Central Bank of Brazil imposed an urgent deadline:
implement identity verification to comply with new regulations.
BACKGROUND CHECKS, FACIAL BIOMETRICS
A CASE STUDY DOCUMENTARY
In collaboration with: Product Managers, Devs and Legal
No room for robust notification systems. I used an existing modal component and worked with legal and marketing to make language readable .
Engineering had limited resources. Legal needed specific language.
Three months post-launch, the modal was underperforming,
I ran funnel analysis, heatmaps and session recordings to find where users were actually dropping off. The data showed two patterns, users were either
Closing the modal immediately
treating it like an ad and never starting the flow or
Clicking through without reading
reaching the personal data page without understanding why they were there, and dropping off.
I had to do something, but within the same constraints:
- It had to stay a modal.
- No backend changes.
-Legal language stayed the same.
Collaborative Ideation:
I ran a collaborative session with the design team to explore improvements within those limitations.
The main goal : increase awareness of why verification mattered before asking users to act.
Collaborative Ideation
2022
it had to stay a modal, no backend or legal text changes
Clearer value and action
The second version states exactly what to do and why, reducing uncertainty and increasing click intent.
Lower perceived effort and time
Listing concrete requirements and reinforcing the short duration helps users feel the action is fast and low-commitment.
Better visual scannability and focus
A cleaner layout with icons and tighter grouping makes the key message and primary CTA easier to spot
2022
HYPOTHESIS
IF the modal gives enough context and redirect users straight to the first form ,
THEN users will better understand the requirement and complete verification at a higher rate.
Success criteria:
When we ran the A/B test, the control group's address page reach had declined to 14.47%, even lower than the initial 21%. That became our baseline.
Each of the variants were exposed to 4000 users.
The Test:
Address page reach increases by at least 20% compared to the control group.
The test variant moved address page reach from
Other than that, some other strong results appeared