DTC technology × growth systems — built to compound.
DTCUP / FIELD NOTES / OPERATE

Make Your Website Navigation and Forms Work for More Customers

Review complete navigation and form tasks, with a repeatable keyboard menu test, visible focus, clear labels and dependable error and success feedback.

Published Aug 30, 2026Updated Sep 20, 202611 min read
Jack Hu
Jack Hu

7 years of experience building and operating brand websites, specializing in website development, paid advertising and SEO. Built a brand website from launch to US$10 million in annual sales.

The Short Answer

Choose important customer journeys and test the controls and states needed to finish them. Follow navigation and forms with a keyboard, inspect labels and focus, and test errors, successful completion, enlargement and relevant assistive technology.

Record reproducible barriers and prioritize essential tasks and shared components. Verify the underlying result as well as the visible feedback, then retest after repairs. This focused review is a starting point, not a full WCAG assessment or legal compliance certificate. For navigation, record the starting state, forward and backward keyboard sequence, opening, dismissal, destination and responsive variant. Match expectations to the implemented disclosure or dialog pattern and retest neighboring controls.

Key takeaways

  • Review complete tasks and difficult states rather than isolated screenshots.
  • Make controls, labels, focus and corrections understandable throughout the task.
  • Success feedback must reflect the actual result, including uncertain outcomes.
  • Repair shared components and retain repeatable customer checks for future releases.

A customer should be able to find a product, choose an option and complete a form without guessing where the controls are or whether the action worked. Yet a website can look polished while its menu only opens with a mouse, its form labels disappear during typing or its error message leaves the customer stuck.

I would review these features as complete tasks. Start at the point where someone decides what to do, follow every control needed to finish, and include recovery from an ordinary mistake. This produces a more useful repair list than a collection of isolated screenshots or an unexplained accessibility score.

The process below is a practical starting review for websites and commerce apps. It is not a complete WCAG assessment or a legal compliance certificate. Involve an accessibility specialist and people who use relevant assistive technology when planning a thorough evaluation.

Choose journeys that include the difficult states

Select a small set of important journeys: find a category through navigation, search for a product, choose its options, create or access an account, submit an enquiry and complete the relevant purchase steps. Include at least one shared component used across many pages, because a single defect there can affect most of the site.

Write the expected result for each task. “The customer reaches the selected category and can identify the current page” is more specific than “menu works.” For a form, identify what record or confirmation establishes completion. The review should distinguish visible feedback from the underlying operation.

Add states that a first screenshot misses: menu open and closed, search with no result, field error, loading, successful submission and service failure. Use safe test data and an appropriate test environment for transactions. Do not create live payments merely to discover whether the button is reachable.

W3C's Easy Checks provides a starting point for manual review. Use initial checks to find issues and prepare a deeper assessment, not to declare the whole website accessible after a short pass. Keep the reviewed pages and states visible in the report.

Follow the task using a keyboard

Begin without using the mouse. Move forward and backward through the interactive controls and try to complete the chosen journey. Observe whether the sequence makes sense and whether you can see which control is active. A control that exists visually but cannot be reached is a practical blocker.

Test opening and closing navigation, following a submenu link and returning to the page. Do not accept a menu that opens only on hover or disappears before the user can reach its content. The W3C menu tutorial explains why structure, keyboard operation and understandable states belong together.

Check the intended interaction pattern. Ordinary site navigation often works well with links and disclosure buttons; it does not automatically need the complex behavior of a desktop application menu. W3C's disclosure navigation example is a reference for that distinction, not a component to paste into production without adapting and testing it.

Record the exact point where progress fails. “Keyboard user cannot reach the second category after opening Products” gives a developer a reproducible task. “Improve accessibility” does not establish what behavior must change or how to verify the fix.

For an ordinary disclosure-based shop menu, run a deliberate sequence. Use Tab and Shift+Tab to move forward and backward. Activate the disclosure button with Enter and Space, check the links it exposes, and test Escape to close it and return focus to its controlling button. Arrow-key behavior depends on the chosen pattern; it is optional in the cited disclosure example. Do not fail a standard navigation solely because it does not behave like an application menubar.

Take a specific shopping task, such as reaching the Bags category through Products, and record each transition. This original test record helps turn a quick inspection into a repeatable acceptance check:

Test moment Evidence to record
Before opening Products The control can be reached, its name is meaningful and focus is visible
While the category links are open The intended category is reachable without hovering or moving the mouse
After dismissing the panel The customer can identify their position and continue the journey
After following Bags The correct destination loads and the customer can proceed into its content
After returning or changing screen width The navigation still has an understandable state and working controls

Keep the URL, viewport, browser, starting state and key sequence with any failure. A short recording can help show a panel closing unexpectedly, but written steps are still useful for reproducing it. Record whether a consent layer, account menu or other overlay was present rather than leaving the next tester to guess.

Repeat the task in the site's small-screen layout using a suitable keyboard setup. A responsive header may use a different component, so the desktop result does not establish that the drawer works. If the drawer behaves as a modal dialog, evaluate its focus and dismissal against that pattern; do not mechanically copy the desktop disclosure expectations.

Finally, retest after the repair in both directions and from a fresh page load. Check a neighboring navigation item as well as the exact broken path. Fixing Products can accidentally change how search or account controls behave when they share a component. Keep the result scoped to the paths and environments actually checked, with a separate assistive-technology review where needed.

Keep focus visible when overlays and sticky bars appear

Menus, cookie notices, chat buttons and fixed purchase bars can cover the control the customer is trying to use. Repeat the keyboard journey with these features present. A page tested only before its overlays load can miss the state actual visitors encounter.

W3C's explanation of Focus Not Obscured, Minimum addresses author-created content completely hiding a focused component. For a practical design review, aim for a clearly visible, usable control rather than treating a tiny visible edge as a satisfying customer experience.

Check what happens when a dialog opens and closes. The focus behavior should match the selected dialog pattern, and closing should leave the person at a sensible place to continue. Hidden panels should not leave users navigating through controls they cannot see or use. Have the implementation reviewed against the appropriate pattern rather than adding arbitrary focus jumps.

Retest after scrolling and after the screen size changes. A sticky header may be harmless near the top and cover a field later. Capture the active control, overlay and viewport in the same evidence so the defect can be reproduced without guessing the page position.

Make each form control understandable before typing

For every field, check whether its purpose is clear before entry and remains clear afterward. A placeholder that vanishes is a poor substitute for a persistent visible label. Explain a meaningful requirement before submission instead of revealing a surprising format rule only after the customer makes an error.

The W3C labeling tutorial explains associating controls with their labels so assistive technology can identify them. A reviewer should check the actual accessible name as well as the visible design. Text placed near a field is not proof that the relationship is implemented.

For option groups, make the shared question and each choice understandable. For icon-only controls, verify that the action has a meaningful name. “Show password” and “Hide password” should describe the current action, while a close control should be distinguishable from a decorative symbol.

Ask whether every required field is needed for the task. Reducing unnecessary questions can reduce effort, but keep legitimate operational requirements. For example, remove a redundant marketing question rather than dropping information the delivery process genuinely needs and asking support to recover it later.

Test errors as part of completing the form

Submit an incomplete form and an ordinary invalid value in the permitted test setup. The response should identify the affected field, explain the problem in plain language and help the customer correct it. A red border alone does not explain what to do, and a generic error at the top may be disconnected from the field.

W3C's input validation guidance covers useful validation and the need for server-side checks as well as browser checks. Align the rules and wording across both. A password or address that appears acceptable in the interface should not be rejected later by a different undisclosed rule.

Preserve appropriate entered information after a correctable error so the person does not have to start over. Treat sensitive fields deliberately rather than repopulating everything without review. Check whether the error can be found with the keyboard and understood with the assistive technology in the test scope.

Use an invented registration example: the customer submits a form, the server rejects the email format and no account is created. A useful message says what needs correction and makes the uncompleted state clear. It should not suggest that an account exists merely because the request reached the server.

Make success and uncertainty equally clear

A button that changes color is not enough to establish success. Confirm the actual result and show what happens next. For an enquiry, this might be a clear receipt confirmation; for an account, it might be the new signed-in state. The message should match the real operation.

W3C's form notification guidance discusses accessible error and success feedback. Test how changes are announced in the chosen screen-reader and browser combination instead of assuming that any visible text will be noticed. Avoid excessive announcements that interrupt the task unnecessarily.

Distinguish a validation failure from an uncertain network result. If a request times out after the server may have accepted it, the system needs a way to verify the result or guide a safe retry. A message saying “failed” can cause duplicate submissions if it is not grounded in what happened.

Prevent accidental repeat clicks without removing the evidence the customer needs. Show a useful in-progress state and provide a recovery route if the process stalls. After testing, inspect the actual record as well as the screen so a friendly message cannot conceal a missing submission or a duplicate action.

Review enlargement, small screens and assistive output

Enlarge the page and check that navigation, labels, errors and buttons remain usable. Look for clipped text, overlapping controls and content that requires awkward movement in two directions to read. W3C's Reflow guidance explains the relevant presentation criterion and its exceptions. A phone preview alone is not the same as checking browser enlargement.

On a touch device, try the task with realistic content and the on-screen keyboard open. Check whether important instructions or the active field become hidden. Closely packed controls and tiny close buttons can make an otherwise straightforward form difficult to operate.

With a screen reader, ask a person familiar with the chosen setup to review the task's structure, names, states and feedback. Record the browser and assistive technology used. A novice tester becoming lost in their own tool is different from a reproducible website defect; the evidence should help distinguish them.

Keep automated scans as an additional source of findings. Investigate their results and combine them with manual task evidence. Neither a clean scan nor one successful screen-reader session proves that every customer, page and interaction is covered.

Turn findings into repairs that survive future releases

Use a concise finding record: page and state, task, reproduction steps, observed behavior, expected behavior, affected users, evidence, owner and retest result. Separate a confirmed defect from a question requiring specialist investigation. Include the actual release or component version where practical.

Prioritize barriers that prevent essential tasks, then widespread shared-component problems and other meaningful friction. Do not rank only by how often analytics detects the issue: people unable to navigate or submit may never create the event you expected to measure.

Fix shared components at their source and retest representative uses. A change to a common field, menu button or modal can repair many pages, but it can also break a less common configuration. Keep a small repeatable set of customer tasks alongside component checks so future design changes exercise both.

The website reliability guide helps assign that ongoing ownership. The product-choice guide connects usable controls with understandable product decisions. Accessibility work should leave the team able to maintain a journey customers can complete, rather than only a report of problems someone once found.

Questions about navigation and form accessibility

Does this review establish WCAG conformance for the website?

No. It is a focused starting review of selected tasks and states. A conformance assessment requires an appropriate scope, method and evidence across the applicable criteria. Record what was tested, repair confirmed barriers and arrange the deeper review needed for the site's users and obligations.

Can a plugin fix all of these issues automatically?

Do not assume it can. A tool may help identify or improve particular behavior, but the actual menu, form rules, focus handling and server result still need testing. Judge any proposed solution by the customer task it repairs and verify that it does not introduce a new barrier elsewhere.

Should we redesign the whole site before fixing a broken menu or form?

Usually a confirmed barrier can be addressed in the responsible component while a wider design decision is considered. Preserve the working behavior, define acceptance criteria and retest the complete task. A large redesign is not a prerequisite for making an essential control operable and understandable.

Continue with the website and commerce platform guides for related work on dependable customer journeys, implementation and ongoing maintenance.

Sources

  1. www.w3.org — preliminary
  2. www.w3.org — menus
  3. www.w3.org — disclosure-navigation
  4. www.w3.org — focus-not-obscured-minimum.html
  5. www.w3.org — labels
  6. www.w3.org — validation
  7. www.w3.org — notifications
  8. www.w3.org — reflow.html

Still have a question?Ask the DTCUP Community ↗

FAQ

Questions related to this insight.

Clear scope, practical constraints and direct answers for the decision ahead.

Does this review establish WCAG conformance for the website?

No. It is a focused starting review of selected tasks and states. A conformance assessment requires an appropriate scope, method and evidence across the applicable criteria. Record what was tested, repair confirmed barriers and arrange the deeper review needed for the site's users and obligations.

Can a plugin fix all of these issues automatically?

Do not assume it can. A tool may help identify or improve particular behavior, but the actual menu, form rules, focus handling and server result still need testing. Judge any proposed solution by the customer task it repairs and verify that it does not introduce a new barrier elsewhere.

Should we redesign the whole site before fixing a broken menu or form?

Usually a confirmed barrier can be addressed in the responsible component while a wider design decision is considered. Preserve the working behavior, define acceptance criteria and retest the complete task. A large redesign is not a prerequisite for making an essential control operable and understandable.

Your workspace

Selected services

DTCUP / YOUR NEXT PERSPECTIVE

Stay curious.
Stay connected.

Ideas for the work behind growth—websites, brands and the apps that connect them.