Case Study: Hotels.com achieves 20x faster ETL and optimized personalized recommendations with Databricks

Hotels.com - Customer Case Study

Hotels.com, a leading online accommodation booking site with 90 websites in 41 languages and over 325,000 hotels, faced a data-scale challenge: millions of photos (with thousands uploaded daily) needed rapid analysis to remove duplicates and low-quality images, classify scenes (kitchen, pool, gym, etc.), and fuel personalized recommendations. Their on-prem Hadoop/SQL/SAS pipeline was slow—processing even 10% of the data took hours—hindering real-time scoring, ML/deep learning model deployment, and efforts to boost customer conversion and lifetime value.

Moving to Databricks’ Unified Analytics Platform gave Hotels.com scalable cluster management, an interactive workspace for cross-team collaboration, and Databricks Runtime for faster streaming and ML processing. The outcome was a 20x increase in ETL throughput, more accurate and contextually relevant image displays and classifications in search results, and higher conversion rates from better personalized hotel recommendations, enabling Hotels.com to run 100% of its workflows on Databricks.

Hotels.com

Matt Fryer

Vice President, Chief Data Science Officer