Lead designer · problem framing + flow + UI
Institutional clients drive 30–70% of an exchange's trading volume, so every company that gives up at verification is lost revenue from the highest-value segment.
A research report flagged low KYB conversion, and a feature request came with it. Before designing it, I went back to the report's funnel.
November KYB self-submission funnel; each % is conversion from the step above. It starts at company info and excludes homepage traffic the report marked as non-institutional.
Users weren't losing interest. They sent the wrong file, got it back and tried again; the research report and internal interviews traced the same loop.
Share of 86 rejected or in-review applications in November with a problem in each document; one application can fail on several. Shown: the document step. Company and contact-person items were 19% or less.
A company with a board needs a board resolution, one without needs a power of attorney. The rule was one line of grey text; 67% of failed applications got this document wrong, and the research report named it a common blocker.
Each shareholder above 25% adds documents, and further conditions add more, all in one paragraph.
Previous design: the key distinctions sat in dense text in exactly these two modules.
04 · ApproachBinance, OKX, Coinbase and Kraken share three patterns: pick the entity type first, show a document checklist up front, and write requirements in plain language. I tested each against what the data showed.
Institutional verification flows at leading exchanges.
Rejections came from comprehension, not irrelevant fields, so a selector wouldn't move today's numbers. I kept it to route each market's document set as EU and AU rules diverge.
Users already skipped long text, so help text only explains terms; it never carries a rule.
With the rule in text, 67% of failed applications got this document wrong, so the choice became a required step that leads straight to the right document.
The first fix targets the board document (wrong in 67% of failed applications), the second the shareholder list (24%). The third tells companies up front that compliance may ask for more.
Designed as required with nothing pre-selected, so users must pick their structure first. The form then shows only the document that applies, with its template.
Before: the rule was one line under the upload. After: the choice comes first.
What the register must include sits in a highlighted list, and the 25% conditions become two options instead of a paragraph to decode.
Before: two or three requirements mixed into the description. After: each one visible.
An optional last step tells companies compliance may ask for more, and lets them add it up front.
Before: no hint that more documents might follow. After: an optional additional-documents step.
06 · Edge casesInstitutional applications rarely pass in one go, so the states after submission matter as much as the form.
High-risk applicants are asked for more after submission, each item with its reason, so the request doesn't feel like a dead end.
A rejection lists specific, fixable reasons, so companies can correct and resubmit instead of guessing.
Institutions need time to gather documents, so progress is saved and the flow resumes where they left off.
The full flow makes the right-document decision at each fork and says what to prepare before companies start.
The full journey: entry, preparation and the four submission steps.
Key screens: the guided steps where the document decisions happen.
Interactive prototype: pick a company type, then walk the four steps. In step 3 the director choice starts empty and must be answered, as designed (the shipped version had a default). Open full screen ↗
08 · ResultBoth fixes shipped, but in the build the board choice came with a default already selected, so companies could move on without choosing and the core mechanism didn't run as designed. The first-time review pass rate for submitted applications went from 41% to 42%, well short of what we expected, so I took the rejections back to customer support.
KYB EU and AU: the same pattern extended to multi-layer ownership, beneficial-owner thresholds and local source-of-funds rules. Full screens on request.
09 · ReflectionI went through the rejections one by one with customer support and compliance. Most companies had picked the right document and still got it back.
Stamped certified copies, a hand-signed POA, the company name on the register, and original ID photos rather than screenshots.
A hand-held ID with a written signature that no one checked, and ECDD asked of every company when only two entity types need it.
Stamp, original and signature rules belong at the upload, not in a help article.
“Required, no default” lived only in the mockups and was easy to lose in the build. It belongs in the acceptance criteria and a design QA pass before release.
One session with the people who handle rejections shows early which causes sit in policy and the back end.