Zach Epstein Would Like to See the Control Group

The Larry David problem

Zach Epstein has a particular objection to a certain kind of sales pitch. Someone promises a remarkably precise answer from a remarkably small advertising experiment, and his patience begins to thin. The confidence is high. The holdout is tiny. The arithmetic, he suspects, is being asked to carry more than it can bear.

He has a name for the disposition this brings out in him: Larry David. People have made the comparison, he says, and he is still deciding whether to accept it as a compliment. “I’m fun like that,” he has written of his tendency to offer an opinion without waiting for an invitation. His version of a conversational provocation involves statistical power and standard errors. Every industry gets the dinner guest it deserves.

That mix of skepticism and amusement helps explain Haus, the company he founded in 2021. Epstein’s subject is advertising, but his recurring concern is evidence. A customer sees an ad and buys something. A dashboard gives the ad credit. The question that interests him comes a little earlier: was that customer going to buy anyway?

It is an awkward question if you sell advertising. It is a useful one if you buy it. Epstein has spent his career getting closer to the place where the two answers diverge.

Before the dashboard, the movies

His route into that question ran through entertainment. Epstein attended Occidental College, where his honors thesis examined political action committee contributions and election outcomes. Even the title posed a spending question: were PACs wasting their money? The setting would change; the difficulty of connecting money to an outcome would remain.

His first job was at Digital Domain, the visual effects company whose founders included James Cameron. He has described that early work as a strategy analyst role. He later worked at Sony Pictures Entertainment before moving to Google. Movies give the measurement problem unusually concrete stakes: someone must decide whether a campaign will persuade an audience to give up an evening and pay for a ticket.

Entertainment still shows up in how Epstein explains his work. A film, in his telling, competes with all the other ways a person might spend time and money. Advertising enters a life already crowded with preferences, habits and alternatives. An audience member can see the trailer and still have planned to go all along. That ordinary fact makes an extraordinary amount of accounting uncomfortable.

An education inside Google

At Google, Epstein worked as a data scientist and product manager on analytics, experimentation and marketing science. He has described leading a data science team that worked with sophisticated advertisers, including Netflix and Hulu. Their questions demanded more than a report of which click came last.

The work brought him into the practical machinery of causal inference: designing a comparison that helps estimate what would have happened without a particular intervention. An advertiser wants to know how many purchases a campaign created. Counting purchases after an ad ran is easier. Separating the campaign’s effect from demand that already existed takes an experiment.

A simple illustration helps. Imagine advertising continues in some regions while comparable regions are held back. Analysts compare the outcomes, accounting for the differences between those places. The extra business attributable to the intervention is the object of interest. This diagram is a conceptual explanation, rather than a report of a Haus customer’s results.

Epstein also worked on privacy-safe advertising technology. That experience mattered to the business he would start: measurement built around first-party business outcomes could answer questions without depending on the same individual tracking infrastructure as older approaches. He had seen both the value of experimentation and the resources required to do it well. Most companies could not simply assemble a specialist team on demand.

Leaving with a question

In 2021 he announced that he had left Google and was building a company in stealth. The message was practical, almost brisk. Privacy was changing the data landscape; his team was developing infrastructure to help companies keep making decisions within it. He needed full-stack and back-end engineers. An idea had become a hiring problem.

The departure did not receive universal applause. Looking back, Epstein recalled people telling him it was a bad idea. He understood the objection. Turning econometrics and experimentation into something a marketing team can use routinely is a formidable product problem. The elegant method must survive imperfect data, impatient meetings and a budget that needs an answer this week.

Early believers joined him anyway. Olivia Kory later described arriving at a stealth startup with five customers and no product. By June 2026, when Haus named her chief marketing officer, that early wager had become part of the company’s history. Her account makes the beginning feel properly unfinished: there were customers to serve before there was a polished platform to show them.

“Answering business questions shouldn’t be harder than a Google search.”

ZACH EPSTEIN · 2023

The investors wanted an experiment, too

Haus’s financing story contains a pleasing symmetry. Insight Partners, which led its $17 million Series A in August 2023, initially questioned whether such complicated methods could become software without requiring an army of specialists. The investor wanted to know whether this was a repeatable product or a consulting operation wearing a product’s clothes.

Its team examined the technology, watched demonstrations and spoke with customers. The firm’s account of the investment describes that skepticism giving way to conviction. For Epstein, the obstacle was also the proposition: build enough expertise into the infrastructure that customers can use the methods without rebuilding the institution behind them.

In July 2024, Haus announced another $20 million, led by 01 Advisors. That brought the company into partnership with a firm led by David Fischer, Dick Costolo and Adam Bain, executives with their own histories inside large advertising and technology businesses. The announcement also introduced Causal Attribution, designed to bring experimental learning into a daily view of marketing performance.

Recognition followed, including a place on LinkedIn’s 2024 list of top U.S. startups. Epstein used the occasion to praise the team and the detail behind its work. For a founder preoccupied with what can be measured, the distinction was a public marker of a much less visible effort: making a difficult scientific process usable, again and again.

Economists in the room

By October 2025, the scientific circle around Haus included Susan Athey as an adviser. A Stanford economist and John Bates Clark Medal recipient, Athey brought experience in both academic methods and digital advertising. Epstein introduced the appointment as an extension of the company’s commitment to econometrics.

Athey’s explanation of the opportunity went to the access problem. Major technology companies had long employed scientists to improve their own systems. Smaller firms could rarely reproduce that expertise. Software offered a way to carry the methods beyond the organizations that first had the staff and budgets to put them to work.

That same month, Epstein announced the availability of Causal MMM, following more than a year of work with design partners. Marketing mix modeling estimates the contribution of different marketing activities; Haus’s approach puts experiments at its foundation. It was another attempt to connect pieces of an advertiser’s working life that often sit apart: tests, models, reports and the meeting where somebody finally approves the spending.

The next dollar

In May 2026, Epstein presented Architect at events in Los Angeles and New York. His June introduction framed the next problem plainly: learning what worked does not automatically tell a business what to do tomorrow. The campaign has ended. The experiment has been read. The next budget still has to be allocated.

He looked to portfolio management for an analogy. Finance had developed automated systems for allocating capital; marketing teams were still reconciling spreadsheets and conflicting measurements. Architect was Haus’s effort to build recommendations and action on top of causal evidence, while keeping marketers involved in the decisions.

The ambition invites a test of its own. Recommendations are useful only if the evidence, business context and constraints underneath them deserve trust. A cheaper campaign may be the wrong answer when a brand is launching a product or pursuing a longer-term goal. Epstein’s insistence on the quality of the underlying signal becomes more consequential as software moves closer to deciding where money goes.

On September 23, 2026, Haus announced Architect’s integration into Causal MMM. Product leader Greg Dale described the immediate goal as reducing the manual work of combining data and operating models. Further recommendations for pricing and promotions were presented as future work. The distinction matters: a roadmap expresses an intention; a working feature has to earn its place in someone’s day.

A life beyond the holdout

Epstein lives in the Bay Area with his wife and two daughters. He also plays golf. Those are modest biographical details, but they put an actual life around a founder’s usual inventory of products and financing rounds. There are other clocks running besides the company’s.

His public humor is similarly domestic in scale. Ahead of Cannes in 2026, he invited people to talk about marketing measurement or swap advice on keeping linen from wrinkling. The same person who gets exercised about statistical claims can acknowledge the less tractable problem of a creased shirt.

For all the machinery Haus has added, Epstein’s work keeps returning to a question a person can ask without a model: what did our decision change? His early academic work examined spending and election results. His entertainment career put him close to the contest for attention. Google gave him experience with experiments at scale. Haus is his effort to put that discipline into ordinary business practice.

There is something companionable about a skeptic who includes a joke with the objection. You can enjoy the campaign, admire the creative and appreciate a persuasive presentation. Epstein would still like someone to check what happened in the places where the ads did not run. Then the conversation can get interesting.