Start with the decision the form supports
Write down what the service needs to know and what the person should be able to do after submitting. For each field, ask whether it is necessary now, whether the requested format is clear, and whether the service already has the information. Every extra question creates work and may require an explanation.
Check the wording before reviewing the layout. A label should still make sense when a person revisits a field after an error; placeholder text alone is a fragile substitute because it may disappear as the person types.
Review the complete field lifecycle
- Before entry: are required fields, formats, and sensitive requests explained in time?
- During entry: can the person use a keyboard, paste, autocomplete, and assistive technology where appropriate?
- On error: does the message identify the field, explain the issue, and preserve valid answers?
- On success: is the next step or confirmation unmistakable, including any expected follow-up?
Test realistic and difficult values
Use plausible long names, international phone numbers, different address formats, optional fields, and values near any documented limit. Test empty, malformed, and corrected input. Do not assume that a field’s validation message will be helpful merely because it appears; check whether it tells the person how to recover.
Avoid treating one locale’s address or name conventions as universal. If the service supports multiple regions, confirm that the accepted input matches the actual audience and the downstream system can handle it.
Check accessibility and trust
Use the keyboard through the whole form. Confirm that labels and instructions are programmatically associated with controls, focus moves sensibly after errors, and errors are available to assistive technology. Explain why sensitive information is needed and how it will be used before asking for it. A quick review is a useful screen, not a conformance assessment.
Record a finding that can be reproduced
Include the page, field, value or state, action, observed result, expected result, and whether the entered data was retained. Redact personal data from captures. If you cannot submit to a production service safely, use a test environment and say which behavior you could not verify.
Use Spec to review a page with your team and keep observations connected to the experience.
Explore Spec