How I see it
One argument in three parts: the instruments we relied on are breaking down, they were never measuring quite the right thing, and decision-making has shifted into a space we can now study directly.
Surveys are running out of road
For the past twenty-five years, market research has largely depended on online survey panels. They still have a role, but they are unlikely to remain the backbone a decade from now. Response quality is declining, incentives increasingly shape behaviour, and AI agents can now complete surveys at scale while passing many quality checks. The risk is not that surveys disappear, but that a new backbone emerges first.
Preference is built, not retrieved
We often assume people hold fixed preferences that good questions can uncover. But decades of research suggest otherwise: many brand judgements are formed at the moment of choice, shaped by the information available at the time. A lot of what we measure is therefore misaligned with how decisions actually happen. The sharper question is which information environment is shaping the decision.
The decision now runs through AI
That information environment increasingly runs through AI. When a model is asked about a category, from mattresses to B2B platforms, its output is often the first input into the decision, before search or reviews. The shift is that category understanding is now formed in these systems, and can be studied directly.