Research Guides
Turn a Research Form Error into a Clear Next Step
A research form should not leave someone guessing whether their website address was rejected, a source was unavailable or a request was saved. Those are different outcomes, even when all three interrupt the same workflow. Review the recovery path as carefully as the successful submission.
This worksheet is a practical review method, not a claim that any particular product has implemented these behaviors or passed an accessibility audit.
1. Classify the failure before writing the message
Use an invented competitor-research form with a website field and a reporting period. Write three test cases: a missing website, an end date before the start date and a temporary source timeout. The first two ask the user to change input. The third does not establish that the input was wrong.
Record the actual condition beside the proposed message. Avoid calling every failed request “invalid.” A reader should not spend time editing a valid address because a service failed elsewhere.
2. Name the field and give a usable repair
W3C's explanation of Error Suggestion says that when an input error is detected automatically and a correction is known, the user should receive the suggestion, except where doing so compromises security or the content's purpose. W3C: Error Suggestion
For the invented date case, compare “Request failed” with “End date must be on or after start date. Change either date, then submit again.” The second identifies the relationship that failed without inventing the date the researcher intended.
For the website field, explain the accepted format only if that is genuinely the problem. Do not silently substitute another company or normalize a value in a way that changes the research target.
3. Make the correction findable
W3C's forms tutorial recommends clear feedback on submission outcomes. Its error-list guidance includes naming the affected control, explaining the problem and correction, and linking to the relevant field. It also discusses inline feedback and ways to communicate dynamically inserted errors. W3C: User Notification
As a review exercise, submit the test form without using a mouse. Can you locate the message, reach the affected field and understand what to change? Repeat with the assistive technology used in your test plan. A visible banner alone does not demonstrate that the update was announced appropriately.
Keep this separate from testing whether the research controls themselves work: see the keyboard research-filter worksheet.
4. Protect the work already entered
For non-sensitive research inputs, check whether unrelated valid fields survive a failed submission. If a form asks for five competitors and only one address needs correction, retyping the other four adds no research value. Do not use this principle as a reason to retain passwords or other sensitive entries unnecessarily.
For an uncertain submission outcome, avoid promising that nothing was saved unless the system can establish that fact. The recovery message should reflect the known state: rejected before processing, still pending, completed, or unknown. A retry should not be presented as harmless when it could create duplicate work.
5. Keep a small acceptance record
- Input: the harmless test values and the condition being exercised.
- Outcome: what the system actually did, not just the banner it displayed.
- Message: the exact text, affected field and proposed next action.
- Recovery: whether the correction succeeds without losing unrelated work.
- Access: keyboard and assistive-technology observations, including untested cases.
Close the test only after submitting the corrected form and checking the resulting state. “The error disappeared” and “the intended research request was accepted” are not interchangeable observations.