Designing Search That Converts: Lessons From Travel
Designing Search That Converts: Lessons From Travel
Search is where most travel journeys begin, and where a lot of them quietly end. At lastminute.com I spent years building destination search for flights and hotels, and the thing I learned over and over is that search isn't a box, it's the first promise you make to a user about whether this product understands what they want.
The hardest part of search is the empty box
A blank search field is intimidating. Users don't always know the exact name of the place, the airport code, or even the region. So one of the highest-leverage things we shipped was a search-box suggester, gently guiding people toward what they meant before they finished typing.
It sounds small. It wasn't. Reducing the moment of "what do I even type here" is the difference between a user starting a journey and bouncing. When we expanded search options with 10+ new features, the ones that helped people *form* a query consistently outperformed the ones that just returned better results.
Polygon search taught me to model the world like a user sees it
We built an in-house polygon-based search and layered it with Mapbox maps, because travel destinations don't fit neatly into a list. People search by "near the beach," "around the old town," "somewhere close to the airport"fuzzy, spatial intent that a flat keyword index can't satisfy.
Drawing search as polygons on a map let us match how people actually think about place. The lesson generalised far beyond travel: when you model your data the way users model the world, conversion follows almost on its own, and bounce rates drop because results stop feeling random.
Conversion is a chain, and search is only the first link
It's tempting to obsess over search ranking and forget that search exists to start a funnel, not finish one. I kept roadmap prioritisation anchored to the whole journey, suggester, results, customised result pages, and the path into checkout, measured against the metrics that mattered to the business, not just relevance scores.
Customised search results reduced bounce rates because they respected context. But the win only counted when it carried through to a booking. Optimising one link in isolation is how you get a beautiful search page that converts nothing.
What I carry into every product
Stay close to the data, I leaned heavily on Google Analytics to see where people actually dropped off rather than where I assumed they did. Write the user story down until it's unambiguous, ship the feature that removes friction at the start of the journey, and always measure conversion across the whole funnel, not the screen you happen to be staring at.
Good search feels like the product read your mind. Behind that feeling is a lot of unglamorous work modelling intent, suggesting before users ask, and refusing to call a feature done until the funnel agrees.
Chabi skipped presentations and built real AI products.
Chabi Anand was part of the April 2026 cohort at Curious PM, alongside 18 other talented participants.