Challenge
A leading car manufacturer wanted to increase sales of digital services directly in the vehicle. The aim was to personalize recommendations based on individual driving behavior and customer preferences. The project was particularly complex, as over 65 billion signals from global sources had to be processed every day. This data needed to be efficiently aggregated and analyzed in order to gain relevant insights and make precise recommendations while maintaining high data protection standards.
Approach
We developed a state-of-the-art recommendation system that enables efficient and effective recommendations. The data was ingested via an event hub and processed in parallel using Spark Clusters, orchestrated by Azure Data Factory (ADF). Parallelized processing on multiple GPUs allowed us to train the model quickly and scalably. The training dataset comprised 37 million entries with almost 1000 features.
Results
The implemented recommendation system achieved impressive results. The car manufacturer was able to generate additional sales and increase the conversion rate by up to 70%. This significant improvement in the conversion rate demonstrates the effectiveness of the personalized recommendation system and our team’s ability to process and use complex data efficiently. The success of the project not only increased sales figures, but also customer satisfaction through relevant and useful in-car service recommendations.