Market research is the backbone of customer-driven decision-making, yet gathering reliable insights has never been more challenging. Recruiting and managing a representative sample takes up 60% of a research project’s time, but despite these efforts, response rates continue to decline, panel fatigue is growing, and operational costs are rising. At the same time, evolving privacy regulations like GDPR (EU’s General Data Protection Regulation) and CCPA (California Consumer Privacy Act) are making it increasingly difficult to collect, store, and utilize consumer data, further restricting access to high-quality insights. Given these obstacles, the question arises: Is there a more efficient way to conduct research? Synthetic Panels – AI-generated panels that mirror real-world behaviours and demographics – could be the answer. But can it truly replace or supplement traditional survey methods? Let’s find out.
Table of Contents
What Are Synthetic Panels?
Synthetic panels are AI-generated groups of virtual respondents designed to simulate the behaviors, preferences, and demographics of real-world consumer segments. Traditional survey panels require time-intensive recruitment, screening, and maintenance. Meanwhile, synthetic panels fastens the process by leveraging large-scale data and machine learning to create digital personas capable of answering survey questions just like humans would.
These panels are not based on fabricated or random data. Instead, they are constructed using deep contextual learning from real-world datasets. This includes historical survey responses, customer reviews, behavioral data, and public opinion trends. The result is a scalable, responsive, and privacy-compliant solution that mirrors actual market segments with surprising accuracy.
In essence, synthetic panels offer researchers a powerful way to simulate and test consumer reactions without needing to directly involve human participants for every new question or product concept. They can even model reactions to hypothetical situations or future product launches, enabling predictive insights that aren’t possible with static historical data alone.
How Do Synthetic Panels Work?
The creation and operation of synthetic panels involve several key steps that combine data engineering, machine learning, and behavioral modeling:
- Data Ingestion and Preprocessing: Synthetic panels begin with the ingestion of diverse datasets. These range from previous survey results and customer support logs to online reviews and demographic insights. These inputs provide the foundational understanding of how different customer segments think and behave.
- Persona Modeling: AI models then use this data to generate synthetic personas. Each persona is built to represent a specific consumer archetype (e.g., tech-savvy Gen Z, cost-conscious retirees, luxury-seeking professionals). These personas are not just demographic shells; they include psychological traits, brand preferences, and behavioral tendencies.
- Response Simulation: When survey questions or research scenarios are presented, the synthetic personas “respond” based on their trained profiles. The AI predicts how each persona would likely answer, using patterns learned from historical data and contextual cues.
- Contextual Calibration: To ensure accuracy, synthetic panels are continuously refined using real-world feedback. This involves comparing synthetic responses with those from actual respondents and adjusting models to reduce any biases or inaccuracies.
- Output Analysis and Insights: Once responses are collected, researchers can analyze trends, segment behaviors, and test hypotheses much like they would with traditional panels – but with greater speed, scale, and flexibility.
Can AI Generate Reliable Survey Panels: A Merkle Case Study
A major Southeast Asian airline, in collaboration with Merkle (a dentsu company), explored this question by implementing an AI-driven approach to generate synthetic survey responses and compare them to real customer feedback.
Phase 1: The Problem with Generic AI Responses
When AI was first asked the survey questions without contextual data, including real human data examples, the responses followed a predictable pattern:
- The Net Promoter Score (NPS) followed a bell curve but skewed toward neutral ratings.
- AI-generated brand perception data did not match real customer sentiment, making it unreliable.
These limitations highlighted a key challenge: without real-world context, synthetic panels lack the nuances of human opinion. This meant that using AI alone wasn’t enough—additional refinement was necessary.
Phase 2: Enhancing AI Responses with Context
To improve the accuracy of synthetic responses, Merkle introduced a crucial element: historical survey responses, customer complaints, and feedback trends. By feeding these contextual inputs, the system began to recognize and replicate real human sentiment more effectively.
- Brand perception scores became more aligned with actual customer opinions.
- Customer segmentation patterns improved, mirroring real-world preferences (e.g., students favouring budget airlines, business travellers prioritizing convenience).
- NPS distributions became more realistic, maintaining a bell curve but better reflecting real human responses.
This marked a turning point – AI-generated responses began to closely resemble actual customer data, demonstrating the potential of synthetic panels in market research.
Why Synthetic Panels Matter?
This case study underscores why synthetic panels are becoming a game-changer in market research. Key benefits include:
- Expanding Sample Sizes: AI-generated responses supplement real data, allowing for richer, more representative insights.
- Predicting Customer Behaviour: AI can estimate how people might respond to new questions without conducting fresh surveys.
- Reducing Research Costs: Recruiting human respondents takes weeks, while synthetic panels are available instantly.
- Reaching Niche Audiences: Hard-to-reach customer groups (e.g., luxury travelers, C-level executives) can be modelled more effectively.
- Overcoming Privacy Barriers: Since it doesn’t rely on actual customer identities, synthetic panels comply with strict privacy regulations.
- Always-On Availability: Unlike human respondents, synthetic personas are available indefinitely to answer or test any further questions.
A Balanced Approach: Human + Synthetic Panels
While synthetic panels show immense promise, a fully AI-driven approach is still evolving. Organizations are currently exploring how much synthetic data can be mixed with real responses—starting with as little as 1% synthetic input and increasing over time.
The best way forward is a blended approach, where companies use both human and synthetic audiences. This allows researchers to compare and optimize synthetic responses based on real human feedback, ensuring reliability and accuracy.
Key Considerations When Using Synthetic Panels
Here are some points to watch out for when using synthetic panels for market research.
- “No Training Required” Claims: Synthetic panels need careful tuning with real-world context to be useful.
- Lack of Validation with Human Samples: Always test synthetic responses against human data to ensure accuracy.
- Over-Reliance on AI: Synthetic Panels should supplement, not replace, real insights—at least for now.
Conclusion
This solution by Merkle proves that synthetic panels, when refined with contextual inputs, can bridge the gap between traditional market research and AI-driven insights. While 100% synthetic adoption will take time, a hybrid approach is the best way forward for now, allowing businesses to enhance decision-making, supplement real responses, and unlock new market opportunities.
At the same time, ensuring transparency and ethical use of AI-generated insights will be crucial in maintaining trust and accuracy in market research. The future isn’t just about gathering data – it’s about generating it intelligently.
This blog is a result of the collective contributions of the following members of the Merkle team: Vinay Mony (Vice President, CXM – Insights & Analytics), Mario Thirituvaraj (Assistant Vice President, CXM – Insights & Analytics), Debasree Bhattacharya (Assistant Vice President, CXM – Insights & Analytics), Rohit Mudukanagoudra (Analyst), Mahima Salian (Analyst), Bharat Shetty (Senior Manager, CXM – Insights & Analytics), and Aneesh Kammath (Head – XM Advisory APAC)
Frequently Asked Questions
Q1. What is the difference between synthetic panels and traditional survey panels?
A. Traditional panels involve recruiting real people to answer survey questions, which can be time-consuming and costly. Synthetic panels, on the other hand, are AI-generated digital personas that simulate real consumer behavior and can respond instantly to surveys based on learned patterns and contextual data.
Q2. How do synthetic panels comply with data privacy regulations like GDPR and CCPA?
A. Synthetic panels do not rely on identifiable customer data. Instead, they generate personas based on aggregated, anonymized insights, making them inherently privacy-compliant and suitable for research under strict data protection laws.
Q3. Can synthetic panels completely replace human respondents?
A. Not yet. While they are a powerful supplement, synthetic panels are currently best used alongside human feedback. A hybrid approach ensures validity, accuracy, and trust in insights—especially when exploring new markets or unfamiliar product categories.
Q4. How long does it take to generate responses using synthetic panels?
A. Unlike traditional research, which can take weeks, synthetic panels can generate responses almost instantly. This makes them perfect for agile research teams needing quick turnarounds.
Q5. How can companies validate the accuracy of synthetic panel responses?
A. Validation typically involves benchmarking synthetic responses against real survey data. Regular calibration, A/B testing, and the inclusion of known data points help ensure the AI remains aligned with actual consumer sentiment.
Q6. What are the risks of over-relying on synthetic panels?
A. Over-reliance can lead to blind spots if synthetic panels are not regularly updated or validated with real data. Additionally, if used without context, the AI may produce generic or biased responses. Human oversight is essential for ethical and effective implementation.
Merkle, a dentsu company, powers the experience economy. For more than 35 years, the company has put people at the heart of its approach to digital business transformation. As the only integrated experience consultancy in the world with a heritage in data science and business performance, Merkle delivers holistic, end-to-end experiences that drive growth, engagement, and loyalty. Merkle’s expertise has earned recognition as a “Leader” by top industry analyst firms, in categories such as digital transformation and commerce, experience design, engineering and technology integration, digital marketing, data science, CRM and loyalty, and customer data management. With more than 16,000 employees, Merkle operates in 30+ countries throughout the Americas, EMEA, and APAC. For more information, visit www.merkle.com
