RADAR: Recall Augmentation through Deferred Asynchronous Retrieval

TL;DR

RADAR enhances large-scale recommender systems by asynchronously pre-ranking a larger candidate set offline, leading to improved recall and increased user engagement during online inference.

Contribution

RADAR introduces a novel asynchronous offline pre-ranking framework that improves candidate retrieval quality without increasing online latency.

Findings

2x recall improvement over baseline

+0.8% engagement lift in online tests

Effective combination of large candidate sets with powerful ranking models

Abstract

Modern large-scale recommender systems employ multi-stage ranking funnel (Retrieval, Pre-ranking, Ranking) to balance engagement and computational constraints (latency, CPU). However, the initial retrieval stage, often relying on efficient but less precise methods like K-Nearest Neighbors (KNN), struggles to effectively surface the most engaging items from billion-scale catalogs, particularly distinguishing highly relevant and engaging candidates from merely relevant ones. We introduce Recall Augmentation through Deferred Asynchronous Retrieval (RADAR), a novel framework that leverages asynchronous, offline computation to pre-rank a significantly larger candidate set for users using the full complexity ranking model. These top-ranked items are stored and utilized as a high-quality retrieval source during online inference, bypassing online retrieval and pre-ranking stages for these candidates. We demonstrate through offline experiments that RADAR significantly boosts recall (2X Recall@200 vs DNN retrieval baseline) by effectively combining a larger retrieved candidate set with a more powerful ranking model. Online A/B tests confirm a +0.8% lift in topline engagement metrics, validating RADAR as a practical and effective method to improve recommendation quality under strict online serving constraints.