On October 20th, Vectorial, Vectorial, a San Francisco startup, raised $3 million to sell you customers who don’t really exist.
They’re called synthetic customers, or synthetic personas: AI-generated stand-ins built to think, shop, and react like the real people you’re trying to reach. Feed the model your data, describe a segment, and you can run a focus group before lunch without a single human in the room.
This isn’t a fringe experiment anymore. Harvard Business Review recently predicted that AI simulation tools will disrupt the $140 billion market research industry this year. The marketing world is aflutter.
We’ve been doing discovery for nearly twenty-five years, so naturally, we have thoughts.
Credit Where It's Due
Let’s be fair. Synthetic customers have real upsides. They’re fast. Research that once took weeks can happen in hours. They’re cheap compared to recruiting, scheduling, and paying real participants. And they’re great for early brainstorming: testing a dozen headline ideas, poking holes in a concept, or spotting objections before you spend a dime.
Used that way, they’re a useful sparring partner. A rough draft of your audience, available on demand.
The Catch
Here’s the problem. A synthetic customer only knows what you feed it.
If your data is outdated, your persona is outdated. If your assumptions are wrong, your persona will politely agree with them. And experts have noticed something else: a room full of synthetic personas can look diverse on paper, with different ages, jobs, and backgrounds, while their answers come out far more alike than real people’s ever would. Where’s the nuance, people?
Real customers are messy because of their humanity. They contradict themselves. They mention the thing you never thought to ask about. They get emotional about your logo, or bored by your pitch, or confused by a word you use every day. That mess is where the gold is.
Synthetic customers are the cheapest dinner guests you’ll ever host. They’ll eat anything you put in front of them and compliment the chef every time. And if you’ve read anything we’ve written lately, you know how we feel about the dangers of believing your own BS.
Imaginary Informants
In the intelligence world, the best information comes from the field. Real sources, real conversations, real surprises. An analyst who only interviews people they invented would be laughed out of the room.
That’s essentially what synthetic research asks you to do. You build the informant, you write their backstory, and then you’re shocked when they tell you exactly what you expected to hear. You’re interrogating a mirror.
It’s why the SCIF Session, our discovery process, is a 90-minute conversation with our founder, Cat Fortel. Real person, real questions. She digs into your origins, the evidence behind your claims, and the challenges you’d rather not say out loud, then turns it all into The Map. Some of the most important insights we’ve ever uncovered came from a single offhand comment a client never would have typed into a survey.
Use Them Without Fooling Yourself
Synthetic customers can earn a spot in your toolkit, as long as you know their job. A few ground rules:
- Use them early. They’re great for brainstorming and stress-testing ideas before real research.
- Feed them real evidence. Current customer data, reviews, and conversations. Garbage in, very confident garbage out.
- Never let them have the final word. Before you bet real money on a decision, put it in front of real humans.
- Watch for agreement. If the synthetic panel loves every idea you give it, that’s a red flag.
AI can imagine your customers. Only your customers can surprise you.
The brands that win the next few years will use every tool available, including the synthetic ones, while staying stubbornly curious about the real people on the other side of the screen.
The Bottom Line
Want to hear what your real customers are thinking?
Book a Chemistry Call and ask us about the SCIF Session. No imaginary informants required.
