SimileThe Story
Simile, which builds simulated populations that enterprises can question in place of real survey respondents, has raised over $200 million at a $2 billion post-money valuation in a Series B round led by Greenoaks. The company announced the round on 30 July 2026. Index Ventures, which led the previous round, invested again. Hanabi, Bain Capital Ventures, A*, Factory and CVS Health Ventures returned, and Definition came in as a new investor. CVS Health is also named by the company as one of its customers, so CVS Health Ventures sits on both sides of the relationship. Simile has not disclosed revenue in dollars, the valuation at which it raised its Series A, how the $200 million divides between the participants, whether any of it is secondary, or what board changes accompany the round. The company says revenue has grown five times since launch without saying what it grew from. The round comes five months after Simile emerged from stealth in February 2026 with a $100 million Series A led by Index Ventures, announced on 12 February, which included Bain Capital Ventures, Hanabi Capital and A* along with angel cheques from Fei-Fei Li and Andrej Karpathy. That round's valuation was never made public. The company was founded by Joon Sung Park, who authored the 2023 generative agents paper behind Smallville, an experiment that placed 25 autonomous characters in a simulated town and observed their social behaviour. His co-founders are the Stanford researchers Percy Liang and Michael Bernstein, along with Lainie Yallen. Simile says it has run tens of millions of simulations for Fortune 100 enterprises, built a model that predicts the accuracy of each simulation it produces, and grown to more than 50 employees. It names CVS Health, Wealthfront, Deloitte and Gallup as customers.
Why It Matters
Asking people what they want is slow and expensive. A focus group has to be recruited, scheduled and moderated, a survey panel has to be sampled and weighted, and the answers arrive weeks after the question was worth asking. For a company deciding whether to launch a product or change a policy, the research often finishes after the decision has been made. Simile sells a population that answers immediately. Its models are trained on interviews with real people and on behavioural data, and the resulting agents can be questioned repeatedly, at any hour, about scenarios that have not happened. There is no recruitment cost per study and no limit on how many times the same simulated respondent can be asked something new. The costs sit in different places from a research agency's. Collecting the interview data that grounds the models is expensive and has to be repeated as populations change. Training and running the models is a compute bill. What the company sells against that is enterprise software, though it has not said how it prices, whether by seat, by simulation or by contract. None of the economics are visible. Simile has published no revenue figure, no customer count and no contract values. Its stated fivefold revenue growth is measured from a company that started selling in February, which makes the multiple a statement about a small starting number rather than about scale. And one of the customers it names, CVS Health, is also an investor through CVS Health Ventures, which is a relationship worth knowing about when reading a client list.
The Strategic Read
The market assumption changing behind this investment is that a simulated person is close enough to a real one that a company will spend real money on the difference between them. Market research has always been sold on method rather than on speed. A panel provider's product is the defensibility of its sample, which is why the industry survived decades of cheaper alternatives. The answers arrive slowly and cost thousands per session, but they came from people who exist. Simile's proposition removes the people and keeps the answers, which means the entire value of the product rests on how closely the simulation tracks what a real population would have said. That is why the most consequential thing in the announcement is not the valuation but the confidence model. A system that tells a customer how much to trust each individual simulation is an admission that some simulations are wrong, and a claim that the company knows which ones. If that calibration holds up against outcomes the model has not seen, Simile has something a competitor cannot copy by training on the same public data. If it does not, the number is a comfort blanket sold alongside the product, and no external party has yet published a check on it. Gallup appearing on the customer list is the sharpest signal available. A firm whose name is a byword for asking people what they think has adopted a tool that stops asking them. That is either an early sign the category works or an expensive experiment inside an incumbent, and nothing published so far distinguishes the two. The valuation prices in the first reading. At $2 billion across a team of just over fifty, five months after launch, with revenue described only as a multiple of an undisclosed base, the round assumes the validation problem gets solved and that pilots convert into recurring contracts. The data underneath also ages. Populations shift, and interviews have to be recollected to keep a simulated public current, which is a recurring cost that behaves less like software than the multiple implies. One published case of a confidently wrong simulation on a decision that mattered would reprice the whole category, not just this company.
For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.
