In the rapidly evolving landscape of market research, synthetic research powered by Large Language Models (LLMs) is emerging as a transformative force. These advanced AI models, such as OpenAI’s GPT-4o, are not only revolutionizing how brands gather insights – making standard research incredibly efficient – but also opening up previously inaccessible areas through traditional methods. Perhaps most intriguingly, they’re producing insights designed not just for human understanding, but for AI consumption itself.
This article explores how LLMs are revolutionizing market research across three key areas: enhancing existing methodologies, unlocking new consumer insights, and pioneering frontier analytics beyond human cognition.
- Enhancing Traditional Research Methods with Efficiency
Traditional market research methods, while proven effective, often involve time-consuming processes and substantial costs. LLMs offer an opportunity to streamline these practices by automating data collection and analysis. Consider sentiment analysis – a crucial component in understanding consumer perceptions. LLMs can now perform this task more swiftly and accurately, processing and interpreting consumer opinions from social media, reviews, and forums in real time.
This represents a natural evolution for sentiment analysis. With their ability to process the entire web, these models can predict unexpected research outcomes with remarkable accuracy. The marketing community has shown particular trust in this capability, as LLMs like GPT-4o and Llama 2 have demonstrated impressive accuracy rates of up to 93% in binary tasks, making sentiment analysis one of synthetic research’s greatest strengths (Source: “Sentiment Analysis in the Age of Generative AI” (2024), by Jan Ole Krugmann and Jochen Hartmann from Technical University of Munich (TUM)).
PHD has identified the potential role of LLMs in predicting market metrics without direct access to proprietary data. Internal studies comparing GPT-4o’s predictions to actual market research data showed an average accuracy exceeding 70% across metrics like unaided brand awareness, consideration, and purchase intent.
While this efficiency gain is exciting, what’s truly compelling is how LLMs allow us to peer into areas where standard research typically struggles.
- Unlocking Consumer Insights Beyond Conventional Research
One of LLMs’ most compelling advantages lies in their ability to uncover consumer insights that traditional methods struggle to capture. While conventional surveys and focus groups often fail to identify unconscious consumer behaviors accurately and need states, LLMs excel at analyzing language patterns and vast datasets to infer these nuanced insights.
For instance, they can estimate the major need states driving purchases in categories like soft drinks – whether it’s thirst, social occasions, or health considerations – and assess the proportion of purchases driven by each state. This depth of analysis has traditionally been challenging due to the limitations of self-reported data and human psychology’s complexities.
While powerful, LLMs shouldn’t be relied upon in isolation. A multi-agent approach, where multiple AI systems interact to simulate varied perspectives, provides more reliable predictions and reduces potential biases. This exciting new area of synthetic research is already being deployed within Omnicom’s marketing technology platform to identify consumer entry points (CEPs) and distinctive brand assets with quantitative measures.
- Pioneering Frontier Analytics for Algorithmic Insights
Perhaps the most valuable aspect of synthetic research lies in generating insights not meant for human minds but for algorithms. The high-dimensional structure of LLMs enables them to detect patterns within unstructured data that might elude human cognition.
These intelligence patterns can serve as bid factors in custom bid optimization, making probabilistic adjustments within real-time decision-making. Annalect, Omnicom’s data and analytics division, leads the integration of LLM-driven insights into their analytics platforms. Adding complex insights from synthetic data provides the competitive edge needed as more brands build their own algorithms for custom optimization and next-best-action decisions.
This also lays the groundwork for creating digital twins of brand marketplaces. These simulations, built from synthetic data, can model and predict entire marketplace behaviors in a virtual world, testing countless strategic approaches across all marketing levers.
By 2025, we’ll likely see increased adoption of agent models with action capabilities for online platforms, with synthetic research providing the closed loop required for agent models to self-optimize.
The Future of Market Research
As we look ahead, synthetic research powered by LLMs represents more than just an evolution in market research—it’s a fundamental shift in how we understand and respond to market dynamics. From enhancing traditional methodologies with unprecedented efficiency to uncovering hidden consumer insights and pioneering algorithmic analytics, LLMs are reshaping the landscape of market intelligence. While traditional research methods will continue to play a vital role, the integration of synthetic research capabilities promises to deliver deeper insights, faster responses, and more accurate predictions than ever before. As these technologies mature, organizations that successfully harness the power of LLMs in their research processes will find themselves better equipped to navigate the increasingly complex world of consumer behavior and market trends.
Authored by
Mark Holden
Chief Strategy Officer, PHD Worldwide
