Recently, I had the chance to work through those exact questions with Priya Corbisier, Senior Customer Insights Manager at Woolworths, a major grocery chain and Australia's largest retailer with a sizable in-house research function. We conducted the first synthetic pilot in Australia with them. They'd heard about synthetic for a while, and they'd seen other vendors try and miss. They were ready to test it themselves, not to be convinced but to learn. "We didn't need that data at the end of this proof of concept (POC) to feed into something into the business," Priya told me. "We were really just doing it to learn."
Priya described a tension I hear from research leaders across the region. Her team supports product development, customer insight, and strategy across a large Australian retail group. The work is rigorous and slow — three-week fieldwork windows aren't unusual. In the meantime, the questions keep coming. From product teams. From UX. From every corner of the business that needs to know something before the next decision lands.
We ran synthetic studies in parallel with the team's human-panel studies. Same questions. Same methodology. Different respondents. The goal was to see where the two lined up, where they didn't, and what each tool was actually good for.
The first thing that hit Priya was the speed. "I knew it would be fast," she said. "But I was really surprised at how fast." From briefing, to questionnaire design, to output dashboards, the workflow moved at a pace that made the three-week human-panel cycle feel like a different era of research.
The fidelity held up too. When our Centre of Excellence team reran the dataset against the human-panel benchmark, the synthetic panel hit 95% on concept tests, 97% on attitudes and perceptions, and 99% on market landscape work. (The rerun used normalized discounted cumulative gain — the same measure our engineering team uses to decide whether a feature is ready to ship.)
But the most useful thing the POC produced wasn't a number. It was a framework. Priya's team came out of it with a sharper view of where synthetic actually belongs in their toolkit and where it doesn't. For early-stage ideation, hard-to-reach audiences, market views, and directional questions that need a fast answer, synthetic is a fit. "That's the role I see for synthetic," Priya said. "Driving directional insight. Not having to wait in a queue to get your piece of research done, but enabling internal stakeholders to make customer centric decisions early on in their concepts design process.”
For the high-stakes decisions, the kind you'd "hang your hat on," in her phrasing, human panels stay. When Priya checked in with the UX team, they told her they wanted more granularity than the current results provided — but in her words, they're "keen to use it as well," because the speed gain alone would be a game changer for how they operate. That's a useful data point. Synthetic isn't trying to replace anything. It's adding a lane.
For Priya's team, the practical implications are already taking shape. They're looking at where synthetic could substitute for some of the nationally representative attitudinal research they currently route through human panels, and at how the method might finally crack the hard-to-reach audiences that have always been a struggle for the team. They're watching what happens to use cases as accuracy keeps climbing.
That's the shift I'm watching across the field. The skepticism conversation is mostly behind us. The toolkit conversation is the one that matters now and the research leaders who get it right will be the ones who treat synthetic as a method, not a magic trick. Priya put it better than I could. "It will be interesting to see how we shift in the way we're using it," she said. "As we're getting closer to 90, 95% accuracy, what does that mean? Can we use it differently? That's what we're curious about."
That curiosity is the right starting point. The answers will come from the work.
