Published on 10 Nov 2026

Can we trust digital twins? | YeeMei Chan

In Episode 3 of the NCMT podcast hosted by the Nanyang Centre for Marketing Technology at Nanyang Business School, YeeMei Chan, Co-Managing Director of the Brand Growth Lab at Kantar in Singapore, unpacks the rapidly evolving landscape of market research. She explores how predictive modelling, synthetic data, and AI-moderated qualitative studies are shifting research to a forward-looking strategy engine.


The Timeless Core of Market Research

Marketing is fundamentally probabilistic, not deterministic. Whether a brand is launching a new bubble tea flavour or a multimillion-dollar ad campaign, the core purpose of market research remains unchanged: increasing the odds of commercial success.

While AI has radically transformed how we gather data, YeeMei Chan insists that the why of research remains firmly rooted in human strategy. Success still hinges on three foundational pillars: knowing what questions to ask, knowing whom to ask, and effectively framing the problem to be solved.

“AI sometimes makes the data-gathering part so easy that people forget some of the fundamental principles,” YeeMei observes. “But ultimately, what decisions you are trying to make with the result really matters. How you frame the problem to be solved is still unchanged, whether you have AI or not.”

From Predictive Modelling to Surprising Insights

Historically, market research looked in the rear-view mirror, tracking what consumers did yesterday or what they think today. Today, AI enables predictive modelling. For instance, tools like Kantar’s LINK AI are trained on vast databases of human ad-testing history. By scanning an untested video file, the AI can predict its persuasion and enjoyment scores before it ever reaches a real audience. Going a step further, it can also advise how to optimise the ad for better performance.

Beyond prediction, AI possesses a unique ability to collide disparate data sources (like social media, search trends, and survey data) to surface insights researchers never even thought to ask for.

YeeMei shares a recent example: if a premium shampoo brand sees sales dropping, the human hypothesis usually jumps to economic downturns or fierce competition. However, AI cross-referenced cultural trends to suggest a completely different hypothesis. “Nowadays, a minimalist beauty routine is in play. And what about people working more from home? Could that mean they shower less than before?” Because AI can synthesise broader environmental factors, it challenges ingrained hypotheses and opens new avenues for innovation.

Navigating the Nuances of Synthetic Data

One of the most polarising topics in modern marketing is “synthetic data”, artificially generated data designed to replace or augment human respondents. Rather than a blanket acceptance or rejection, YeeMei breaks synthetic data into three distinct, risk-adjusted use cases:

  • Synthetic Boosting: Used to augment hard-to-reach demographics. If a survey captures only 30 young males living in specific public housing, AI can study their data and historically similar cohorts to synthetically generate 30 more “lookalike” respondents. This creates a statistically significant subgroup without the prohibitive cost of niche recruitment.

  • Digital Twins: Kantar maintains a panel of 7.2 million “Digital Twins” globally. These are not random AI creations; they are built upon the longitudinal, historical data of actual human panellists. Because the AI is grounded in an individual’s real past behaviours and income shifts, brands can confidently use these twins to test new product concepts predictively.

  • Synthetic Personas: These could take the form of in-house sandbox GPTs for knowledge management such as using LLMs to prompt and query what a company already knows from existing research, e.g. what unmet needs an Indonesian woman has when it comes to shampoo. Or, it could take the form of persona bots built around a generalised understanding of a particular customer profile. While excellent for early-stage brainstorming or hypothesis generation, YeeMei warns against relying on them for final validation. “Would I trust what she said before I launched my million-dollar ad campaign? Absolutely not. You need to understand what the AI is built on and whether you are risking your immediate decision on generalised information.”

The Unexpected Rise of the AI Moderator

Qualitative research, traditionally reliant on the deep empathy and probing skills of a human moderator, is also experiencing an AI renaissance.

Kantar is increasingly deploying AI-moderated chatbots to conduct qualitative interviews. Surprisingly, rather than feeling alienated by a machine, consumers often feel more liberated. Because respondents can chat with the bot via their phones at night in the privacy of their own bedrooms, they are highly engaged.

“This girl was so cute to show us what she's actually using,” YeeMei recounts of a recent skincare study. “She's doing her night routine while having the conversation with us. I actually feel that people are getting much more comfortable talking to a camera, and maybe because of the environment, they're a lot more comfortable sharing a lot of details.”

The Enduring Value of Business Acumen

As AI democratises data collection and analysis, the technical mechanics of research are becoming baseline hygiene. The true competitive advantage for human researchers now lies in business acumen and strategic storytelling.

A researcher’s value is measured by their ability to drive action. If an insight sits unread on a shared drive, it has failed. Researchers must use tools and their own strategic empathy to frame the exact same data point differently for the CFO focused on margins, the R&D lead focused on packaging, and the CMO focused on channels.

Ultimately, AI can gather the data and simulate the consumer, but it is the human researcher who must translate that data into the language of the boardroom to drive tangible business growth.