Accessibility Statement

1. Our commitment

1.1 This website exists to make skincare information available to people who would otherwise be guessing. That is not much use if the information is unreachable.

1.2 We aim to meet Web Content Accessibility Guidelines (WCAG) 2.1 Level AA.

1.3 We do not currently meet it. Clause 4 says where, and clause 5 says what we have done about it. A statement that claimed full conformance would be easier to write and would be untrue.

2. What this statement covers

2.1 The website at sensible.com.sg and its subdomains, as at 4 September 2026.

2.1A What has and has not been assessed of the account work, as at 17 September 2026. Accounts were built on 16 September. On 17 September we checked the signed-out half of the results page against the live site: every visible control was checked for an accessible name, every input for a label, the page for duplicate identifiers and for skipped heading levels, and every control against the 24 by 24 pixel target-size minimum. We found no unnamed controls, no unlabelled inputs, no duplicate identifiers and no skipped headings. We found one thing we have not yet changed: the tick box for marketing email is 18 pixels square. The 24-pixel minimum we measured it against comes from WCAG 2.2, and this statement declares conformance to WCAG 2.1 (clause 3), so it is not a failure of the standard we claim — we are telling you because it is a real thing about a real control, and holding it back on a technicality would be the wrong instinct in a document like this one. The whole line of text beside the box can be clicked, which helps.

2.1B What is still unassessed, and it is about half of it. The parts that only appear once you are signed in — the three choices offered when you save a routine, the panel that asks whether you want to hear from us, the strip that appears on a phone when you are signed in, and the account menu in the header — have not been checked, because checking them needs a signed-in session. Nor have we yet checked colour contrast, keyboard focus order, behaviour at 200% zoom, or how any of it sounds on a real screen reader. The sign-in page itself is hosted by Shopify and is outside our control in the same way the checkout is — see clause 2.2. We would rather list this than let the 4 September scope imply a coverage it does not have.

2.2 It does not cover third-party services reached from the website. In particular, from the Shop Opening Date the checkout is operated by Shopify and its accessibility is governed by Shopify's own conformance, which we do not control and have not independently assessed.

3. What works

Assessed by automated testing across every page at seven viewport widths, and by keyboard navigation, on 2 September 2026.

• Keyboard. Every step of the Routine Generator can be completed using only a keyboard. Every interactive control can be reached by tabbing, and activated with Enter or Space.
• Focus. No stylesheet on the website suppresses the browser's focus indicator, and 84 explicit focus styles are defined for custom controls.
• Reduced motion. The website honours the prefers-reduced-motion setting on every page. With it enabled, animations resolve immediately and no content becomes unreachable — verified page by page.
• Reflow. No page scrolls horizontally at any width from 320 px upwards. Content reflows to a single column on small screens.
• Touch targets. Interactive controls meet a 44 × 44 px minimum, with hit areas expanded behind the visual control where a smaller appearance is wanted.
• Text alternatives. Every image carries an alt attribute. Decorative illustrations are marked as decorative so screen readers skip them rather than announcing a filename.
• Structure. Pages use semantic headings, landmarks and lists, so a screen reader can navigate by structure.
• No hover-only content. Nothing on the website is reachable only by hovering.
• Zoom. Text can be resized to 200% without loss of content or function.

4. What does not work yet

We are naming these because you should not have to discover them.

4.1 Colour contrast — the main one

Measured on 2 September 2026 by rendering every page in a headless browser at 1280 px and 390 px, walking every visible text node, and taking the actually painted background pixel rather than the nearest ancestor in the markup:

83 failing text nodes, across 35 distinct colour-on-ground combinations, on 9 pages. Most are small uppercase labels, section eyebrows and captions at 10–13 px, measured against a required 4.5:1 (or 3.0:1 for large text).

Most of the affected text is supplementary — section labels, figure captions, the kicker above a heading. Two exceptions we are not going to bury:

(a) The "see a doctor" alert panel on the rosacea page measures 4.47:1 against a required 4.5:1. It is a marginal failure of the most safety-critical copy on the site. When it is fixed it will be taken to real headroom, not to 4.51:1.

(b) The large decorative words on the acne page (ZAP!, BAM!, POP!, OW!) measure 1.04:1 to 1.30:1 against a required 3.0:1. They carry no information that is not also in the body text. Whether they are decoration or content is a design decision, and until it is taken they are recorded here as a failure rather than argued away.

No body text and no clinical instruction other than (a) is affected. That is a mitigation, not an excuse.

Status: measured and registered in full, with every failing pair listed. No colour has been changed. Not yet fixed.

4.2 Small controls

Four controls fall marginally below the 44 px minimum: two buttons at 42 px, and two information buttons whose expanded hit area is partly overlapped by an adjacent card.

Status: identified. Low impact, scheduled.

4.3 Interactive figures

Several pages carry interactive instruments — a trigger dial on the rosacea page, a sun-exposure simulator on the pigmentation page. These are operable by keyboard, but their output is conveyed largely visually and their text alternatives do not fully describe what the illustration shows.

Status: identified. These need proper descriptions rather than a technical fix.

4.4 Screen reader testing

Our assessment is automated and manual-keyboard. It has not included end-to-end testing with JAWS, NVDA or VoiceOver by a person who uses those tools daily.

We are not claiming conformance we have not tested for. Automated testing catches perhaps a third of accessibility problems. Until a real screen-reader user has been through the Routine Generator, our Level AA claim is an aim, not a finding.

Status: not yet done. This is the largest gap in this statement.

4.5 Language

The website is published in English only. A Chinese translation exists in the project's archive but is not published, and the language selector in the header is not currently functional.

Status: a product decision, not a defect to be fixed.

5. What we have done

Changes made on 2 September 2026 following an accessibility review:

1. The Routine Generator could not be completed by keyboard at all. The selection cards on six of its steps were unlabelled, non-focusable elements. A keyboard-only visitor could not answer the skin-type question, the makeup-and-sunscreen question, the moisturiser question or the sun-exposure question, and so could not reach a result. Fixed. Every card is now a labelled button, reachable by tab and operable by Enter or Space, with its selected state exposed to assistive technology.
2. prefers-reduced-motion was ignored on eight pages, including the home page. Fixed site-wide, and verified not to hide content.
3. Navigation items appeared twice on mobile, once in a collapsed bar at 16 px tall and again inside the menu. Fixed.
4. Three header controls were below the touch minimum. Fixed.
5. The home page scrolled horizontally at 320 px. Fixed, at source rather than by clipping the overflow.
6. A product that must not be sold to a pregnant customer was prevented from being added to the basket by a visual rule only, which a keyboard user could bypass. Fixed — the control is now genuinely disabled.

6. How this was assessed

7. Tell us if something does not work

7.1 If you cannot use part of this website, we want to know. It is more useful to us than almost any other feedback.

7.2 Email hello@sensible.com.sg. Tell us the page, what you were trying to do, and what happened. If you can, say what browser and what assistive technology you were using — but do not let not knowing stop you writing.

7.3 We will acknowledge within 5 working days and tell you what we intend to do and roughly when. If we are not going to fix something, we will say so and say why, rather than leaving you waiting.

7.4 If you need information from this website in another format — larger text, a plain document, or read to you — ask, and we will provide it at no charge.

8. If we do not put it right

If you are not satisfied with our response, tell us and it will be escalated. We are not aware of a statutory accessibility complaints body for private websites in Singapore, so we would rather resolve it directly than have you find there is nowhere to go.

9. What this statement does not cover

Everything in clause 3 was measured on 2 September 2026. Two parts of the site have been built since and have not been assessed at all: the shop pages and the Contribute page. Do not read clause 3 as covering them.

The figures in clause 4 are from that same assessment. They will be re-measured, and this statement updated, before we describe the site as conforming.

We would rather fix things than list them. Where something in clause 4 can be fixed, the intention is to fix it and delete the entry, not to leave a tidy disclosure standing in for the work.

Version 2.1 · In force from 17 September 2026. Standard: WCAG 2.1 Level AA — partial conformance, as set out in clause 4.

Assessed 2 September 2026. We review this statement on any material change to the site, and in any event every 12 months. Clause 4 says where the site does not yet meet the standard and clause 9 says what has not been tested at all. Both are there because you should not have to find out by running into it.