Research Guides
Explain a Research Upload Before Asking for the File
A helpful upload form starts explaining itself before the first error. Someone preparing a research dataset needs to know what the form accepts, which information is required and what will happen after selection. Hiding those answers until submission wastes work and can encourage people to share the wrong file.
This is a manual design-review worksheet for an authorized form or test environment. It is not a description of Colquhoun.shop's upload behavior, a privacy assessment or a complete accessibility audit.
1. Separate requirements from examples
Write down the actual accepted format, size limit and required fields from the implementation or verified product specification. A filename ending in CSV does not establish that every spreadsheet export will work. If the parser expects particular headers or a particular date representation, the instructions should say so.
Mark examples as examples. In a fictional research form, a sample containing one website and one month should not imply that the real upload must describe that same website or reporting period. Give the user a way to understand the structure without borrowing demonstration figures as evidence.
2. Put help where the decision happens
W3C's forms guidance recommends explaining relevant requirements, optional input and data formats. It describes associating additional instructions with a field, including through aria-describedby, and warns that placeholder text is not a substitute for a label. W3C: Form Instructions
Place the file requirements before or beside the selection control. Keep essential guidance available after a file has been chosen. A user reviewing the selection should not need to clear the field to recover an example that disappeared.
3. Test an instruction-only preparation task
Give a reviewer harmless invented data and the published instructions, but not an explanation from the development team. Ask them to prepare a valid sample. Record each question they have before attempting submission.
For an invented monthly research import, the reviewer might ask whether “visits” means a monthly count or daily average, whether dates identify a month or a day, and whether a missing measurement can be blank. Those are meaning questions. A file can satisfy a parser while answering them incorrectly.
Revise the instructions to resolve the questions that the product can actually answer. Do not invent accepted values or silently broaden the claimed format merely to make the prose friendlier.
4. Verify what selection and submission do
Check whether choosing a file begins processing immediately or whether a separate action is required. Explain the observed behavior accurately. Do not write “nothing leaves your device” or “your file is deleted immediately” without verifying the relevant processing and retention behavior.
Use synthetic, non-sensitive fixtures for the review. If the form triggers a paid analysis, sends a message or creates a real record, obtain the appropriate authorization and safe test procedure before exercising it. Stop at the boundary you cannot safely test and record the gap.
5. Keep an instruction-to-behavior checklist
- Requirement: exact format, limits and field meanings.
- Evidence: implementation or specification supporting each instruction.
- Placement: where help appears before and after selection.
- Association: which control each instruction describes.
- Preparation: whether a reviewer could create a valid synthetic sample.
- Behavior: what selection and submission actually trigger.
- Limits: untested accessibility, processing or retention claims.
Recheck the instructions when the accepted schema or processing behavior changes. Clear pre-submission guidance and useful error recovery solve different problems; a good error message should not be the first place a reader learns the rules.