HPRO: New LLM framework boosts sales lead scoring by 39.7%

Sales lead conversion in high-stakes domains like automotive and real estate is fundamentally different from e-commerce recommendations. Traditional methods – rule-based scorecards, machine learning models, or pointwise CTR predictions – struggle with sparse supervision, unstructured CRM logs, and the inability to capture relative lead priority. Large Language Models (LLMs) offer superior semantic understanding of customer interactions, but general-purpose LLMs are ill-suited for lead ranking because they generate text rather than comparable scores and lack alignment with the hierarchical priorities of sales funnels.

To address this, researchers propose HPRO (Hierarchical Preference Ranking Optimization), an LLM-based discriminative framework that jointly models structured CRM features and unstructured customer interactions. HPRO augments lead scoring with a hierarchical preference ranking objective, using a margin-aware Bradley-Terry formulation to transform sparse binary labels into dense, funnel-aware preference pairs. Experiments on large-scale data from a leading NEV brand demonstrate state-of-the-art classification (AUC 0.8161) and ranking performance (+39.7% precision among top-ranked leads). A 132-day online A/B test validated a 9.5% sales volume uplift, confirming real-world commercial impact.