A redesign of Despegar's home and search — Latin America's leading OTA for flights, hotels and packages. The challenge: stop the home from competing with itself. Instead of serving three users badly at once, it asks the essentials first and reveals the rest in layers.
Design goalStrip friction from the search-and-book flow, without losing the information density travelers rely on.
Today's home tries to serve three users at once — the one who already knows where they're going, the one who's browsing, and the one coming back to finish something — and ends up serving all three poorly. Too many fields, flat hierarchy, zero memory. Friction from the first second.
"A home that talks to everyone at the same volume talks to no one. The user doesn't choose a path: they get all of them at once, open and competing."
The search box shows every tab and every field at once. The user has to dismiss before they can even start. The primary action — choosing a trip — competes with nine nav items and seven promos.
Banners, installments, coupons and carousels fill half the first view. What should be secondary visually buries the one action 80% come to do: search.
Someone returning to finish a search starts from scratch, just like a new visitor. No recent searches, no price alerts, no saved trips. The home remembers nothing.
Instead of showing everything to everyone, the home figures out the situation the user is in and offers a single dominant path. Three intents, three answers — with a crystal-clear hierarchy between them.
A dominant search box, one vertical visible at a time, minimal fields that expand on interaction. The action 80% come to do sits center stage, with no competition.
Recent searches, price alerts and saved trips. It only appears for the returning user: if the home recognizes you, it brings you right back to where you left off.
Inspiration by budget and destination, no form required. For the one who feels like travelling but has no plan: discover first, search later.
The search box starts with the minimum — from, to, when — and expands class, baggage and flexibility only when the user asks. Less up front, more when it's needed.
The home changes based on who you are. A new visitor sees Search and Explore; a returning one also sees their last search and their alerts. The interface adapts, not the other way around.
A single primary action dominates each screen. Secondary things live in bands below, never fighting for attention. The promo noise that competed with the task is gone.
Four screens of the redesigned flow. The search box is always the dominant action; "Resume" and "Explore" live below as secondary bands. The marked interactions work: try expanding the search box and hovering over the results.
From, to, when. The four fields everyone needs. Nothing more.
Passengers, class, baggage and flexible dates unfold when you tap "more options." They animate in, in context, with no modal.
The fast booker sees no friction; the one who needs control finds it one click away. The same box serves both.
The same home, before and after. Switch between the two views: what used to be nine fields and seven promos competing is now a single question, with everything else in layers.
AI didn't replace the design decisions: it accelerated them. It compressed the mechanical part — synthesizing, varying, drafting, questioning — to free up more time for judgment. Four key moments in the process.
I gathered patterns from six OTAs and my usability notes, and had them summarized into the user's three modes — Search, Resume, Explore — and the friction points of each. From hours of reading to an actionable map.
I generated layout variants of the search box and the home hierarchy to compare fast: dominant search vs. tabs, bands vs. grid. I dropped what competed with the task and kept what served it.
I iterated the tone in Rioplatense Spanish: "Where do you want to go?", "Pick up where you left off", field labels and empty states. I tested several voices until I landed on one that's close and clear, without sounding like a banner.
I used AI as a devil's advocate: what if the user isn't logged in? what about multi-city? what if there are no direct flights? Questioning every decision surfaced the edge cases before prototyping.