Real-time exact audience sizing with conversational AI for marketers
REALMS: An AI-Assistant Conversational System for Real-Time Exact Audience Sizing over High-Dimensional Nested Profiles
Computation and LanguageInformation Retrieval
Summary
Digital marketers need to know exactly how many people fit certain detailed profiles to plan campaigns well, but this can take a long time and be inaccurate with traditional methods. The authors created REALMS, a tool that lets marketers ask questions in everyday language and get precise counts almost instantly, even when dealing with millions of profiles and many attributes. It uses smart search to find relevant details, turns natural language into exact database queries, and works across different industries. Their tests show it finds the right information quickly and correctly, enabling real-time decisions.
What this means in practice
- •For digital marketing teams: Provide marketers with instant, accurate audience counts from large customer profiles using natural language queries.
- •For enterprise data platform engineers: Integrate conversational query capabilities to enable real-time exact analytics over complex nested customer data schemas.
Authors
Haixu Ma, Aditya Bansal, Shubham Lohiya, Sumit Ranjan
Abstract
Audience sizing is a critical component of digital marketing. It enables precise resource allocation, campaign planning, and performance optimization. Traditional approaches using skeleton audiences, sampling, or predictive modeling suffer from significant delays, estimation errors, and poor scalability over high-dimensional profile data. We present REALMS (Real-time Exact Audience sizing via LLM-based Multi-attribute Search), a conversational system for exact audience sizing deployed in production on an enterprise customer data platform. REALMS enables marketers to query massive profile stores with millions of profiles and thousands of attributes using natural language and receive precise counts in seconds. The system introduces three key components: (1) a categorical attribute retrieval mechanism using embedding-based vector search to dynamically identify relevant schema attributes without manual configuration; (2) an LLM-powered NL2SQL pipeline with template-based in-context learning for accurate query generation over complex nested schemas; and (3) schema standardization enabling industry-agnostic deployment across diverse enterprise environments. Evaluation on real enterprise data demonstrates strong recall for attribute retrieval, high SQL execution accuracy, and low latency, which enables real-time interactive audience insights where prior methods required hours.