Most product decisions in 2026 are still being made the way they were in 2016. A discovery phase that takes six weeks. Five customer interviews, if you're lucky. A persona deck nobody opens after the kickoff. A roadmap built on the segments that answered the survey, not the ones that didn't.
The cost of that hasn't changed either. You ship the feature. You watch the metric. You find out which segment you misread, and by then it is too late to make changes.
Synthetic personas are not invented in a random Tuesday brainstorming session. They are not produced by asking an LLM to "pretend you're a 32-year-old SMB owner in Texas." A synthetic persona is a statistically grounded profile built from verified public data. Census microdata. Population-level distributions of income, household composition, geography, and housing. Joined so a single profile reflects the constraints of a real person, not a plausible one. The inputs trace back to a source you can interrogate.
Validation stops being a question of who you can find to interview. It becomes a question of what the population actually looks like, and how different segments will respond to what you're building.
Three things look different.
You can interview the segments you can never get on a call. Renters in rural counties. Households below the median income. People who never answer recruiter emails. Survey research systematically underrepresents them. A synthetic population does not.
You can run the same idea against multiple segments in an afternoon. Not because the AI is fast, but because the segmentation is already there, grounded in real data. You compare. You see why two segments that look the same on the surface make completely different decisions.
You can stress-test the riskiest assumption before you commit to the build. Pricing. Positioning. Which segment to ship for first. Cambium AI's product manager, MCP, returns a go, no-go, or partial verdict with the reasoning behind it, along with the persona interviews that informed it.
Consider a global streaming service weighing an ad-supported tier at $6.99, below its existing $9.99, $15.49, and $22 ad-free plans. The team already has deep data on its current subscribers. What it doesn't have is a picture of the people who aren't subscribers yet, the audience the new tier is supposed to capture.
A synthetic persona run on Cambium AI's product manager, MCP, fills that gap. The matching U.S. population comes back at roughly 7.5 million people, concentrated in young hourly workers, renters, students living with parents, and households below $50,000. The constraint profile is what makes the segment useful to design for. It skews heavily mobile-first. Many households are on capped or metered broadband rather than fixed wired connections. Account sharing is common.
That changes the design brief. The ad-tier experience needs to work on mobile data, in shared-account contexts, with reliable downloads, before it is optimised for the smart TV in the living room. That is a different product than the one the existing subscriber base implies, and it is not a finding most internal data sets would surface on their own.
Combined with the customer data the company already holds, the picture sharpens further. Existing data tells you which of your current subscribers are most likely to downgrade, which markets over-index on price sensitivity, and which content drives retention. The synthetic population tells you who isn't your subscriber yet and what they would actually need from the product.
An example of the full report can be viewed here.
The persona deck from 2016 told you who your customers were supposed to be. Synthetic personas tell you who they actually are, where the constraints sit, and what changes when you decide for one segment over another.
Validation stops being a discovery exercise and starts being a decision-making one. You walk into the prioritisation meeting knowing which segment the data points to, why the second-place segment loses out, and what would have to change about your product or pricing for that to flip.
Try out our product manager, MCP, and bring synthetic personas into your agent environment. For more information and access requests, click here.