Simile raises $200M for synthetic users. You have real ones

Simile raised $200 million at a $2 billion valuation on July 30, five months after a $100 million Series A. The product: "agentic twins" — AI stand-ins for real consumers that companies survey instead of running actual market research. It's a serious round from serious investors. It's also a diagnostic, and what it diagnoses is a problem you probably don't have.

What actually happened

Per TechCrunch, Greenoaks led the round, with Index Ventures returning alongside Hanabi, Bain Capital Ventures, A*, Factory, and CVS Health Ventures, plus new investor Definition. The company is roughly a year old and based in Palo Alto. CVS Health is described as a marquee customer.

Founder Joon Sung Park did the underlying work at Stanford — the "Smallville" research, where generative agents were dropped into a simulated town and went about simulated human lives. Simile's stated ambition is to model all eight billion people on earth. A competitor, Aaru, raised a Series A at a $1 billion valuation in December, so this isn't a one-company thesis.

TechCrunch is openly skeptical, and the skepticism is the right shape: the entire reason to do market research is that humans are unpredictable, driven by emotion as much as reason. A model trained to predict the average response is structurally worst at the thing you most need research for — the surprise. We'd add the obvious caveat that every accuracy claim in this category is currently vendor-supplied.

Why synthetic users matter for your business

Ask why a company pays for simulated customers. It's not that talking to real ones is impossible. It's that at CVS scale, reaching a representative sample takes a panel vendor, six weeks, and a budget line — so a fast, cheap approximation genuinely beats the alternative of guessing.

You are not at that scale, and that is a structural advantage, not a limitation. You can text eleven customers today and have real answers by dinner. The enterprise is buying a synthetic version of an asset you already own outright.

So use the actual thing:

• Your support inbox is a research corpus. Six months of tickets tells you what confuses people, in their words, unprompted. Nobody has to simulate it.
• Your churn list is the highest-signal panel you'll ever have. People who left will tell you why with a bluntness no survey respondent offers. Ten calls beats ten thousand simulated responses.
• Ship the small version and measure. A landing page with a real price and a real button answers "would you buy this" better than any twin. Intent is cheap; a completed checkout isn't.
• Log the pattern once you find it. The reason enterprises lose this signal is that it never got written down anywhere queryable. Put it in a database you own, tagged and searchable, before it's a folder of screenshots.

Where synthetic users would earn their keep at your size: pressure-testing copy or pricing before you burn a real customer conversation on it. That's a drafting tool. It is not evidence. Don't let a simulation be the last thing you talk to before you build.

Sources: TechCrunch, FinSMEs.