Idea
Model simulating realistic consumer purchase intent for scalable, cost-effective market research.
Research Paper
Core Innovation
This paper presents semantic similarity rating (SSR), which elicits textual responses from LLMs and maps them to Likert scale ratings using embedding similarity to reference statements. This method overcomes unrealistic numerical rating distributions from direct LLM queries and achieves high reliability and realistic response patterns.
Why It Matters
Consumer research is costly and limited by panel biases and scale constraints. This approach reduces expenses and expands scalability by generating synthetic yet realistic consumer responses, preserving traditional survey metrics and interpretability. It transforms market research workflows by enabling rapid, large-scale consumer insights without extensive human panels.
Market Size (TAM)
$10–20B TAM for global consumer research and market insights; $2–5B SAM from consumer goods and market research firms. Driven by demand for cost reduction and scalable consumer data.
Potential Customers & Pain Points
- Consumer goods companies – High cost and bias in consumer research
- Market research firms – Limited scalability and slow data collection
- Retailers – Need for rapid consumer feedback
- Advertising agencies – Require rich qualitative consumer insights.
Business Model
Subscription-based SaaS platform offering synthetic consumer research simulations with tiered pricing based on survey volume and customization; potential for API integration with existing market research tools.
Competitive Landscape
- Qualtrics
- NielsenIQ
- SurveyMonkey
- Toluna
Implementation Challenges
- Acceptance of synthetic consumer data by traditional market research stakeholders
- Ensuring model generalizability across diverse product categories
- Regulatory and privacy concerns around synthetic data usage
Validation Strategy
- Pilot deployments with consumer goods companies to compare SSR outputs against traditional survey results
- Longitudinal studies measuring predictive accuracy of synthetic responses on actual purchase behavior
- User feedback collection from market researchers on interpretability and usability
Research Paper Overview
LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings
Summary
This paper introduces semantic similarity rating (SSR), a method that uses large language models to simulate consumer survey responses by mapping textual answers to Likert scale ratings through embedding similarity. Tested on 57 personal care product surveys with 9,300 human responses, SSR achieves 90% of human test-retest reliability and realistic response distributions, enabling scalable and interpretable consumer research simulations.