SalesLoop Boosts Lead Ranking with Reinforcement Learning

Traditional lead ranking models often fail to translate high offline accuracy into strong production performance due to issues like mismatched metrics, objective misalignment, and data drift. A new framework, SalesLoop, aims to bridge this gap by employing reinforcement learning. SalesLoop establishes a continuous feedback loop, directly linking model predictions to actual business outcomes. It introduces a novel performance-aware reward system that considers conversion results, ranking position, and conversion speed. Additionally, it utilizes Discriminative GRPO, a listwise optimization objective adapted for discriminative ranking models. Empirical results show SalesLoop significantly improves NDCG@K and P@K compared to static baselines. A 160-day production A/B test involving 16.5 million leads and 280 sales specialists at a New Energy Vehicle manufacturer demonstrated statistically significant cumulative lifts in conversion rates, proving its effectiveness in real-world scenarios.