The directions on a payroll announcement seem almost too simple to test. A text box contains a short message, and a button opens formats the employee can use. But a button can look real and still be absent from keyboard navigation. A stack of blue links can make sense beside their headings yet sound identical when a screen reader lists them alone. Run two distinct passes, starting from the message the employee actually receives.
This walk-through is for employer-controlled information pages and test copies. It is not a way to test a stranger’s account, and it does not require anyone’s personal or payment details. The question for this review is whether an employee can locate and understand the employer’s own instructions and distinguish them from third-party steps. Do not infer behavior in the provider’s systems from what happens on your page.
Pass one: put the pointer aside
Start from a fresh copy of the rollout email. Press Tab until you reach the link to the notice, open it, and continue from the top of the destination. Use Shift+Tab to check the reverse path. When you encounter a control, use the keyboard operation its role implies. A conventional link follows on Enter; a conventional button can be activated by Enter or Space. Record custom widgets for the development team rather than guessing how Workday, a mobile browser or another environment should behave. For scripted controls, document expected keyboard behavior and test it.
Look at the screen while moving. Can you tell which control has focus? Is the focus indicator visible against the background? A cookie or privacy panel, newsletter banner, or sticky toolbar can cover a focused link even when pressing Enter would still take you somewhere. Trigger the panel that opens and closes alternate formats. After closing it, check where focus landed. If the panel starts nested navigation, make sure you can exit back to the path. The WCAG 2.2 keyboard guidance is useful for relating an observed defect to a criterion.
Pass two: listen to the document
Choose a screen-reader and browser pairing your organization can support. Record both and the operating system. Begin with the page title, then use heading navigation. Does the reading order give you the timeline, the document choices, and the applicable contact without making you listen to every sidebar? Then use your screen reader’s list of links. Three “Learn more” links are harder to distinguish in a list than “Compare notice formats,” “Read the Spanish notice,” and “Ask HR for a different format.” The link label should be accurate even when read apart from nearby styling.
Check the dynamic messages. In a test fixture, expand the resources panel, then simulate an unavailable version. Can the employee discover the state without moving through the whole page again? Is a visible error also exposed to assistive technology? Do not depend on one noisy announcement to explain several unrelated problems. If an error interrupts the task, make the next action clear in text. The W3C’s first-review checklist includes helpful checks for headings, focus and other foundations. The team should plan broader evaluation for a release.
Tell a developer exactly how to see what you saw
A report called “broken for screen readers” will lead to a meeting about tools. Instead write “From the test-email link, open Notice v4 in Firefox with NVDA; press H to jump to the third heading. The alternate-format title is skipped, and its links appear under the previous section.” Add what should have happened and whether the normal employee task can still be completed. A sanitized screen capture or audio-free recording can help, but do not capture real account pages, personal information or user speech without separate approval.
Once the fix arrives, start at the same point and use the same test combination. Then check the page with the other pass as well. A fix that restores arrow-key movement but takes away visible focus only trades one problem for another. For a supported mobile route, try a phone with its built-in reader and practical text settings, rather than declaring a desktop result transferable.
For a repeated result, check that magnification or voice output did not mask a slower state change. Note what happens immediately and again when the network settles.
Know when a quick pass is not enough
These passes establish only what you observed. Include people who actually use assistive technology when planning a fuller evaluation, and avoid making one participant a stand-in for all users. Give a recruited tester a realistic task and a private way to flag a barrier, not a list of keys they must press. Keep research notes separate from accommodation case records.
If you are testing an attachment instead of an HTML page, start with the PDF reading-order check. If controls disappear only on a small screen, use the notice-at-large-text test. For release owners, the rollout review connects those results to a sign-off decision.