Existing methods for displaying customer-oriented recommendations do not always reveal the causal relationships between customers and the products they purchased, as most of the times they focus on the customer profile space or the product profile space. As a result, customers especially those with little knowledge of these insurance products, could hardly get a clear vision of those recommendations, if provided, nor could they discover the reasons and mechanisms behind those recommendations.
UX Solutions
We propose an interpretable visualization on insurance product recommendations, in collaboration with datebao.com, one of the leading online insurance platforms in China. We develop user-friendly visualizations in 3 dimensions: user profile, insurance profile, and the link space based on data collected in a 3-year timespan. By providing an enriched sankey diagram, new customers could easily browse the links between clustered customer profiles, product categories, as well as the causal relationship of those two spaces, which could largely facilitate them deciding which one to buy based on their own personalized needs.
We also provide an interactive visualization, using the metaphor of a planetary system, which extends the 2D sankey diagram. The extended 3D system is particularly suited for customers interested in detailed information of the mechanisms behind those recommendations, meanwhile providing a possible solution for visualizations of a similar platform, featuring animated visual outputs.
My Role
Wrote the paper as 1st author.
Implemented the visualization system with Unity, HTML and Javascript.
Contributed in extending the confounder balancing algorithm of causal inference with Python.