A lending client was losing two-thirds of applicants during onboarding. Eleven screens, an average of 14 minutes to complete, and a 31% completion rate. We rebuilt it to four screens and 68% completion over eleven weeks.
Where people actually dropped
Session recordings and funnel analytics disagreed with the team's assumptions. Everyone believed document upload was the killer. It was fourth. The three biggest drops were:
Screen 2 — income details (24% loss). It asked for gross annual income, net monthly income and employer PAN on one screen with no explanation of why. People left to find documents and never came back.
Screen 6 — address (19% loss). Four address fields with strict pin-code validation that rejected valid rural codes.
Screen 9 — review (16% loss). A wall of text with every field repeated and no way to edit inline. Users hit back, lost state, and gave up.
The three changes that mattered
We asked for less, later. Only what was needed for a provisional decision came first. Everything else moved after the applicant had an answer worth completing for.
We fixed validation before we fixed layout. The pin-code regex alone recovered several points. Unglamorous, and the highest-ROI change in the project.
We made the review screen editable. Inline edit, no navigation, no state loss.
Numbers
Completion 31% → 68%. Median time 14m → 5m20s. Support tickets tagged "onboarding" down 41% in the first month.
The visual redesign happened too. We can't attribute much of the gain to it.
- #Case Study
- #Fintech
- #Onboarding
- #Conversion