Synthetic Data Under the Hood: Technologies, Models and Research Implications

As synthetic data becomes increasingly embedded in research workflows, researchers need a deeper understanding of the technologies that generate it.

This course moves beyond applications and use cases to explore how synthetic data is actually created. We'll examine the statistical, machine learning and generative AI techniques that underpin today's synthetic data solutions and consider how these approaches differ from—and build upon—many familiar modelling techniques already used within market research.

We'll explore two major branches of synthetic data technology. First, we'll examine machine-learning approaches used to augment, project and complete datasets. We'll discuss imputation, data augmentation, model-based estimation and other methods designed to extend existing information.

We'll then explore the rapidly expanding world of large language model (LLM)-based synthetic data. Participants will learn how personas, digital twins, synthetic respondents and synthetic panels are constructed, how they are calibrated and validated, and the assumptions that underpin their outputs.

Throughout the course we'll focus on the practical implications of these technologies for research quality, validity, transparency and decision-making. Rather than turning researchers into data scientists, the goal is to provide enough technical understanding to evaluate methodologies, interpret findings appropriately, and engage confidently with vendors, stakeholders and technical teams.

Learning outcomes / learning objectives

We'll make sure you'll come away being able to:

• Understand the major technical approaches used to generate synthetic data.
• Explain the differences between statistical, machine learning and generative AI methods.
• Understand how synthetic data augmentation and imputation techniques operate.
• Describe how LLM-based personas, digital twins and synthetic respondents are created.
• Evaluate strengths and weaknesses of different synthetic data generation methods.
• Understand key concepts relating to validation, calibration and model performance.
• Recognise how assumptions embedded within models can influence research outcomes.
• Assess methodological transparency and identify important questions to ask suppliers.
• Connect technical approaches to practical research applications and limitations.

Who will benefit?

This course is designed for researchers, insight professionals, research managers, consultants and innovation leaders who want a deeper understanding of synthetic data technologies and their implications for research practice.

The course assumes a basic understanding of synthetic data concepts but does not require advanced statistical or programming knowledge.

Learning method

The session will combine seminar-style teaching, visual explanations of technical concepts, methodological comparisons, industry case studies and interactive discussion. Complex concepts will be explained using research-focused examples rather than mathematical detail.

Course delivery mode

Online

Trainer Biography

Scott Worthge brings more than four decades of hands-on experience designing, managing, and delivering market and social research for complex, decision-critical projects across public- and private-sector organizations. His work spans consumer goods, technology, healthcare, financial services, government agencies, and nonprofit organizations, with engagements ranging from localized program evaluations to large-scale, multi-market domestic and international studies.

In parallel with his professional practice, Scott has taught market research and applied analytics at the university level for more than 20 years, including appointments with UC Berkeley Extension’s Marketing program and Michigan State University’s Master of Science in Market Research program. His teaching reflects the same applied orientation as his consulting work—bridging methodological rigor with the realities of implementation, procurement constraints, and stakeholder decision-making.

Scott is also a Board member of the Insights Career Network, a nonprofit organization dedicated to peer-to-peer support for research and insights professionals, where he contributes governance oversight and program guidance on a volunteer basis.

In 2025, Scott formed mktresearchguy llc to support project-based research engagements requiring senior-level design leadership, independent oversight, and practical execution. His consulting work includes advising on RFP-driven studies, leading primary research programs, and supporting organizations adopting emerging methodologies. Recent original research has focused on synthetic data modeling, voice and video data collection effectiveness, multicultural research design, and evolving definitions of representativeness in U.S. populations—areas increasingly relevant to public-sector, regulatory, and large-enterprise research initiatives.

A frequent speaker at professional conferences, webinars, and association forums, Scott regularly presents for organizations such as ESOMAR, IIEX, the Insights Association, and the Market Research Society. He has delivered subscribed webinars with record-setting attendance, reflecting sustained demand for practical guidance grounded in real-world research delivery rather than abstract theory.

Across all engagements, Scott’s work is characterized by a pragmatic, collaborative approach—focused on delivering research that is defensible, transparent, and actionable for organizations operating under formal procurement, compliance, and accountability requirements.