Accessibility statement

Last updated: August 2026

Conformance target

FluTrack is built against the Web Content Accessibility Guidelines (WCAG) 2.2, Level AA. That is a target, not a certification, and this page does not claim full conformance. What follows separates what has been measured, what the build enforces on every change, and what no person has yet checked with assistive technology.

What has been measured

The figures below are computed from the site's own design tokens and can be re-derived from them.

Text contrast

Text-and-background pairings were computed in both themes: body, secondary and muted text, links, and the five activity labels on their tinted cards. Every pairing checked clears the 4.5:1 that Level AA asks for at body size, with the tightest measuring just under 5:1 — the “High” activity label on its tinted card, at 5.00:1. Body text sits far above the threshold in both themes.

The state map

The map's five fill colours are ordered by lightness, not hue alone. Measured relative luminance runs 0.550, 0.353, 0.215, 0.115 and 0.042 from the lowest step of the scale to the highest — strictly decreasing, so the order of the scale survives a greyscale print. The ramp was also run through a standard dichromat simulation: under both deuteranopia and protanopia the luminance order remains strictly decreasing. An earlier ramp inverted under simulated deuteranopia, so a tile at the high end of the scale rendered lighter than one at the low end. That is why this ramp was rebuilt, and it is recorded in the changelog.

Colour is not the only channel. Each tile prints its two-letter state code, and its numeric level 0 through 4 wherever a level was published, and carries a text label naming the state and its level. The state code measures at least 4.77:1 against every fill, because the tile ink switches from dark to white partway up the scale.

What the build checks on every change

These run as part of the build and stop it when they fail:

• Every button in the generated HTML must have an accessible name — visible text, an aria-label, an aria-labelledby reference or a title. An icon-only control with none of these fails the build.
• Every indexable page must carry a title, a meta description and a canonical; a page excluded from indexing must not declare one. A page missing any of these fails the build. The heading count is reported as a warning rather than a failure.
• Every internal link and asset reference must resolve to a file that was actually generated.
• The theme control's dark palette and the operating system's dark palette must declare the same set of colour tokens — neither may define one the other leaves out — so a token cannot exist in one dark context and be missing from the other.
• Every generated icon and the share image must be decodable and reach the edge of its canvas, so a truncated or blank asset cannot ship.
• No published contact address may use a reserved domain that never resolves.

A unit-test suite runs alongside them. Several tests are accessibility-specific: that the map emits one keyboard-reachable link for each of the 51 tiles, that each tile carries its numeric level as a non-colour cue, and that the bypass link is present.

What is built into the interface

Each item was confirmed in the current stylesheet and scripts:

• A skip-to-content link is the first focusable element on every page.
• The map is 51 consecutive tab stops. A “Skip the state map” link sits before it and moves focus past the whole grid in one keystroke.
• Every tile is an ordinary link, so the map works with scripting off. The state directory lists the same pages as plain links.
• Keyboard focus is a three-pixel outline held clear of the element it marks. On map tiles it sits outside the tile, so its contrast is against the page rather than the fill underneath.
• Animation is opt-in: tile and gauge animations are declared inside a reduced-motion “no-preference” query, so a system set to reduce motion never applies them.
• In forced-colours mode the severity fills, meter segments and legend swatches opt out of colour substitution, so the five steps do not flatten into one.
• A “prefers more contrast” setting strengthens card and control borders, thickens tile outlines and underlines every link.
• The theme control cycles through light, dark and follow-the-system, rather than pinning a visitor to one override after a single press.
• Links inside running text are underlined, not distinguished by colour alone.
• Tables in running text sit in a focusable, labelled scrolling region so they can be reached and scrolled from the keyboard. The vendor register is the one wide table without a minimum width, so it compresses rather than scrolling on a narrow screen.
• Form errors mark the field invalid, point it at its message, and announce the result in a polite live region.
• Every state page restates its level, composite score and week-over-week direction as a sentence, not only as a graphic.

What has not been verified, and is therefore not claimed

The project's release checklist requires named manual sign-off for the items below. Every one is currently unsigned, so this statement makes no claim about them.

• Screen readers. No screen-reader pass has been signed off. NVDA with Chrome and VoiceOver with Safari are both on the checklist and both outstanding. No claim of screen-reader compatibility is made here.
• Keyboard-only operation. Not signed off. The skip links, the map bypass and the focus handling described above are present in the markup and stylesheet, but no person has driven the site start to finish using only a keyboard and recorded the result.
• 200% zoom. Not signed off.
• Forced colours and reduced motion on real hardware. Both are implemented in the stylesheet, as described above; neither has been confirmed on a device.
• Viewport widths. Six widths — 320, 375, 390, 414, 768 and 1024 pixels — are unsigned. The browser used for an earlier measurement pass would not honour a 390-pixel viewport override, so no automated run has certified reflow at handset width.

Separately, and not part of that checklist: no external accessibility audit has been commissioned. Everything on this page was measured or checked in-house.

Known limitations

• Form control borders. Both the text input's resting border and the dropdown's fall below the 3:1 that WCAG 2.2 asks of a control boundary. For the dropdown, the chevron clears that threshold and marks the control at rest. For the text input there is no such second cue: its boundary is below threshold until it receives focus, at which point the focus outline clears it.
• The tile level digit. The small 0-to-4 digit on each map tile is drawn at reduced opacity. Against two of the five fills it clears the 3:1 asked of a graphical object but not the 4.5:1 asked of small text. It is a redundant cue — the state's level is also given in the tile's text label and in prose on the state page — but it is not itself text-contrast compliant.
• The map is a tile grid. It is a deliberate simplification, so relative position is approximate. Below roughly 26rem of width it scrolls sideways inside its own container rather than shrinking tiles under the 24-pixel minimum target size.

Report a barrier

Date basis

The contrast figures and the colour-vision simulation were computed from the design tokens in the build dated below, and the automated checks listed are the ones in that build. The sign-off status was read from the release checklist the same day. This statement is revised when the code or that status changes, not on a fixed cycle.