--- title: Vectorial Models Synthetic Users on Real Behavior Data so Product Teams Can Test Before Launch description: Vectorial has raised a $3M seed round from Smilegate and SenseAI Ventures for synthetic users that let product teams test features before launch. author: Darie Nani (Editor-in-Chief) date: 2026-10-20T13:34:58.092Z updated: 2026-10-20T13:35:11.819Z canonical: https://www.sovereignmagazine.com/article/synthetic-users-vectorial image: https://cdn.nanimediahouse.com/vectorial-founders-garg-kaur-singh.webp categories: Artificial Intelligence, Startups content_type: Spotlight region: San Francisco publication: Sovereign Magazine about: - type: Organization name: Vectorial description: Vectorial is a San Francisco human behavior simulation platform for companies building software products and digital experiences. Its behavior model, SAPIENS, builds modeled users of a company's target audience from real behavioral data, and product, design and marketing teams run simulations with them to test product journeys, campaigns and customer-facing AI agents before launch. It raised a $3 million seed round in October 2026 from investors including Smilegate and SenseAI Ventures. url: https://www.runvectorial.com/ industry: Artificial Intelligence --- Vectorial, a San Francisco startup that lets product, design and marketing teams try out a new feature, campaign or AI agent on thousands of synthetic users before any real person sees it, has raised a $3 million seed round. Smilegate and SenseAI Ventures invested alongside Berkeley SkyDeck, the Berkeley AI Research Lab, a Stanford GSB fund, Dent Capital, Motus Ventures and SUM Ventures, and product leaders from Meta, Shopify and Mastercard joined as angel investors. The modeled users come from SAPIENS, a behavior model Vectorial trained on real data. The company has modeled more than 2.5 million real people and run more than 30,000 simulations for customers in healthcare, gaming and e-commerce, including GoodRx, MIDI Health, Suki AI, the Indian consumer electronics brand boAt and the Canadian listed company Everkind. The new funding will pay for data partnerships to widen its behavioral datasets and for SAPIENS 2.0, the next version of the model. ## Traditional User Research Before a Product Launch A typical user research study draws on 50 to 150 paid participants who are recruited, interviewed one at a time and then summarized into findings that a company applies to millions of customers. Each round takes weeks, and any follow-up question means starting a new study. AI-moderated interviews have cut that to days, but teams using [AI coding tools](https://www.sovereignmagazine.com/article/anthropic-and-openai-release-competing-ai-coding-models-within-minutes-of-each-other) can now change a product within hours, so many skip the research and decide on instinct. Others test ideas on live customers through A/B tests and beta releases, which need enough traffic to give a reliable answer and leave real users to deal with whatever does not work. "Building a successful consumer brand involves hundreds of decisions, from product features and pricing to positioning, messaging, and campaigns. Yet traditionally, consumer research has only informed a fraction of those decisions," said Vedansh Kumar, chief marketing officer of boAt, which has $350 million in revenue and a $2 billion valuation. "What excites me about Vectorial is the possibility of changing that equation entirely, bringing consumer understanding into everyday decision-making at a speed and scale that simply wasn't possible before." ## Synthetic Users Grounded in Real Behavior Data Synthetic users are AI-generated stand-ins for real customers that a team can question or send through a product instead of recruiting people. Most synthetic-user tools are [general-purpose large language models told to play a persona](https://www.sovereignmagazine.com/article/the-matrix-gets-more-factual-every-year), with no grounding in data on how real people behave. Vectorial says those models plateau at 51 to 56 percent accuracy when predicting human behavior. Simile and Aaru, the simulation companies Vectorial is most often compared with, mainly answer strategic questions such as which market to enter or how the public will respond to an event. Vectorial works on the decisions that come after, such as which onboarding flow ships, which feature is built next and which customer segment a campaign should reach. SAPIENS learns the socio-demographic, psychographic and behavioral traits of a company's target customers. It draws on what people say in public online communities, what they reveal in panels and interviews, and what they do, as recorded in product analytics and customer conversations. Those sources cover more than 30 million reachable people across eight behavioral dimensions, and a company can segment its modeled audience by age, income, geography or life stage. Vectorial says SAPIENS predicts behavior 30 percent more accurately than general-purpose LLMs and outperforms two academic models of human behavior, [Be.FM](https://arxiv.org/abs/2505.23058) from the University of Michigan and [Centaur](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12390832/) from Helmholtz Munich. In customer work, its simulations have reproduced the findings of earlier user studies with more than 85 percent accuracy and predicted survey results to within eight percentage points, the company says. Each simulation comes with the anonymized profiles of the modeled users, the traits the model inferred and the reasoning behind each response, and Vectorial checks its accuracy against parallel studies with real users. Because the model keeps taking in new data, companies can also use it to follow how their customers' behavior changes and to find new customer segments as they appear. "Enterprises we serve treat a continuously enriched model of their target audience as a strategic investment," said Jasmine Kaur, co-founder and chief business officer of Vectorial. "We see organic adoption across product, design and marketing, because they all design for the same audience." ## Synthetic User Testing for Products, Campaigns and AI Agents > "We continue to invest in our behavior model, but a model alone doesn't solve the problem. That's why we put behavior models to work inside the product development loop." > — Rahul Garg, co-founder and chief product officer of Vectorial Once Vectorial has modeled a company's target audience, teams hand it a product journey, a landing page, a marketing campaign or a customer-facing AI agent on web, mobile or AI-native apps. The modeled users work through it as the real audience would, and the results show how each person reacted and why. Customers run around 600 simulations a month, which Vectorial says amounts to about 20 times the research they did before. ![Vectorial journey timeline showing how modeled users reacted at each screen of a product test](https://cdn.nanimediahouse.com/vectorial-synthetic-users-journey-timeline.webp) Everkind used Vectorial to run 10,000 simulated sessions with its AI agents in a matter of hours, each with a different modeled user drawn from a range of populations. The sessions found more than 100 gaps in how the agents behaved with different kinds of people, which Vectorial estimates would have taken more than 1,500 hours of expert review to find. "At Everkind, we hold an exceptionally high bar for AI behavior and alignment. But a conversation that works for one person may land very differently for a first responder coming off shift or a parent running on no sleep," said Supreet Pal Singh, chief technology officer of Everkind. "Vectorial lets us simulate those real-world differences end to end, extending our evaluation across diverse human behaviors before an update ever reaches a real user." Health technology companies use the platform to model patients and test how a new service would be taken up at each stage of the patient journey. Gaming studios use it to aim campaigns at the right player segments, find audiences beyond their existing players and try out game mechanics with modeled players before committing development time. Concrete Software, the developer behind the PGA Tour and NHRA Drag Racing games, is one of them. "The challenge in game development is knowing what to build first. Vectorial gives our teams a way to engage with modeled players throughout development and live ops, test feature and content ideas against different audiences and get a first read in minutes instead of days or weeks," said David Nixon, vice president of data services at Concrete Software. "It helps us make better decisions about where to deploy our resources, without relying solely on the instincts of a few experienced people interpreting feedback from a small handful of players." ## Vectorial Founders Taranveer Singh, Jasmine Kaur and Rahul Garg ![Vectorial's three co-founders under the company sign at its office in Menlo Park](https://cdn.nanimediahouse.com/vectorial-founders-menlo-park-office.webp) Taranveer Singh, Vectorial's co-founder and CEO, has spent a decade modeling human behavior. He began in defense research, modeling how large populations might behave over time, then at Carnegie Mellon worked with Andrew Ng's team at Coursera on predicting how learners from different backgrounds would learn. In a Chegg-funded research group with PhDs from Stanford, CMU and Columbia, he used billions of résumés to model career decisions. In each project he found people more predictable than is commonly assumed, and that work became SAPIENS. He is also a visiting research fellow at UC Berkeley's Berkeley AI Research Lab. Jasmine Kaur, a Stanford GSB graduate and Singh's partner, led AI work at Pfizer on synthetic users grounded in thousands of real-world user studies, to help the business understand its customers. Rahul Garg, the chief product officer, studied at BITS Pilani and the Indian School of Business and has led more than ten new products at several unicorn startups. At his previous startup, Zime AI, he worked on a "behavior graph" to model the psychology behind sales conversations. The three had been friends for years and decided over drinks one New Year's Eve to start the new year building Vectorial together. The company now has 10 employees. ## Vectorial's $3M Seed Round and SAPIENS 2.0 Smilegate is a South Korean games company, and SenseAI Ventures runs a $25 million fund for early-stage AI startups. "Building software has evolved to lightspeed, but understanding user needs remains painfully slow," said Rahul Agarwalla, managing partner of SenseAI Ventures. "Vectorial replaces the slow theater of surveys with millions of simulated humans using the product and providing insights. This is a new class of AI to predict human behavior and not just the next word like an LLM. That is why we invested." Vectorial will use the money to add behavioral data through new partnerships and to develop SAPIENS 2.0, which it wants to become the standard simulation infrastructure for building and launching products. The company expects to reach $1 million in annual recurring revenue within six months. Its longer-term plan is aimed at AI coding agents, which can already run hundreds of experiments on their own to meet goals for speed, cost and performance that engineers set. Vectorial wants those agents to test each design, flow and message against its modeled users while they build, with every simulation showing which users were modeled, how they reacted and why the agent changed course. The founders expect that within four years, running a behavior simulation with modeled audiences before building or launching a consumer product or AI agent will be standard practice, as simulation already is in aerospace, manufacturing and chip design. "We believe AI is creating two distinct frontiers. One focuses on reasoning and solving complex problems, while the other focuses on simulations and decoding human behavior," Singh said. "Two pieces are missing: the ability to simulate many scenarios, and bringing your users into the same room as you build." **About Vectorial** Vectorial is a San Francisco human behavior simulation platform for companies building software products and digital experiences. Its behavior model, SAPIENS, builds modeled users of a company's target audience from real behavioral data, and product, design and marketing teams run simulations with them to test product journeys, campaigns and customer-facing AI agents before launch. It raised a $3 million seed round in October 2026 from investors including Smilegate and SenseAI Ventures. [Website](https://www.runvectorial.com/) ## FAQ **Q: What is a synthetic user?** A synthetic user is an AI-generated model of a real customer that a team can question or send through a product, campaign or AI agent instead of recruiting participants. Vectorial's synthetic users come from SAPIENS, a behavior model trained on real data from more than 2.5 million people, rather than from a general-purpose language model playing a persona. **Q: How accurate are synthetic users?** It depends on the data behind them. Vectorial says general-purpose LLMs predict human behavior with 51 to 56 percent accuracy, while SAPIENS is 30 percent more accurate and has reproduced the findings of earlier user studies with more than 85 percent accuracy. Each Vectorial simulation is also checked against parallel studies with real users. **Q: Is UX research being replaced by AI?** Not entirely. AI-moderated interviews have cut studies from weeks to days, and synthetic users such as Vectorial's give a first answer in minutes, which its customers use to run about 20 times more research than before. Vectorial still validates its simulations against studies with real people. **Q: Who invested in Vectorial?** Smilegate and SenseAI Ventures invested in the $3 million seed round alongside Berkeley SkyDeck, the Berkeley AI Research Lab, a Stanford GSB fund, Dent Capital, Motus Ventures and SUM Ventures, with product leaders from Meta, Shopify and Mastercard investing as angels.