Achieving a deeper understanding of customers enables businesses to create the right products, experiences, and personalization. The existing approach to understanding customers across industries is dependent heavily on using explicit feature representations of specific behaviors and attitudes of interest. However, with more customer data being generated and collected than ever before through multiple customer engagement modes (digital, call centers, sales transactions), traditional approaches to characterizing customers are becoming a significant bottleneck. This makes learning latent representation an attractive proposition for understanding customers based on their expressed behaviors.
In this talk, ZS will showcase:
How to use graph neural networks to learn rich, latent representations for customers (customer embeddings) to encode interactions and behaviors.
Graph neural network’s superior performance on a variety of customer-level ML tasks such as prediction of brand adoption (pre-launch, early launch), channel preferences, and engagement with digital channels (e.g. email) compared to existing approaches.
Abstract & Bio
Customer2Graph: Powering Customer Analytics with Graph Representations