A client sent over a Q1 research budget last week with a line item I had to read twice. Synthetic respondents. Two hundred simulated consumers, delivered in forty-eight hours, for roughly a ninth of what the eight in-person interviews cost us back in October.
I asked what the vendor had promised. The panel matched our segment on twelve attributes. It never no-showed. It could be re-run as many times as we liked, with the questions adjusted between passes.
Re-run as many times as we liked.
That is the pitch, and it is not a pitch about accuracy.
A simulated respondent can only hand back what somebody already fed the model, which means the thing you are buying is not insight, it is agreement at speed.
Here's the part that gets lost. The value of real fieldwork was never the tidy data at the end. It was the friction on the way there.
One. The woman who answers your carefully worded concept question with a four-minute story about her mother's kitchen. That story is where positioning comes from. A model does not produce it, because a model has never had a mother or a kitchen.
Two. The pause. I have sat behind the glass and watched somebody say they loved a package while their shoulders said the opposite, and everything useful in that session lived in the gap between the two.
Three. Accountability. Real respondents are expensive, so you ask only what actually matters, and then you have to live with what they tell you. Answers that cost nothing get treated as though they are worth nothing, which is usually correct.
Consider what an averaging machine would have done to the brands we all cite in decks. Ask two hundred modeled consumers whether they want water in a heavy metal tallboy and Liquid Death never leaves the room. Ask them about a milk alternative that argues with you in tiny type on the side of the carton and Oatly never leaves the room. Ask them about pear and blue cheese ice cream and Salt & Straw never leaves the room. Every brand I admire was built on a thing the average said no to.
Every brand I admire was built on a thing the average said no to.
I am not against the tools. Use them to rough out a discussion guide, or to work out which six questions are worth a real person's hour. What I will not do is put a simulated quote on a slide and let a room treat it as a human being.
Truffles has exactly one strong opinion, which concerns the left corner of the couch. Barnaby has never once agreed with her about it. Six years of genuine, unresolvable disagreement in my living room, and I still learn more from it than from anything built to say yes.