Uber
Redesigning Uber's Australia tax flow to turn a 24-hour identity bottleneck into a provisional fast path
Results
Results
100%
Task completion across usability testing, zero facilitator intervention
Results
0%
Abandonment across all six participants
Results
6.35/7
Average ease-of-use score
Team
Product Design
Product Management & Engineering
Tech stack
Design & AI
Role
Led end-to-end design, partnering closely with our UXR — who ran usability testing — and the PM throughout, under an active company-wide supply-growth escalation. Owned the provisional-activation mechanism and its fraud-safe validation logic, from concept through the current pre-launch stage.
Overview
28% of Delivery signups raised a support ticket tied to their ABN, and 41% of those earners never activated. The cause was structural: a fast tax ID check sat gated behind a slower identity-matching process that could take up to 24 hours, with generic errors and no upfront disclosure of required documents. I redesigned the flow around provisional activation — validating the tax ID fast, letting earners start earning immediately, and letting the slower identity match complete in the background with an automatic block on any mismatch.
Insights
1 gate, 2 checks
ABN validation and identity name-matching were bundled into a single step — slow enough to create real friction exactly when intent was highest.
Drop-off after the hardest step
Drivers who'd already cleared a paid background check still fell out at the ABN stage — the point that should've felt safest was where people quit.
No way to recover from a mistake
No upfront disclosure of required documents, delayed validation, and no in-flow error recovery meant one wrong step sent people back to the start.
Locked fields read as broken
Disabled styling on read-only fields and an unscannable disclaimer made the form feel unreliable before drivers even reached the real question.
Testing & iterations
Participants understood every screen's purpose on sight — the checklist, the form, the review. What broke them was ambiguous status labels, tax jargon, and error states that named a problem without explaining it. From there, I focused on specific fixes that replaced ambiguity with clarity at each failure point.
