Pharma Insights Leaders Weigh In: The Future of Synthetic Data in Market Research

A Glimpse into the Future of Pharma Market Research

In late July, Day One Strategy hosted a webinar that brought together over 100 pharmaceutical insights leads to discuss a topic at the cutting edge of market research: synthetic data and respondents. This article summarizes the key takeaways from this engaging discussion, offering a window into how industry leaders are thinking about the future of data-driven insights in pharma.

The Current Landscape: Challenges Voiced by Pharma Insights Leaders

The webinar opened with a frank discussion of the challenges facing insights teams across the industry. Several key themes emerged:

1. Doing More with Less: "We're facing budget cuts of 10-15% across the board," shared a senior insights manager from a top-10 pharma company. "It's pushing us to be more innovative in how we gather and use data."

2. Maximizing Existing Data: A director of global insights noted, "There's a wealth of information in our historical data that we haven't fully tapped. We need to find ways to extract more value from what we already have."

3. Seeking Clarity from AI: "We're intrigued by AI, but we need it to provide clear, actionable intelligence," emphasized a market research lead. "We can't afford to add more confusion to the mix."

Synthetic Data and AI: A New Frontier

Imagine a focus group that never sleeps, a therapy expert available 24/7, or a strategist capable of processing millions of data points in seconds. 

The future of generative AI is specialized AI agents:

1. AI Market Research Moderator: Trained by expert moderators, this agent can conduct and analyze focus groups and interviews with the finesse of a seasoned professional.

2. AI Therapy Expert: Armed with verified knowledge on specific therapeutic indications, this agent provides deep, specialized insights on demand.

3. AI Strategist: A master synthesizer, capable of connecting dots across vast datasets to generate strategic insights that might elude even the most experienced human analysts.

"The idea of having 24/7 access to 'synthetic respondents' is appealing," said one insights manager. "But how do we ensure the insights are truly reflective of our real-world stakeholders?"

This question led to a nuanced discussion about the strengths and limitations of different AI approaches.

Large Language Models (LLMs): Promising but Risky

Participants acknowledged the buzz around models like ChatGPT and Claude but highlighted concerns:

1. Data Hallucinations: The generation of convincing but factually incorrect information.

2. Lack of Data Ownership: Insights generated are not proprietary, raising concerns about competitive advantage.

3. Generalist Knowledge: A breadth of information that often lacks the depth required for specialized pharma applications.

Small Language Models (sLMs): A Targeted Approach for creating synthetic respondents for specific use cases.

Many attendees showed interest in the potential of smaller, more focused AI models:

"If we could train these models on our own data, ensuring accuracy and maintaining ownership, that could be a game-changer," enthused a digital innovation lead.

Opportunities in Specific Areas:

Rare Disease Research

The potential of synthetic data in rare disease research generated significant excitement:

"With such small patient populations, gathering sufficient data is always a challenge," explained a rare disease research specialist. "Synthetic data could help us fill crucial gaps in our understanding."

Multi-Modal Engagement

Looking to the future, many participants were intrigued by the possibility of more interactive AI engagements:

"Imagine being able to show an images of a new campaign, or speak to synthetic respondents like we're doing a live interview," mused one forward-thinking insights director. "That could open up entirely new avenues for insight generation."

Balancing Optimism with Pragmatism

While the overall tone of the discussion was optimistic, there was a clear desire for practical, actionable next steps:

"We need to start small, perhaps with pilot projects in non-critical areas," suggested a senior VP of market research. "Let's prove the concept before we scale."

For brands which have been around for years to those preparing to launch, data quantity differs considerably however what matters most is quality of data, not volume. “Insight data in most organizations tends to be well organized by therapy and brand” proclaimed an AI engineer, so that helps in the data preparation phase of using data for a LLM or sLM. 

The data audit is the first step and that will determine if there are any major gaps which primary market research will need to fill.  All formats can be used – PDFs, PPTs, Excel, SPSS, Word etc. and this is improving considerably at new powerful models launch into the market.  The latest being Meta’s Llama 3 405bn parameter model.

Others emphasized the importance of maintaining ethical standards: "As we explore these new technologies, we must ensure we're upholding the highest standards of data privacy and research ethics," reminded a compliance officer.

A Future Shaped by Collaboration

As the webinar concluded, there was a palpable sense of excitement about the potential of synthetic data and AI in pharma market research. However, this was tempered by a recognition of the challenges ahead.

"This isn't a journey any one company can make alone," summarized the webinar host. "It will take collaboration across the industry – insights teams, data scientists, ethicists, and regulators – to realize the full potential of these technologies."

The discussion made clear that while synthetic data and AI respondents are not yet mainstream in pharma market research, they are firmly on the radar of industry leaders. As one participant put it, "The future of insights is coming into focus, and it's looking increasingly synthetic."