Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Conversion Rate Optimisation › Landing Pages

Accessibility as Conversion Work

Why accessibility belongs in conversion work: the barriers that stop people completing a page, how to find them, and how to fix them without a test.
Guidebeginner13 min readUpdated

A barrier is a lost conversion

Conversion work is mostly about friction: the extra field, the unclear price, the slow page that some visitors push through and others abandon. A barrier is different. It is a step that some visitors cannot complete at all, whatever their motivation. A delivery-slot picker built so that only a mouse can select a slot ends the journey for anyone using a keyboard or a screen reader. A consent banner that traps keyboard focus is worse: the consent banner QA checklist puts it plainly, a visitor who cannot dismiss the banner cannot use the site at all.

The people affected are not a separate audience. They include visitors who use screen readers or keyboards, visitors who zoom the page to read it, visitors with a temporary injury, and visitors on a phone in bright daylight with one hand free. The forms guide makes the point for forms, and it holds for every page: most of what makes a page accessible also makes it convert. Clear labels, errors that say what went wrong, contrast that can be read outdoors and buttons large enough to tap help everyone, which is why accessibility belongs inside the conversion programme rather than in a separate compliance project.

Barriers are also hard to see in the data. A visitor who cannot get past a step simply leaves, and the funnel records an exit like any other. Analytics is precise about where people leave and silent about why, as the conversion research methods guide explains, and people who meet a barrier rarely write in to say so. A step whose conversion rate (CVR) has always been a little low can hide one for years. That is why barriers are found by checking the journey directly, not by waiting for the numbers to point at them. The academy checkpoint on shipping, redesigns and accessibility without a split test asks you to spot barriers that look like ordinary drop-off.

What the guidelines ask, in brief

The Web Content Accessibility Guidelines (WCAG), published by the World Wide Web Consortium and listed in this page's sources, are the widely used standard for web content. They are organised around four principles: content should be perceivable, operable, understandable and robust. They apply to a Landing Page, a checkout, an email or an interactive tool as much as to the main website, as the approvals and compliance guide notes. Conformance requirements vary by jurisdiction and sector, so treat the guidelines as the shared reference and check what applies locally; nothing here is legal advice.

The table translates the four principles into what they mean on a page that is trying to convert, and points to the guide on this site that covers each in more depth.

PrincipleWhat it means on a conversion pageTypical failureWhere it is covered
PerceivableInformation is available to people who cannot see the image, hear the video or distinguish the colourAn offer shown only in an image; an error shown only in redWriting for the web, alt text and captions
OperableEverything can be done with a keyboard, without hover, without racing a clock, with targets large enough to tapA slot picker that needs a mouse; a session that expires with no warningForms, speed and mobile
UnderstandableLabels, instructions and errors say what is needed and how to fix itA generic "invalid input" far from the fieldForms, checkout
RobustThe page works with assistive technology, not only with one browserA custom control a screen reader announces as nothingSupport channels and self-service

Several specific requirements recur in the site's guides. The forms guide lists them for forms: every field has a label that is programmatically associated with it, not just text nearby; instructions and format requirements are given; errors are identified in text, with suggestions for correction where known; focus is visible as a person moves through the page with a keyboard; colour is never the only way information is conveyed; touch targets meet the documented minimum size; and nothing depends on a time limit the person cannot extend or on content that flashes or moves unexpectedly. The writing for the web guide adds real headings in order, link text that makes sense on its own, and alternative text that says what an image conveys. The brand guidelines template settles colour contrast once, in the palette, so every page inherits pairs that pass.

Where barriers sit along the journey

Barriers cluster where pages are most interactive, which is also where the conversion happens. Walking the journey step by step, the usual suspects are these.

StepCommon barrierQuick check
First screenConsent banner or pop-up that cannot be dismissed with a keyboard; headline or offer baked into an imageTab from the address bar; can you reach and close everything, and is the offer readable as text?
Navigation and searchMenus that open only on hover; links that read "click here"Move through the menu with the keyboard; read the link list out of context
Product or service informationDetails only in images or video without text alternatives or captions; stock or sale status shown only by colourTurn images off or use a screen reader; check every colour signal has words beside it
FormsPlaceholder text instead of labels; errors in colour only; focus that jumps or disappearsComplete the form with the keyboard alone and make a deliberate mistake
Checkout and paymentCustom selectors for delivery slots or options; session time-outs with no warning; an embedded payment step nobody checkedComplete a purchase with the keyboard and a screen reader, slowly
ConfirmationSuccess shown only visually, so a screen reader announces nothingListen to what is announced after the final click

The checkout deserves the slowest walk. The checkout optimisation guide already asks that time-outs and session expiry be long enough for a slow shopper and that the basket be restored when the shopper returns, and the guidelines add that a time limit should not be one the person cannot extend. Payment steps and booking widgets are often supplied by a third party, which is a reason to check them, not a reason to skip them: the visitor experiences one journey, whoever built each part of it. The same applies to chat widgets, which the support channels guide expects to work with a keyboard and a screen reader, and to the variants of any running test, which change the page as much as a release does. Pages are only half of it for someone who needs help: the guide to supporting customers in vulnerable circumstances covers how a support team adapts the conversation itself, from the channel to the time allowed.

Finding barriers: the checks that catch most of them

The forms guide states the principle that applies everywhere: accessibility problems are found by testing, not by inspection. A small set of checks, run on the journey rather than on single pages, finds most of them.

  • The keyboard alone. Unplug the mouse and complete the journey from landing to confirmation using tab, shift-tab, enter, space and the arrow keys. Every control should be reachable, the focus should always be visible, and the order should match the order a person would expect.
  • A screen reader. Use the screen reader built into the operating system on a desktop and on a phone. Listen to how headings, buttons, form fields, errors and the confirmation are announced. Silence after a click is as telling as a wrong label.
  • Zoom and reflow. Enlarge the text and the page; content should reflow rather than spill off the screen or hide the call to action (CTA).
  • A phone held in one hand. The speed and mobile guide sets out the layout rules for a thumb: reachable actions, nothing that depends on hover, text readable without zooming, and tap targets at least as large as the documented minimum.
  • An automated scan as a first pass. Scanners catch what a machine can see, such as missing labels or low contrast, and miss what needs judgement, such as whether alternative text says anything useful.

Fold these checks into the research the programme already runs. The landing page audit scores accessibility against the WCAG success criteria as a guide to what to check, not as a legal opinion, and the landing page checklist carries the four checks every new page should pass. Usability sessions, described in the research methods guide, are stronger still when some participants use assistive technology themselves, because they find problems no checklist anticipates. Record every barrier in the research backlog with its location, the check that found it and the steps it blocks, exactly as any other finding. Kept in the research backlog, barriers sit beside every other finding rather than in a separate accessibility list.

Fixing without a test

Barriers are bugs. The CRO pillar and the conversion research hub both make the rule: broken things are fixed, not tested, and a step that some visitors cannot complete is broken for them. Running an A/B test on whether to remove a barrier keeps it in place for half the traffic for the length of the test, and the result would not answer the question anyway. The visitors affected are usually a small part of any sample, so a test sized for the whole audience would rarely detect a change in their behaviour, and a reading of "no measurable difference" would be taken, wrongly, to mean "no problem". The sample size guide explains why small effects in small groups are invisible at ordinary sample sizes.

What can be tested is the choice between accessible alternatives: two layouts for an error summary, two ways of presenting delivery slots, two wordings for a field's instructions. Both arms must pass the same checks, and the test answers which accessible version converts better, not whether accessibility is worth having.

Order the fixes by where they sit. Barriers on the conversion path itself, in the consent banner, the forms, the checkout and the payment step, come first, because each one ends journeys outright. Barriers in content, such as missing alternative text on supporting images or vague link text, come next. After each fix, repeat the check that found the barrier, then watch the step's rate in the funnel analysis without expecting it to prove anything on its own; the fix was justified by the barrier, not by the number. The conversion rate calculator turns the step's counts into rates, so the periods before and after the fix are compared the same way. The academy exercise on evaluating a website relaunch for a cookware maker practises shipping barrier fixes on the current site rather than holding them for a relaunch.

Building accessibility into how pages are made

Fixing barriers one page at a time is endless if the templates keep producing them. The approvals guide recommends the practical approach: build accessibility into templates, covering colour contrast, heading structure, alternative text fields and link text, train the people who build pages to check new work against a short list, and keep specialist review for new templates and complex interactive work.

The rest of the production chain can carry the same checks at little cost.

  • Briefs. The landing page brief template already requires that the page be usable with a keyboard and a screen reader; keep that line in every brief so it is designed in, not bolted on.
  • Brand decisions. Contrast pairs and type sizes chosen once in the brand guidelines reach every page, advert and email afterwards; chosen later, they have to be fixed everywhere. The guide to designing an accessible brand identity sets out those decisions for the palette, type, logo, imagery, motion, sound and voice.
  • Publishing. The web page publishing SOP includes a live check that the page is readable with keyboard navigation, in line with WCAG.
  • Tests and third-party scripts. The A/B test launch checklist is run before any variant goes live; add the keyboard and screen reader checks to it, because a variant built in a testing tool can introduce a barrier the original page did not have.
  • Speed alongside access. Page speed, read through Core Web Vitals, is a separate quality bar with its own guide, but the two meet on phones, where a heavy page and an awkward control are often met by the same visitor.

Writers and designers carry a large share of the work. The writing for the web guide notes that several of the guidelines' success criteria are met or failed by the writer alone, through headings, link text, alternative text and plain language, and the guide to alternative text and captions covers images and video. Many visitors arrive from a social post, and the guide to accessible social media posts applies the same guidelines to the descriptions, captions, graphics and links in those posts. For sectors where an inaccessible form decides who can apply at all, such as admissions, the education industry guide shows what is at stake. To run these checks across every channel with named owners, monthly spot checks and quarterly journey tests, the marketing accessibility programme playbook joins this guidance with the brand, email, social and support guides.

An online grocer's checkout without a mouse (illustrative)

Illustrative example — not a real result

Pantrywhim Grocer checks its checkout without a mouse

Pantrywhim Grocer, an invented online grocer taking about 1,900 orders a week, adds accessibility checks to its quarterly research sprint. A marketer and a developer complete the journey from the homepage to the order confirmation using only a keyboard, then again with a screen reader on a phone. They log four barriers. The delivery-slot picker is built from clickable boxes that the keyboard skips entirely, so a keyboard user cannot choose a slot. The card form shows errors only as a red outline, which the screen reader does not announce. The checkout session expires after 15 minutes with no warning and empties the basket. The order confirmation appears visually but nothing is announced. The consent banner, by contrast, passes. The team logs each barrier in the backlog as a bug, rebuilds the slot picker with standard buttons, adds text errors linked to their fields, adds a time-out warning with an option to extend and keeps the basket, and announces the confirmation. The fixes ship over two sprints without a test. Over the following two months, chat contacts that mention the slot picker fall from 38 to 5, and the share of visitors who continue past the slot step rises from 74% to 76%, a movement the team reads cautiously because a delivery-price change shipped in the same period. The picker, the error pattern and the time-out warning go into the shared component library so new pages inherit them.

Common mistakes

  • Treating accessibility as a separate project. Barriers are conversion problems; record and fix them in the same backlog as everything else.
  • A/B testing whether to remove a barrier. Fix it; test only between accessible alternatives.
  • Relying on an automated scan. It finds what a machine can see; the keyboard and screen reader checks find the rest.
  • Fixing the page and not the template. The next page built from the same template brings the barrier back.
  • Skipping the parts someone else built. Consent banners, chat widgets, payment steps, booking widgets and test variants are part of the journey the visitor experiences.
  • Colour as the only signal. Errors in red, sale prices in red and stock levels as coloured dots need words beside them.
  • Offers and headlines as images. Text in an image is invisible to a screen reader and blurs when zoomed.
  • Time limits without warning. A session that expires silently costs the slowest visitors their basket.
  • Reading silence as success. People who cannot complete a page leave without writing in, so an absence of complaints proves nothing.
  • Claiming conformance nobody has checked. Say what has been tested and what has not, and leave statements about legal obligations to whoever handles compliance.

Frequently asked questions

Is website accessibility a legal requirement?

In many places and sectors it is, but the requirements differ by jurisdiction and by type of organisation, and they change. The Web Content Accessibility Guidelines published by the W3C are the widely used reference that most requirements point to, so building and checking against them is the practical starting point. Whether a particular site must meet a particular level is a question for whoever handles compliance in the business; nothing on this page is legal advice.

Should accessibility fixes be A/B tested before they ship?

No. A barrier that stops some visitors completing a page is a bug, and conversion programmes fix bugs rather than test them. A test would also keep the barrier in place for half the visitors while it ran, and because the people affected are usually a small part of any sample, a result showing no measurable difference would say nothing about them. Test the alternatives for how to fix something, such as two accessible layouts for an error summary, not whether to fix it.

Does an automated accessibility checker find everything?

No. A scan is a useful first pass for things a machine can see, such as missing labels, images with no alternative text or text with too little contrast. It cannot judge whether the alternative text says what the image conveys, whether an error message makes sense, or whether the order in which a keyboard moves through the page matches the order a person would expect. Completing the journey with the keyboard alone and with a screen reader is what finds those.

Terms used on this page

Tools to try

Where this page is used

Part of these learning paths

Also related

Sources

  1. Web Content Accessibility Guidelines (WCAG) — W3C

Related