contact

Adaptive Booking Guidance

Adaptive Booking Guidance

Improving hotels.com natural language search experience for ambiguous queries using guided destination resolution.

Improving hotels.com natural language search experience for ambiguous queries using guided destination resolution.

Improving hotels.com natural language search experience for ambiguous queries using guided destination resolution.

Oct 2026

TL;DR ⋆✴︎˚。⋆

Adaptive Booking Guidance (ABG) is an intelligent guidance layer for Hotels.com natural language search (NLS), designed to work alongside NERD, the service that resolves destinations from a traveller's query. It addresses a pivotal moment in travel search: when a traveller knows the kind of trip they want, but are still unsure of where to go. Instead of silently collapsing an ambiguous request into one destination, ABG gives the traveller meaningful, explainable paths to explore, then handing their choice back to the existing hotel search flow.


Note: This was an internal hackathon prototype, not a live production launch. Hotel content, pricing and availability shown in the prototype are illustrative; no post-launch or experiment results are claimed here.


Context

Hotels.com is currently rolling out Natural Language Search in the USA.

NL Search Product and Tech Metric Review, 29 Sept 2026

Research

The project drew on existing internal research and an assessment of the current search interface. The interface review highlighted three main issues.

“

AI Agent on Hotels.com,

June 2025 report, Kira Pinard-Welyczko.

Natural-language travel search returning a single destination, with interface annotations
The use of the term ‘exact match’ indicates a level of certainty for a query which lacks a specific enough intent, creating an unhinged sense of confidence.
Collapses the whole world into one city; shows no alternatives or indications that other options could be available to the traveller.
The current map takes up a large amount of page real estate despite unidentified user destination intent which is not useful.
How has the agent decided that features such as ‘outdoor pool’ or ‘bar lounge’ are appropriate for this query? An alternative property shows different amenities such as ‘full service spa’ and ‘southern cuisine restaurant’. Currently a lack of explainability.

Our hypothesis was that making these possibilities visible could help travellers refine their intentions, choose a relevant destination, and continue towards a booking.

Design

ADAPT THE GUIDANCE TO THE STARTING POINT

MAKE DIFFERENT INTERPRETATIONS VISIBLE

KEEP EXPLORATIONS CLOSE TO THE RESULTS

USE FAMILIAR COMPONENTS

We reused cards, carousel, and tab patterns to connect the guidance to the existing search experience. There was a focus on when each pattern should appear

Familiar search interface components

The destination carousel allows users to explore a selection of suggested destination that best fit the users ambiguous to assist in narrowing for destination intent.

Build

Frontend designs and interactive prototypes were built in Figma and Cursor, working alongside teammates responsible for machine learning, data, backend development and product strategy.

Reflection

this website was made by nyree marsh >_<

Prototype A

Prototype B

Prototype C