A smooth ride: Online car buying and selling at mobile.de
summary
mobile.de runs a suite of production data products: price prediction using XGBoost/H2O with BigQuery + Airflow and a Scala REST API, a personalized search ranking using historical snapshots and a learning to rank lambdaMART model deployed to Elasticsearch LTR, weekly-retrained fraud detection stacks, and financing-intent scores. Site stats: 12 million monthly users, 1.6 million live listings, 40k dealer customers. Early LTR A/B tests show a 6–7% uplift versus business-rule ranking, later iterations net 1–2% gains, and user exposure to fraud falls below 1% overall.
AI-written summary · about
chapters (6)
- 0:00Introduction to mobile.de
- 7:40Search ranking with ML
- 13:22Market price prediction
- 19:01Fraud detection system
- 23:55Account takeover prevention
- 24:35Q&A session
related talks
20mArchitecture of relevancy search at mobile.de
43m“Which car fits my life?” - mobile.de’s approach to recommendations
25mBetter search relevance using Learning to Rank at mobile.de
44mRobust AI Search Ranking for Radical C2C Marketplace Growth
42mWhich Car Fits My Life? - mobile.de’s Approach to Recommendations
39mBuilding a personalised home feed for Germany’s biggest classifieds website
42mHow mobile.de brings Data Science to Production...
58mReal Data Science Examples at OLX's Online Marketplace
