LLM-Powered Virtual Population for Demand Simulation and Pricing

2026-06-15 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

An LLM-powered virtual population model has been developed to simulate demand for pricing decisions, particularly for products described by rich unstructured information like text and images. This model addresses the need for not only mean-demand predictions but also uncertainty estimates for counterfactual prices. It operates by representing exposed customers as draws from a finite mixture of customer personas. For each persona, product, and candidate price, a Large Language Model elicits a persona-level purchase probability using both structured persona information and unstructured product details. These probabilities are then aggregated through calibrated mixture weights to form a predictive distribution of aggregate demand. The resulting simulator can evaluate counterfactual prices under various pricing objectives, including expected revenue and risk-aware criteria such as conditional value at risk. Tested on an online H&M fashion dataset, the calibrated LLM-based simulator achieved the best overall predictive performance among models considered and supports sample-efficient pricing decisions. This framework offers a practical way to use LLMs as demand simulators for products with limited historical demand data but rich product information, providing a full predictive demand distribution.

Key takeaway

For product managers or data scientists tasked with pricing new products or optimizing existing ones, this LLM-powered demand simulator offers a robust tool, especially when dealing with rich product descriptions but limited historical sales data. You should consider integrating this approach to move beyond simple point forecasts, enabling you to quantify demand uncertainty and evaluate counterfactual prices against both expected revenue and risk-aware objectives like Conditional Value at Risk. This allows for more informed, resilient pricing strategies.

Key insights

LLMs can simulate complex demand dynamics for pricing by modeling virtual customer personas and product interactions.

Principles

• Incorporate unstructured data for richer demand modeling.
• Model demand uncertainty, not just point forecasts.
• Use persona-based probabilities for granular predictions.

Method

An LLM elicits persona-level purchase probabilities from structured persona data and unstructured product information, which are then aggregated via calibrated mixture weights to predict aggregate demand distribution.

In practice

• Evaluate counterfactual prices with demand uncertainty.
• Optimize pricing for risk-aware objectives like CVaR.
• Simulate demand for new products lacking historical data.

Topics

• Large Language Models
• Demand Simulation
• Pricing Optimization
• Virtual Population
• Unstructured Data
• Counterfactual Analysis

Best for: AI Engineer, NLP Engineer, AI Scientist, Machine Learning Engineer, Data Scientist, Director of AI/ML

Related on AIssential

• The Shadow Price of Reasoning: Economic Perspective on Optimal Budget Allocation for LLMs — Optimal LLM inference budget allocation can be modeled economically using a global shadow price to maximize accuracy under resource scarcity.
• Poisson-MNL Bandit: Nearly Optimal Dynamic Joint Assortment and Pricing with Decision-Dependent Customer Arrivals — The Poisson-MNL model optimizes dynamic assortment and pricing by accounting for decision-dependent customer arrival rates.
• A Few-Shot LLM Framework for Extreme Day Classification in Electricity Markets — LLMs can effectively classify electricity price spikes with limited data by interpreting natural language prompts of system conditions.
• What Can You Learn from a Deterministically Chaotic Market Simulation? — Agent-based market simulations can reveal emergent financial phenomena like the co-existence of value and momentum strategies.
• Distribution Shift Alignment Helps LLMs Simulate Survey Response Distributions — Aligning LLMs with distribution shifts across backgrounds enables more accurate and data-efficient survey response simulation.
• I Built a Demand Forecasting Model at a Real FMCG Company. It Failed First. (Part 1) — Segmenting diverse demand patterns before forecasting significantly improves model accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.