Context
A recurring compliance requirement within a supplier verification platform was generating a high resubmission rate, with a meaningful share of users failing to recover even after a second attempt. I led a data-informed content design review to diagnose the cause and recommend fixes.
The Problem
Initial assumption was that friction was spread evenly across the requirement. Analysis of ~150,000 question submissions across 5,000+ organisations showed otherwise: friction was concentrated in a small handful of document-upload questions, which accounted for the large majority of rejections. Of those rejected, most did not pass on their next attempt, pointing to a deeper issue than a one-off mistake.
What the Data Showed
- A small subset of upload questions accounted for roughly 70% of all rejections in the requirement
- Nearly a quarter of users who resubmitted failed again
- The same uploaded evidence, submitted across multiple related questions, was frequently accepted for some and rejected for others suggesting inconsistent evidence interpretation rather than a genuine quality gap
- Guidance text described the intent of what was required but rarely defined what an acceptable upload actually looked like
- Rejection feedback relied on generic, reusable messages that didn't identify the specific issue with a submission
My Approach
Rather than treating this as a compliance-only problem, I reframed it as an experience design issue: users were engaging with the requirement, but the content and structure weren't consistently helping them complete it correctly. I:
- Audited question wording for ambiguity, inconsistent subject framing (e.g. shifting between "you" and "your company"), and unnecessary cognitive load
- Identified where answer taxonomies (e.g. Yes/Not Applicable, with no No option) were forcing users into inaccurate responses
- Mapped where multiple questions were requesting effectively the same underlying evidence, creating duplicated upload effort
- Recommended more defect-specific rejection feedback, so users understand what to change rather than just that something was wrong
Recommendations Delivered
- Rewrite question text to remove ambiguity and align to the content design system
- Rewrite guidance to define acceptable evidence in plain, actionable terms
- Correct answer taxonomies so exemption, negative response, and uncertainty aren't collapsed into one option