Accessibility

Why this page is written this way

Public-sector buyers read accessibility pages properly, and they have been trained by a decade of vendors to treat "fully WCAG compliant" as a warning sign. Conformance is a per-page, per-release, tested statement, not a badge a product carries permanently. A survey tool cannot honestly claim conformance for surveys it has not seen, because a great deal of the final experience is decided by the person building the form.

So this page splits into three: what we do, what you control, and what we have not done. If you are completing a procurement questionnaire and need a formal position, use the accessibility statement, and ask us directly rather than inferring from marketing copy.

The standard we target

WCAG 2.2 Level AA, published by the W3C in October 2023, is the benchmark. In the UK it is the standard referenced by the Public Sector Bodies (Websites and Mobile Applications) (No. 2) Accessibility Regulations 2018, so a council, NHS trust or university running a consultation carries a legal duty that a survey embedded in its site does not escape. WCAG 2.2 added criteria that matter for forms in particular — visible focus that is not obscured, target sizes, dragging alternatives, consistent help, and not asking someone to re-enter information they have already given.

Decisions we have taken

Keyboard operation

Everything a respondent needs to do is a real form control or a real button: inputs, radios, checkboxes, selects, links. Native controls are used because they arrive with keyboard behaviour, state reporting and platform conventions already correct, and every re-implementation of a checkbox is an opportunity to get one of those wrong.

Ranking questions are presented as items you drag into order, and dragging is a pointer gesture — WCAG 2.2 success criterion 2.5.7 expects a single-pointer alternative. There is one: every option carries an Add button that moves it into the ranked list, and every ranked item carries a control to take it back out. The question can be completed entirely with buttons, so dragging is a convenience rather than the only route through.

Focus indicators

Focus is never removed, and the focus ring is deliberately amber rather than the brand purple. A purple ring on a purple button is not a visible focus indicator; it is a decorative outline that happens to fail 2.4.11. Choosing a colour that clashes with the brand is the point.

Motion

All non-essential motion sits behind prefers-reduced-motion. If your operating system is set to reduce motion, entrance animations and hover transforms do not run — the content is simply there. Nothing important is communicated by movement alone, and nothing auto-plays.

Semantic markup and labelling

Pages use headings in order, lists that are lists, tables that are tables, and landmarks that let a screen-reader user skip to content. Every question is a labelled control, and label text comes from the question title you write — which is exactly why the wording you choose in the builder is an accessibility decision, not a style decision.

Not relying on colour or images alone

Option tiles carry their label as text, questions can take an optional description line, and images attached to a question are supplementary to the wording rather than a replacement for it. The builder offers a plain list as an alternative to the default bento option tiles; a plain list is often the better choice for long option labels and for respondents using magnification.

Errors and validation

Validation covers required fields, minimum and maximum length with a live counter, numeric ranges, email, phone, URL, UK postcode and custom patterns. A custom pattern can carry a custom message, and it should: "invalid format" tells nobody anything, whereas "enter a postcode like SW1A 1AA" tells everybody what to do. Write the message you would want if you could not see the field.

Reducing the load on the respondent

Accessibility is not only mechanics. Cognitive load is the barrier that goes unmeasured, and three product features exist partly to lower it: conditional logic so nobody reads questions that do not apply to them, save and continue so a long survey can be done in short sittings, and piping so a later question refers back to what the respondent actually said. There is also no CAPTCHA anywhere in the respondent flow — spam protection is handled instead by a honeypot, a dwell-time floor and a per-minute rate limit, which is one fewer challenge for a disabled respondent to fail.

What you control

The tool can only take you so far. These are yours:

• Layout choice. Classic is a single scrolling page in document order and is the most predictable for assistive technology. Conversational moves one question at a time.
• Question wording. It becomes the accessible name of the control. Long, compound or ambiguous questions are an accessibility problem before they are a data-quality problem.
• Images. If an image carries information the question text does not, describe it in the question or the description line.
• Accent colour. Per-survey theming can be set to a colour that fails contrast against white. Check it.
• Ranking and matrix questions. Both are harder work than a simple choice. Use them where they earn it.
• Testing. Complete your own survey with the keyboard alone before you publish it. It takes five minutes and finds most problems.

What we have not done

Stated plainly, because the omissions matter more than the list above:

• No independent accessibility audit has been carried out.
• No accessibility certification is held, and none is claimed.
• No VPAT or formal conformance report is published.
• Ranking questions offer a pointer-free route — each option carries an Add button to move it into the ranked list and a remove control to take it back out, so the question can be completed without dragging. Dragging remains available for those who prefer it.
• Multi-language surveys exist, but you write the translations, and accessibility is unverified in any of those languages — because it is unverified in all of them.
• There is no offline or kiosk mode for assisted completion without a connection.

If you find a barrier, tell us through the contact page with the survey layout and the assistive technology you were using. A specific report is worth more than a scan report, and we would rather fix a thing than argue about a criterion.

Related

See NumoForms for local government for how consultations tend to be run, per-survey partner branding for co-badged work with other authorities, and the security page for storage location and sub-processors. The product overview gives the wider picture.