All posts
Engineering
Normalising hotel content across suppliers
How we map duplicate properties across bed banks into a single property ID, and why it matters for conversion.
Data Engineering 8 min read
The same hotel often appears several times across suppliers, with slightly different names, coordinates and room descriptions. Without mapping, customers see duplicates and you lose the chance to show the best rate.
Our approach
- Candidate matching on geo-distance, normalised name and address tokens
- Scoring with supplier codes such as GIATA where available
- Human review for low-confidence matches
- Continuous re-mapping as suppliers update their content
Rates from every supplier are returned under one property_id, so rate comparison works without extra code.
