Customer Data Platform (CDP) at Qunar Travel: Business Background, Construction Practice, Applications, and Future Outlook

Qunar Travel, a leading Chinese online travel platform, built a Customer Data Platform (CDP) to support fine‑grained operations across its multiple business lines, achieving billion‑level revenue gains and winning a company gold award.

The CDP addresses key operational pain points such as data silos, low tag accuracy, high development cost, and slow iteration cycles by abstracting the entire operation process into four core steps: image tagging, audience selection, strategy configuration, and effect analysis.

Architecturally, the CDP sits in the data‑application layer, leveraging a foundation built on Hive, Trino, Flink, Hadoop, HBase, Kafka, and other components to provide both offline and real‑time tag generation, with services exposed via Dubbo and HTTP APIs backed by Redis, HBase, and ClickHouse.

Core capabilities include a flexible tag system (statistical, rule‑based, model‑based tags for users and products), rigorous tag quality inspection, and interactive audience segmentation using bitmap operations in ClickHouse to achieve sub‑second query responses.

The platform ensures high availability and performance through containerized deployment, horizontal auto‑scaling, multi‑level caching, circuit‑breaker and rate‑limiting mechanisms, as well as automated versioning and failover strategies.

Real‑time tags are built with FlinkSQL and visual configuration, lowering development barriers and enabling seamless integration with offline tags, while the system supports diverse user reach channels (push, SMS, in‑app messages) and detailed effect analysis for continuous optimization.

Business outcomes include a five‑fold ROI increase, three‑fold operational efficiency gains, 99.99% service availability with peak QPS over 300,000, and a closed‑loop workflow that drives revenue, user activation, and fraud mitigation.

Future plans focus on expanding model‑driven tags, covering more scenarios through BI integration, and introducing intelligent strategy selection to further automate and personalize marketing efforts.