The problem
A few hundred people, surveyed months ago, still guide many CPG launch decisions. That data is useful, but it is often too sparse, too dated, and too aggregated to answer the questions that matter most: which format to launch, at what price, for which shopper, in which channel.
Real surveys, panels, and sales data are still essential. They keep you tied to reality. The mistake is asking summary-level data to support decisions that require individual-level precision.
The traditional way
The usual approach is to start with a study, look at the averages, cut by segment, add assumptions, and drill down to a recommendation. It feels rigorous because every step has logic behind it. But each step also adds error.
By the time the team gets to the specific decision, they are often least certain at the exact moment they need the most confidence. That is how you end up with a precise answer built on a shaky foundation. No one gets fired for using Nielsen / Kantar / Ipsos. But 3 in every 4 launches still fail. Why?
The better approach
Flip the direction. Start granular, then calibrate up.
Build synthetic respondents at the individual level and simulate the actual trade-offs shoppers make: price, format, pack size, brand, claims, occasion, and channel.
Then use real data, including surveys, panels, sales, and segmentation, to calibrate the model. The real data anchors the system. The simulation adds the resolution and freshness the real data cannot provide on its own. Instead of forcing averages to answer specific questions, you use known reality to tune a more granular view of demand.
The future
This changes the economics of insight. You do not need to refield a full study every time you want to test another price point, format, claim, or audience. You can compare scenarios quickly, update assumptions as the market moves, and keep the model grounded in real-world signals.
That is the shift: from static research as the foundation to real data as the calibration layer. The insights leader should not have to defend a launch decision using research that was never designed to answer it. The analysis was not wrong. It was just starting from the wrong place. Data, technology, and human expertise combined allows to start from a better one.
