The signals before
anyone touched
a screen.
Desktop was 25% of traffic. Eighty percent of transaction value. The platform was built for the wrong audience — and no one had named it yet. I built the research case, secured the mandate, and redesigned it. $50B+ GMV ran through every decision.
A mobile-first company
sitting on a desktop problem.
32 cross-functional interviews, a Baymard audit, and gaze tracking on $500+ order sessions all pointed to the same conclusion: the desktop was built for consumer shoppers. The buyers using it were running procurement.
Research signal → design bet → measured outcome.
That’s the structure. The work follows.
The three failures from the diagnostic each became a sprint. First impression, search, transaction — in that order.
B2B credibility
above the fold.
First-time buyers on $500+ orders called it a flea market. No distinction between a consumer browsing and a business sourcing at scale — no logged-out/logged-in state, no B2B signal above the fold. I designed the gate: B2B credibility above the fold, browse-first entry without a credential wall, and a personalized returning-buyer state. The PM worried sign-in would drop. It went up.
Trust signals
at card level.
Search was built for browsing, not sourcing — no trust signals at the card, no B2B sort, MOQ buried three page-loads deep. I redesigned the card to carry what buyers need at a glance: Trade Assurance, response rate, OTD — as primary sort controls, not sidebar filters.
The PDP as
the spreadsheet.
Buyers were building spreadsheets before they’d commit to a conversation because the PDP couldn’t answer the question that mattered: does the math work? I put the math on the page — tier pricing inline, a live order calculator, Trade Assurance in the primary scan zone.
The same three surfaces.
A complete sourcing loop.
I’d design the loop from intent to transaction. Natural language in Chat becomes a structured brief — the agent parses what to buy, at what price, under what conditions — then traverses Search and PDP in real time, evaluating every trust signal against those parameters. When the buying signals clear, the transaction completes autonomously. The buyer defined the intent once. The AI decided what, when, and how to pay.
No form, no re-entry.
lead = ≤20d · cert = trade_assurance
———————————————————
Scanning 12,483 results → filtering TA + RTS…
12K filtered, zero clicks.
Brief confirmed in ms.
Every surface closed.
🛡 Trade Assurance · Lead time 14d
Total $4,950 · budget $5,000 ✓
Every surface in the sourcing loop carries structured data, accurate trust signals, and inline pricing. The loop closes on the platform — no spreadsheet required.