Bengal is five holidays inside one state, and nobody sells it that way
Hills at 12,000 feet, a tiger delta, a 200-kilometre coast, terracotta temple towns and a colonial river all sit inside an eight-hour radius. The composer scores 32 places in West Bengal against your month, your budget and your constraints — and tells you what it left out.
Indicative shoulder-season rates for the tier selected, not live inventory. A production build prices against connected supply and the WBTDC property list.
The question a booking form can't take
"My mother can't manage stairs and we have four days in November" is not a set of filters. The concierge answers from the same Bengal dataset the composer uses — so it can be specific without being invented.
The reply is assembled from those fields. In production the same fields go to Claude as tool results, so it writes freely and still cannot invent a road that isn't there.
What a Bengal property should charge on the 14th
Occupancy forecast from booking pace, competitor rates, the mela and puja calendar and search interest — then the rate that maximises revenue rather than the rate that fills rooms.
Next 21 nights
Nine hundred reviews, four things to fix
A star rating tells an owner they have a problem. This tells them the problem is the 20-minute check-in queue at 2 PM, that it costs them 0.4 of a star, and that it started in March.
Fixing the two operational themes below moves the modelled rating to 4.5. Neither needs capital spend.
Themes are clustered and scored live in this page from the review text using an aspect lexicon. Production uses embeddings for clustering and Claude for the action write-up.