# Customer Lifetime Value Prediction Using Embeddings

**Authors:** Benjamin Paul Chamberlain, Angelo Cardoso, C. H. Bryan Liu, Roberto Pagliari, Marc Peter Deisenroth  
**arXiv:** [1703.02596](https://arxiv.org/abs/1703.02596)  
**Date:** 2017-03-07  
**Summary page:** https://www.summarizepaper.com/en/paper/1703.02596v3/

## Key points

- Customer Lifetime Value (CLTV) prediction system at ASOS.com is crucial for estimating future customer value accurately
- System allows retailers to allocate marketing spend effectively and identify high-value customers while minimizing losses
- Provides daily estimates of each customer's future value and enhances personalized shopping experience
- Recent advancements in various domains have shown progress by using automatically learned features from data instead of handcrafted features
- Study titled "Customer Lifetime Value Prediction Using Embeddings" details the innovative system at ASOS that leverages feature representations learned from data
- Authors propose a novel approach to generate embeddings of customers to address challenges posed by ever-changing product catalog
- By utilizing embeddings instead of handcrafted features, the system achieves substantial improvement in predicting customer lifetime value

## Summary

The Customer Lifetime Value (CLTV) prediction system deployed at ASOS.com, a global online fashion retailer, is crucial for accurately estimating the future value of customers. This estimation allows retailers to effectively allocate marketing spend and identify high-value customers while minimizing losses. The system at ASOS provides daily estimates of each customer's future value and plays a pivotal role in delivering a personalized shopping experience. Traditionally, CLTV prediction has relied on large numbers of handcrafted features and ensemble regressors to forecast value, predict churn, and evaluate customer loyalty. However, recent advancements in various domains such as language, vision, and speech have demonstrated significant progress by replacing handcrafted features with automatically learned features from data. In this study titled "Customer Lifetime Value Prediction Using Embeddings," authors Benjamin Paul Chamberlain, Angelo Cardoso, C. H. Bryan Liu, Roberto Pagliari, and Marc Peter Deisenroth detail the innovative system deployed at ASOS that leverages feature representations learned from data to enhance CLTV modeling. The authors propose a novel approach to generate embeddings of customers that address the challenge posed by the ever-changing product catalog. By utilizing these embeddings instead of an exhaustive set of handcrafted features, the system achieves a substantial improvement in predicting customer lifetime value. This research was presented at the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining in 2017 and offers valuable insights into enhancing CLTV prediction through advanced feature learning techniques.