The brief was clear and the first version matched it precisely: a single, thorough onboarding form covering everything a new seller needed to provide before listing their first product. It worked as specified and almost nobody finished it.
What the brief got right and still missed
Every field in the form was genuinely needed eventually — banking details, tax information, product categories, shipping policies. The brief was correct about what was required. It was silent on when, and "all of it, up front, before you can do anything" turned out to be a decision nobody had actually made, just one that fell out of building a single form for a single requirements list.
What the drop-off data showed
Roughly two-thirds of sellers who started the form abandoned it before the banking details section, which sat in the middle. Sellers who'd come to list one product were being asked to complete tax paperwork before they'd seen any evidence the platform would work for them.
The rebuild's actual change
Not a redesign of the form's fields, which were fine — a redesign of its sequencing. Sellers could now list a product and see it live with only the minimum required to do that safely, and were prompted for banking and tax details at the point they actually needed them: before their first sale could pay out, not before their first listing could go live.
- Defer anything not required for the immediate next step, rather than front-loading everything the eventual complete profile needs.
- Make the deferred requirement visible but not blocking, so a seller knows banking details are coming and isn't surprised, without being stopped by them prematurely.
- Measure completion per step, not just overall, because an aggregate completion rate hides exactly which step is losing people.
Why we didn't catch this in the original scoping
The brief described a complete, correct set of requirements and we built precisely that. What discovery hadn't asked was in what order a seller would tolerate providing them, because that's a behavioural question that a requirements list doesn't naturally surface — it takes real usage data to answer.
A form can collect exactly the right information and still fail, if it asks for all of it before the person filling it in has any reason to trust the process.
What changed after the rebuild
Completion through to a first live listing improved substantially, and the client's support queue saw fewer questions along the lines of "why do you need my bank details to just list a product", because by that point in the flow, sellers understood why.