Every major booking platform claims to offer personalised recommendations. None of them do, in any meaningful sense. What they offer is targeted re-marketing: you looked at Barcelona once, so now you see Barcelona hotel ads everywhere for three weeks.

Real personalisation is a harder problem. It requires understanding the purpose of a trip — is this a long weekend with a partner, a solo conference trip, a family holiday — and adjusting every dimension of the recommendation accordingly. A business traveller and a backpacker don't just want different price points; they want different proximity to transit, different amenity profiles, different cancellation policies.

The structural reason existing platforms fail at this is incentive misalignment. Their revenue comes from commissions on bookings and sponsored placements. A truly personalised result might surface a smaller hotel with lower commission rates over a premium chain that pays more for visibility. The platform's interests and the user's interests are in direct conflict.

The second reason is technical debt. Most legacy booking systems were built on top of GDS infrastructure from the 1990s. Layering modern machine learning on top of that is expensive and slow. The incumbents have opted for cosmetic personalisation — 'based on your recent searches' — rather than investing in the underlying data infrastructure needed for genuine fit prediction.

Roavo is built without those constraints. We do not take placement fees. Our ranking function has one objective: match the traveller to the best available option given their stated and inferred preferences. The business model depends on users trusting the results, which means the incentives point in the right direction.