Travel Offer Data Lineage

Travel offer lineage means being able to explain which supplier response, mapping rules, normalization steps, cache state and ranking decision produced the offer shown to a traveler.

Production scenario

A traveler clicks a €210 refundable offer but lands on €228 non-refundable inventory. UI logs alone cannot explain the root cause; the complete transformation chain is required.

Lineage flow

Supplier Request
 -> Raw Response
 -> Parser Version
 -> Property Mapping
 -> Room/Rate Mapping
 -> Price Normalization
 -> Offer Fingerprint
 -> Cache Entry
 -> Ranking Features
 -> Rendered Result
 -> Click / Reprice / Booking

Lineage identifiers

Useful identifiers include correlation ID, raw-response reference, provider, parser/schema version, canonical property ID, mapping decision ID, offer fingerprint, price-observation ID, cache version, ranking decision ID and click/booking reference.

Raw-data retention

Keeping every raw payload forever may be too expensive. Short retention, immutable object storage and hashed references can preserve evidence while controlling cost.

PII and contractual retention rules still apply.

Versioning

Lineage is incomplete without transformation versions. Record parser, normalization and ranking-rule versions.

Failure modes

Common gaps include lost source references, unaudited mappings, unknown cache versions, missing feature snapshots and unnecessary PII copied into lineage records.

RCA use

An operator should be able to answer: “Which raw source, rule version and observation timestamp produced the exact offer shown to this traveler?”

Observability

Track lineage completeness, missing source/mapping references, untraceable rendered offers, RCA resolution time and replay success.

Alternatives

For small systems, correlation IDs plus structured logs may be sufficient. Distributed multi-provider platforms benefit from explicit lineage records or stores.

Production checklist

Use stable correlation IDs, source references, transformation versions, mapping decision IDs, fingerprint/observation links, ranking-decision links, retention policies, PII minimization and replay tooling.