Synthetic data has entered the industry conversation. AI-generated respondents, model-built datasets, and synthetic personas are no longer purely theoretical. Across the WIN network, the overarching theme is one of genuine curiosity, but not without hesitation. Members are exploring, testing, and in some cases cautiously applying AI-generated data, but no one has yet committed. They are keenly aware that the great power of synthetic data comes with great responsibility to the trust and rigour fundamental to the industry.
From Peru, Japan, the Netherlands, Sweden, Argentina, France, the Philippines, Vietnam, Canada, and Germany, WIN members share where their markets stand, and where they believe the technology is, and isn’t, ready to go.
Cautious curiosity but still experimenting
The most resounding view from the WIN members across the network is that while synthetic data is certainly on many clients’ radars, it has not (yet…) found its footing in live project work. Karin Nelsson (CEO, Demoskop, Sweden) echoes this from within her own firm, explaining that further testing is being done: “We’re actively running tests and experiments, though we have not yet applied synthetic data in live client projects.”
“Clients are still cautious… and continue to place greater trust in insights coming from real customers” says Urpi Torrado (CEO, Datum, Peru). In Argentina, Constanza Cilley (Executive Director, Voices!) echoes this, adding that hesitation is driven by a combination of concerns: “limited information, uncertainty about the methodology, and… data quality and representativeness.” Though it is currently “niche”, “some innovation-oriented clients are beginning to explore pilot projects and use cases.”
However, in the Philippines, Elaine Japitana (Senior Research Manager – Innovations Lead, Philippine Survey and Research Center) explains that interest has barely registered: “It is only mentioned a few times and does not proceed further.” Yet Olivier Guillon (Deputy General Manager, OpinionWay, France) offers a more optimistic counterpoint: despite limited actual usage, “there is a genuine interest from many clients.”
Complementary, not competing
Where synthetic data is finding traction, one principle consistently from nearly all members guides views of its use: it works alongside traditional research, not instead of it. Olivier Guillon puts it simply: “I don’t see it as a threat, it will just be another way to collect data.” Constanza Cilley agrees but also warns that “there is a danger of overestimating what synthetic data can deliver and underestimating the value of direct consumer engagement.”
Pieter Paul Verheggen (CEO, Motivaction, Netherlands) draws a useful boundary: “What it is and how it should be used must become clearer. (For example) If we use synthetic data to fill in missing values in questionnaires, nobody will complain. But if we replace real respondents for synthetic ones to make decisions on political or democratic topics, we must be very careful.”
Ludger Rolfes (Division Director, Produkt+Markt, Germany) is already working within those limits, noting that clients are more open to using the technology to fill missing survey data, and that synthetic data has real potential to streamline surveys, as well as “enable more robust analysis and efficient multivariate analysis.” And at the moment, Xavier Depouilly (General Manager, DXL Research and Consulting, Vietnam) uses it in a supportive role: “we use it only at the proposal stage to prepare example deliverables,” but does not push for clients to use synthetic data.
Sarah Mottet (Vice-President, Transformation and AI, Leger, Canada) states that success will not come from seeing the two methodologies as opposing each other but will come from “frameworks that leverage each where it creates the most value”. She captures the shared direction of travel: “The future of insights is not synthetic data versus real consumers. It is synthetic data and real consumers working together.”
The conversation is shifting
Even when adoption is still cautious, the direction of client sentiment is changing; “awareness is increasing, but adoption is still relatively low”, notes Ludger Rolfes, with trust, quality control, client education, and data governance being key challenges.
Karin Nelsson explains that these are risks that “undermine the trust that underpins all research”, so her team (and their clients) are in the wait-and-see mode while experiments are still underway.
For Sarah Mottet in Canada, there is a more active shift, “clients are no longer asking whether synthetic data will play a role in the future of insights. They are asking how to leverage it effectively, responsibly, and in ways that create tangible business value.” She sees a broader opportunity too, “the opportunity is not simply to do research faster, it is to ask better questions, explore more possibilities, and democratise access to insights.” Though she maintains that trust must be built, and transparency “is a prerequisite for trust”.
In Argentina, Constanza Cilley reflects that at this stage of synthetic data technology, “an experimental mindset and a commitment to continuous learning are probably more valuable than definitive answers” ,a reminder that the industry doesn’t need all the answers yet to make meaningful progress.
Hesitation persists despite progress
The barriers to adoption are real, and the network names them clearly. For Urpi Torrado, the challenge is especially acute in Latin American markets, where “understanding cultural nuances, emotions, and social context is critical,” and where synthetic training sets might not reflect local realities.
Chie Michihiro from Japan (Executive Researcher, Nippon Research Center) is succinct and sees more negatives in synthetic data’s current form, urging the need for more “reliability and theorization”.
Constanza Cilley points to structural gaps, the “limited availability of local expertise” and the “lack of proven local case studies” that could demonstrate when and how synthetic data actually works within local markets.
Elaine Japitana highlights that in the Philippines it’s a trust problem that runs deeper than methodology, “for those unaware of the process, they might consider the answers as fabrication or a fake response,” making building awareness and trust “a crucial step” before adoption can meaningfully progress, particularly when trust in heavily digital is already low.
Sarah Mottet and Ludger Rolfes both focus on quality. Mottet warns that “the quality of synthetic outputs is entirely dependent on the quality and representativeness of the underlying real-world data” used to train them, and that “not all synthetic data is created equal.” Rolfes reinforces this from practice, “clients need confidence that synthetic outputs accurately represent real-world behaviours,” and “synthetic responses can sometimes appear overly consistent or rational compared with real consumers”, requiring ongoing human oversight of every output.
For now, the industry finds itself at a productive edge, curious enough to explore, rigorous enough to question. That balance, the WIN members suggest, is exactly where it should be at this stage.