AI-generated representative user profiles grounded in real demographic data. Unlike marketing personas (fictional characters), synthetic personas are statistically generated from population survey data and evaluated by AI models to simulate diverse perspectives. Modest Idea uses 250 Census PUMS-grounded personas per analysis, each with demographic attributes, occupation, income, and a Big Five personality profile.
Traditional user personas are invented. A product team brainstorms who their user is — usually a slightly idealized version of the founding team's social circle — names them "Sarah, 32, urban marketing manager," and makes product decisions based on Sarah. Sarah is a guess dressed up as a user.
The problem isn't that user personas are useless — understanding your user is essential. The problem is that invented personas don't include the people you didn't think of. When your founding team is three software engineers in their late twenties, your invented personas probably don't include a 52-year-old night-shift nurse in suburban Ohio who turns out to be your highest-PSF user. Synthetic personas include her because the Census data includes her.
Synthetic personas also enable systematic evaluation at scale. Once you have 250 demographically representative personas, you can evaluate each one against your product concept using AI models — asking, in effect, "does this person have this problem, and would this solution address it?" Across 250 personas, patterns emerge that no amount of founder intuition or 10-person user interview study would surface.
Each persona in Modest Idea's database starts from a real record in the US Census Bureau's American Community Survey (ACS) Public Use Microdata Sample (PUMS). This provides ground-truth demographic attributes: age, sex, race/ethnicity, education level, occupation, income, household composition, and geographic region.
These raw demographic records are then enriched with:
When a user runs an analysis, Modest Idea samples 250 personas using IPF weighting to ensure the sample matches real US population distributions. Each persona is then evaluated by multiple AI models against the product concept.
When evaluating a habit accountability app against 250 personas, the analysis includes a persona like: Maria, 38, registered nurse, Houston TX, night shift, income $68K, household: partner + 2 children, Conscientiousness 3.2/5, Neuroticism 3.8/5, technology adoption: early majority.
This persona isn't invented — her demographic profile is derived from real ACS data on nurses in Texas. Her personality scores are generated to be consistent with known occupational patterns. When evaluated against the habit app concept, Maria scores high on problem recognition (rotating schedule destroys routine), high on pain severity (can't maintain exercise habits), and high on solution gap (mainstream apps assume 9-to-5 schedules). PSF: 84.
Without synthetic personas including shift workers, this analysis would return results skewed toward the founding team's social circle — and the highest-PSF segment would remain invisible.
Synthetic personas are not real people and don't represent real individuals. They're statistical simulations designed to surface demographic diversity in problem recognition — not to predict any individual's behavior. The insights they generate are probabilistic and directional, not deterministic. Use them to identify which segments to investigate further, not as a substitute for talking to real users.
Synthetic personas are AI-generated representative user profiles grounded in real demographic data. Unlike marketing personas (fictional characters invented by a team), synthetic personas are statistically generated from population survey data and evaluated by AI models to simulate how diverse real-world people would respond to a product concept.
Traditional marketing personas are fictional archetypes created by product teams, often reflecting the founders' own demographic assumptions. Synthetic personas are generated from real survey data (like Census PUMS), so they represent actual population distributions — including demographic groups the founder might not think to include. They're designed to surface blind spots, not confirm existing assumptions.
Get our free PSF Framework guide — a 5-step process for evaluating problem-solution fit, with scoring templates and real case studies.
Get the Free Guide →Explore our demo analyses to see how 250 Census-grounded personas evaluate real product concepts across diverse segments.
Read the methodology →