Payment checkout — redesigned at scale, live.
Redesigning the most critical step in the airline booking funnel — at scale, across markets, with zero margin for error.
Four windows. Zero clarity.
LATAM's payment page spread one decision across four surfaces: a main window just to read which payment method was selected, plus three modals stacked on top of it — one to select, one to add, one to fill out billing information. Every step meant losing the context of the one before it — generating friction, errors, and customer claims on the final gate before a booking became revenue.
Highest-traffic flow, zero margin for error.
For the business: this was the final conversion gate for millions of booking journeys across Latin America — one of the highest-traffic checkout flows in the region, and any redesign had to roll out without disrupting it. For the user: travellers already committed to a purchase were losing the thread of their own payment, across two distinct journeys — regular checkout and miles redemption — on both desktop and mobile web.
The redesign — and the rollout.
Senior Product Designer leading the checkout redesign end to end: the single-page flow, a reusable payment-method component adopted company-wide, and the phased rollout strategy — coordinating product, engineering, QA and market operations across countries.
One page, not four windows.
The redesign collapsed all four surfaces into a single page, with what's selected and what's available made unambiguous at a glance instead of inferred window by window. Rather than solve this once for this page, I designed it as a single custom component with variants per payment method — the same building block reused across every online payment touchpoint, web and mobile, replacing whatever each product had improvised on its own.
Four windows, one page. Start to finish, no context lost.
Selected and available, clear at a glance.
- 01Single page — one main window plus three modals (select, add, bill) collapsed into one flow
- 02Payment-method clarity — selected vs. available states redesigned to remove ambiguity
- 03Time on task — down from 6 to 4 minutes
- 04One component, company-wide — a single payment-method component with variants, reused across web and mobile touchpoints
The company-wide rollout of that component took a year and a half, reached 7 other digital products, and each one saw its own conversion rate improve.
Ship it like a live experiment.
A phased rollout strategy treated the release as a live experiment — scaling traffic per market only when conversion data confirmed improvement, and pulling back when issues emerged.
The regular flow took six iterations across three weeks. Starting at −10% with 10% of traffic and landing at +13% at full traffic wasn't turbulence — it was the process working exactly as designed. We fixed a debit offsite issue mid-flight and kept moving. The miles redemption flow followed a cleaner path across four markets, closing at +3% at full rollout.
Scale traffic when the data agrees. Pull back when it doesn't.
Up at full traffic, on both flows.
Beyond conversion: faster delivery and fewer inter-team PR review cycles, a byproduct of a rollout process that surfaced issues early instead of at launch.
Coordination as design work.
At this scale, redesign is also a coordination problem. Aligning product, engineering, QA, and market operations across multiple countries — while maintaining rollout discipline — demands as much design thinking as the interface itself.
The interface was the easy part. Holding the rollout together wasn't.