Hotel Recommendation Engine · Documentation

User Journey

The end-to-end path through Hotel Recommendation Engine, from trigger to outcome.

Live diagram — browsing signals reorder the rail; the recorded output feeds #31 Upsell and #11 Segmentation off the one shared behaviour log.

1
Behaviour → ranked rail
As the guest browses, `behavior.recorded` / `guest.viewed_property` signals land on the shared behaviour log. `rankProperties` reads them, computes region affinity and per-property affinity (weighting bookings > saves > detail-expands > views), adds rating and popularity, and boosts properties whose facilities match the guest's profile interests. It sorts, applies the ≤2-per-region diversity cap, and returns the top 6 with reasons.
2
Ranking → the rest of the ecosystem
The ranked output the guest saw is recorded (`demo_eco_recommendation`, with reasons + mode) and feeds downstream: it gives #31 Upsell its personalisation context and #11 Segmentation a behavioural signal. Because everything reads the one shared `behavior_event` log and one property catalog, the recommendations stay consistent with what the guest actually did — no app keeps a private copy.