Vectorial Raises $3M Seed to Test Products Against Simulated Customers

AI can help a team build a new product feature in hours. Finding out whether people actually want it can still take weeks. Vectorial is betting that human behavior models can narrow that gap, and has raised $3 million in seed funding to put simulated audiences inside the product development process.

The San Francisco company announced the financing on October 20, with backing that includes Smilegate, SenseAI Ventures, and Dent Capital, according to its announcement. Vectorial also identifies support from the Berkeley and Stanford ecosystems and participation from product leaders at Meta, Shopify, and Mastercard. It plans to expand its behavioral datasets through new data partnerships and develop the next version of its SAPIENS model.

The proposition is straightforward: let product, design, and marketing teams examine how different audiences might react to a digital experience before exposing real customers to it. The harder question is whether those simulated reactions are reliable enough to improve decisions.

Building an Audience Model That Can Be Reused

Vectorial describes its platform as a two-step system. First, it builds a model of the target audience using behavioral data. Then teams run simulations against that audience, evaluating product journeys, campaign concepts, brand messages, or customer-facing AI agents across web and mobile experiences.

Its SAPIENS model represents demographic, psychological, and behavioral characteristics, including traits that are not directly visible in a customer profile. The aim is to capture differences between people rather than treat a market segment as one average customer. A design team and a marketing team can then work with the same underlying audience model while asking different questions.

On its website, Vectorial explains that customers can calibrate populations using CRM segments, earlier surveys, support transcripts, and product analytics. It also describes back-testing against a previously completed study before using the modeled audience for a new question. That creates a practical way to examine whether the simulation reflects a particular customer base.

For a product team, this could mean comparing alternative onboarding flows before committing engineering resources. A game developer could investigate how different player groups respond to a proposed feature. A company deploying an AI assistant could examine how its conversations change across user circumstances. These are potential applications of the platform, rather than evidence that every predicted response will occur after launch.

What the Accuracy Claims Measure

Vectorial says it has modeled more than 2.5 million people and completed over 30,000 simulations. Its funding announcement reports that simulations have reproduced earlier user-study findings with more than 85% accuracy and describes SAPIENS as 30% more accurate at predicting behavior than general-purpose language models.

Those are company-reported results. Its current website separately gives an 87% accuracy figure for held-out studies and a 35% improvement for matching real opinions. The company explicitly distinguishes opinion matching from behavioral accuracy. These figures should therefore be read as results for different evaluations, rather than a single interchangeable measure of how well software predicts people.

For prospective customers, the useful test is how the model performs on their audience, their product, and decisions with known outcomes. A simulation that ranks two messages correctly is answering a different question from one that predicts the exact percentage of users who will buy. Vectorial itself acknowledges weaker performance for precise predictions in categories with limited behavioral data.

This is also where simulated research differs from AI tools that automate market-research workflows. Faster survey preparation and analysis can improve the process of gathering human evidence. A behavior simulator instead generates predictions from a model, making calibration and validation central to the value of its output.

From Research Expertise to Product Decisions

Vectorial was co-founded by Taranveer Singh, its CEO; Jasmine Kaur, its chief business officer; and Rahul Garg, its chief product officer. Their backgrounds combine behavior modeling, enterprise AI adoption, and software product development.

The company’s founder biographies describe Singh’s work across behavioral research and AI, including Carnegie Mellon and Berkeley AI Research. Kaur previously worked on synthetic users for marketing at Pfizer, while Garg has built products at companies including Blinkit and Housing.com. That combination helps explain Vectorial’s focus on inserting research into recurring product decisions, rather than selling only occasional market studies.

“We continue to invest in our behavior model, but a model alone doesn’t solve the problem,” Garg said in the announcement. “That’s why we put behavior models to work inside the product development loop.”

Vectorial lists customers across healthcare, gaming, and commerce, including Midi Health, GoodRx, Suki AI, boAt, Everkind, and Concrete Software. The announcement says customers run around 600 simulations a month, which it equates to roughly 40 user studies. Those usage figures indicate the frequency of testing the company is targeting; they do not by themselves establish the commercial effect of each simulation.

A Faster First Read, With Real-World Validation Still Essential

The seed funding will support more behavioral data partnerships and SAPIENS 2.0. Expanding data coverage matters because the platform’s usefulness depends on how faithfully it represents the people a business wants to reach, including groups that may behave differently from the majority.

The strongest case for this approach is earlier, more frequent testing. Teams can explore more options, identify assumptions, and arrive at human research with sharper questions. Vectorial’s own website says real participants remain important for unfamiliar situations and work requiring direct human evidence.

That leaves a clear challenge for the startup: show that repeated simulations lead to better product choices, not simply more research output. If it can demonstrate that connection across different customers and markets, audience modeling could become a regular step between generating an idea and shipping it.