Case study: AI search for a travel marketplace

Context

Users were spending time on complex filters instead of simply describing the trip they wanted. On top of that, offers from over 100 suppliers had to be brought together in one place.

Task

Build an AI search that parses a request written in natural language and produces personal recommendations without complex filters.

Implementation

• Request parsing: the module breaks a request down into trip parameters, meaning dates, budget, type of holiday and preferences
• Content search: the parsed parameters are matched against supplier offers to produce a relevant result set
• Recommendation engine: offers are ranked in real time and assembled into a personal selection for the individual user
• Supplier data integration: data from 100+ suppliers is processed and normalised for a single search across the platform
• Supplier catalogue: status, ratings and revenue per service
• Booking schedule: slots, load and orders per supplier

Delivered in 3 months.

Results

• Search takes a request in plain language, with no filters
• A single supplier base: data from over 100 sources in one search
• Search is 3 times more accurate than filter and keyword solutions
• Deep personalisation: the selection reflects each user's preferences

The platform owner sees bookings, revenue and registrations in one dashboard.