Artificial Intelligence1 min read

Build a real-time next-best-offer recommender using Databricks Feature Views

By · Published by Everything Blog

In short

The open-source architecture combines stable customer attributes like credit balances with live event streams from Kafka, enabling immediate reactions to active user behavior. A LightGBM ranking model registered in Unity Catalog powers the pipeline to generate relevant product suggestions mid-session. The entire system is packaged as runnable Databricks notebooks designed to eliminate the engineering gaps typical between model development and real-time inference.

Key points

  • Declaring features once for both offline training and online serving prevents train/serve…: Declaring features once for both offline training and online serving prevents train/serve skew and costly data leakage.
  • Combining long-term profile data with immediate clickstream intent allows banks to surfac…: Combining long-term profile data with immediate clickstream intent allows banks to surface financial offers before sessions conclude.
  • Managing the LightGBM ranker via Unity Catalog streamlines the retrieval of batch and str…: Managing the LightGBM ranker via Unity Catalog streamlines the retrieval of batch and streaming features during inference.

Sources checked by our analysis pipeline: 1

Author a Feature once, serve it everywhere — one declarative definition powers point-in-time training and warm online ranking at p95 34 ms, with in-session signals live in under 130 ms.

by Tian Tan, Sixuan He, Nick Joung and Julia Powell

Real-time personalization depends on understanding what customers want in the moment. When a retail banking customer checks their accounts, explores credit cards, or opens a loan calculator, each interaction signals their current intent. A relevant offer needs to reach them while they're still in the session.

The challenge is making that intent available to the model. Raw clickstream data must be aggregated into time-windowed features for two purposes: training on historical values as they existed at a given point in time, and serving predictions using current values with millisecond latency. Building these pipelines separately creates duplicate feature definitions that can drift over time and degrade model accuracy.

Databricks Feature Views solve this by letting you define features once for both training and serving. In this blog, we'll walk through an open-source Solution Accelerator that uses Feature Views to build a…

Artificial Intelligence