Steve Yi and the value of a second try

Steve Yi once built a search-toolbar company with four business people and a problem that no amount of business expertise could solve: Google wanted the same market. Google could pay distribution partners more. Google could reach more people. The startup had a business model, an idea, and rather too many people qualified to discuss both. Yi later called the arrangement “four chefs.” The kitchen did not prosper.

That failure belongs near the beginning of his story because the successful company came afterward, with a different team and a less extravagant ambition. Before MediaAlpha became a publicly traded advertising marketplace, Yi and his partners wanted to earn enough to pay themselves. A sensible objective. Also a surprisingly unfashionable one in a business where losing money can acquire the social standing of a graduate degree.

The respectable exits were still open

Yi was born in Seoul. His family moved to Minneapolis when he was seven, and Minnesota became the place he considered home. He studied East Asian Studies at Harvard, then entered management consulting at Mercer. “I opened my first spreadsheet working in management consulting,” he recalled. It is a useful detail for someone whose later advice would put considerable faith in arithmetic.

Law school followed. He had considered working on social justice issues and saw a law degree as a route that could also take him back into business. By graduation in 1999, he wanted to build a company. Goldman Sachs offered a starting point; a year later he left for an internet company in Los Angeles.

The timing was unfortunate. The technology crash arrived within months. By 2002, that company had sold for much less than the founders had once imagined. Colleagues returned to consulting, banking, and law. Yi stayed in internet businesses. Los Angeles had an advertising industry worth learning, and he had decided that building something was worth another attempt.

A better use for a failed company

The toolbar venture ran from 2004 to 2007. Its problems included a founding team whose skills overlapped too much. Four people could see the commercial opportunity; that did not give them four different ways to build it. The experience made complementary abilities a practical requirement for his next partnership.

At Oversee.net, where he became senior vice president and general manager of marketing services, Yi regained his financial footing and learned more about search marketing and lead generation. A salaried job supplied something the entrepreneurial mythology tends to overlook: room to recover. He could use the skills a failed venture had taught him without immediately financing another lesson.

Then came Fareloop, a travel comparison business he co-founded and led from 2009 to 2011. Its sale was modest, but useful. It gave the next venture breathing room. The size of the exit mattered less than what the team could do with the proceeds.

First find the people. Then find the idea.

Yi’s next company began with a team before it had a settled purpose. He and two former colleagues wanted to work together again. Their abilities complemented one another, and the travel business had supplied enough capital for six to nine months of concentrated work without salaries becoming the immediate problem.

He went to conferences, met advertising contacts, spoke with venture capitalists in Los Angeles, and tested possibilities with anyone willing to listen. This was an unusually candid version of company formation: the founders knew one another and knew their skills, while the idea remained under discussion.

Auto insurance eventually offered a manageable starting point. The industry spent money on advertising. The team understood how to attract visitors, design a shopping experience, and match shoppers with advertisers. Travel had already given them practice with comparison. They could apply familiar tools to a new market and learn its particular frustrations by working inside it.

Their initial ambitions were almost domestic in scale: generate income, pay themselves, perhaps establish a comfortable business. Yi recalled thinking, “Let’s make some money because our money is about to run out.” The eventual public company began with a deadline more persuasive than a pitch deck.

Twenty dollars out. Two or three back.

The early arithmetic was unpleasant. Buying a visitor through Google could cost $20; the business might initially make only $2 or $3 back. Yi remembered a first month in which the numbers seemed impossible. He also remembered reaching profitability in the second month. Between those points came experiments with keywords, the shopping experience, advertiser relationships, and pricing.

He resisted reducing the turnaround to one clever intervention. A smoother conversion process helped. Giving advertisers more control helped. So did the accumulated experience of people who had done similar work before. His explanation was “the execution of a hundred things.” That is an inconvenient answer for anyone shopping for a single entrepreneurial trick.

The discipline was to improve enough small parts for the whole transaction to work. A visitor had to cost an amount the resulting advertising revenue could support. The founders were learning about insurance while keeping the business alive, which gave every experiment a rather immediate audience: their own bank balance.

“Trust the math.”

Steve Yi, on advice to technology startups, 2021

Why let a competitor into your shop?

The more interesting business emerged from a question that sounds commercially impolite. Why would an insurer show another insurer’s advertisement on its own website?

Because shoppers compare. A visitor who requests a quote may be unlikely to buy the policy on offer. The insurer has already paid to attract that visitor. An advertisement for an alternative carrier can produce revenue from a shopping journey that might otherwise end with a closed browser tab.

Esurance became a turning point. Its existing advertising arrangement was based on spreadsheets, and it wanted a more dynamic system. MediaAlpha was already considering an exchange. The conversation gave that idea an early outside publisher around which to build.

That changed the job. The founders could supply infrastructure to other businesses, extending beyond the visitors they attracted to their own site. Buyers could distinguish between publishers and consumer segments, then decide what each opportunity was worth. Publishers could make their inventory visible. Yi’s company had found a problem whose solution could be reused.

The person who will probably leave

Yi later explained the apparent contradiction through predictive analytics. A carrier can use historical information to estimate which shoppers are unlikely to buy. Those visitors may be appropriate candidates for alternative offers. A shopper likely to purchase the carrier’s own policy presents a different decision.

The distinction gives an unglamorous commercial activity some intellectual interest. A business can learn as much from the customer it probably will lose as from the one it expects to win. The departing visitor becomes part of the economics rather than simply an unsuccessful sale.

Yi also stressed that a model needs people to trust it. Good data and a sound prediction accomplish little if the team refuses to act. His prescription included bringing relevant colleagues into the process. Selling software, in this account, involves helping an organization become comfortable with decisions it previously made another way.

The humanities turn up at the auction

For all the emphasis on data, Yi has defended the usefulness of a liberal arts education in technology. His East Asian Studies degree required him to move between literature, politics, and economics. Running a company required a comparable movement between sales, technology, and marketing.

He saw value in being able to hold several perspectives at once, especially when a decision had no single correct answer. Technical accuracy mattered, but so did understanding people’s motives and the other forces operating around a product. An advertising exchange, after all, has to work for organizations as well as algorithms.

That is a revealing connection in his career. The student who moved between disciplines became an executive selling greater visibility between parties. Both jobs demand translation. A carrier wants profitable customers. A publisher wants revenue. An engineer needs a system that can express those preferences. Yi’s work sits where those different descriptions of the same transaction meet.

Growing without a hurry to announce the ending

Self-funding gave the founders a particular relationship with growth. Yi said they wanted the business to support itself rather than raise a large sum and accelerate immediately. That pace suited the traditional insurers they hoped to serve. Their customers had reasons to move carefully; matching that rhythm could be commercially useful.

Outside ownership did arrive. White Mountains bought a majority stake, giving the founders liquidity, and MediaAlpha completed its IPO in October 2020. Yet years earlier, when asked about the eventual exit, Yi had been open to several outcomes. Continuing to operate, selling, and going public were all possibilities. Growing the company came first.

There is a certain freedom in declining to give the ending a date. It leaves the next customer, the next improvement, and the next workable transaction with something useful to do.

The next calculation

By the second quarter of 2026, the business Yi led reported $316.9 million in revenue and $41.8 million in net income. His account of the opportunity ahead centered on more carriers putting more advertising through the marketplace. He also expected AI to support more direct purchases and more precise targeting, describing possibilities whose eventual scale still depended on adoption.

The organization continued to change around him. Tigran Sinanyan returned to the chief financial officer role effective October 1, 2026. The next earnings release was scheduled for October 28. Public-company life supplies its own regular appointments with arithmetic.

Yi’s path makes those appointments more interesting than the figures alone suggest. He has been the founder with too many chefs, the employee rebuilding his footing, and the partner looking for an idea before the runway expired. The enterprise that followed grew out of those ordinary constraints. His advice has remained fittingly compact: earn the customer’s trust, and let the results survive inspection.