A Hotel's Website Loses 98% of Visitors Without Booking — Lighthouse's Predictive AI Recovered €30,545 of It in One Test

The average hotel website loses 98% of its visitors without a booking, and the usual reason is that every visitor sees the same homepage, the same offer, and the same urgency messaging regardless of where they actually are in the decision — a browser three months out gets the same push as someone ready to book tonight. Lighthouse Direct (formerly OTA Insight) built a predictive-AI layer to fix that: five real-time scoring algorithms — intent, spend, destination flexibility, date flexibility, and length of stay — drawing on more than 400 behavioral and contextual signals across 80,000-plus hotels and hundreds of millions of sessions.

The number that makes this more than a pitch deck: THE THIEF, a hotel in Oslo, tested the intent algorithm alone and reported a 25% conversion uplift among low-intent visitors, €26,000 saved in promotional spend that would otherwise have gone to visitors unlikely to book anyway, and €30,545 in incremental revenue converted from visitors the site would previously have written off. That’s personalization measured the way a revenue manager actually measures things — in euros, not engagement.

It fits a broader shift in how PMS and BI platforms are wiring together. Cloudbeds deepened its own integration with Lighthouse this month, auto-flowing reservation, occupancy, and rate data straight into Lighthouse Performance and Distribution — Cloudbeds’ Sebastien Leitner frames the goal as killing “hours wrangling spreadsheets just to understand where next month’s revenue is coming from.” Radisson’s real-time AI price-matching, live across all 1,575-plus of its hotels, shows the same pattern from the distribution side: replacing a manual, claim-based process with something the system just does continuously, without a guest filing a claim or staff reviewing it.

The common thread across all three: AI is moving out of the marketing team’s toolkit and into revenue management’s actual workflow — not generating content, but making the pricing and personalization decisions the old workflow required a person to make one visitor at a time.