to the stylesheet gives this:
| Reported as failing | Ratio | Elements | Where it actually comes from |
|---|---|---|---|
#a67c2e on white |
3.79 | 5 | The --gold brand variable |
white on #a67c2e |
3.79 | 1 | Same variable, used as a background — not a second problem |
#b07415 on white |
3.92 | 4 | The --warn brand variable |
#c3bcbc on white |
1.86 | 3 | Not a colour in the stylesheet at all. It is --muted seen through
opacity:.42 on the booked-date tiles. |
#a49e9f on white |
2.63 | 2 | The same opacity:.42, this time over --ink |
#9a8d8c on #fdf8f4 |
3.03 | 1 | --muted through opacity:.7 on one footer paragraph |
#c3bcbc returns nothing, and a developer can lose an afternoon to that. The
real cause is two opacity rules. Sixteen findings, six reported combinations,
four changes.
| # | Change | From | To | New ratio | Fixes |
|---|---|---|---|---|---|
| 1 | --gold |
#A67C2E |
#957029 |
4.54 | 6 |
| 2 | --warn |
#B07415 |
#A26B13 |
4.52 | 4 |
| 3 | .d.gone opacity | .42 | .88 |
4.59 | 5 |
| 4 | footer paragraph opacity | .7 | .9 |
4.60 | 1 |
| Metric | Before | After |
|---|---|---|
| Problems found | 16 | 0 |
| Serious | 16 | 0 |
| Checks passed | 12 | 12 |
Same tool, same page, same settings, after the four changes above. This re-test is included in every engagement, so the work is evidenced rather than asserted.
Automated testing reliably detects roughly 57% of accessibility issues by volume and covers 20–30% of WCAG success criteria. The remaining criteria — keyboard operation, focus order, screen reader announcement, whether alt text is actually meaningful, and error handling — require manual testing with assistive technology. Every finding above was additionally checked by hand to confirm the element is visible and the failure is real; findings caused by scroll animations, hidden elements or backgrounds the tool cannot resolve were excluded rather than reported.