George Carollo Lost Product-Market Fit. Then He Went Looking for It Again.

In early 2022, George Carollo was walking home past the Empire State Building when the numbers began to look almost respectable. Dover, the recruiting company he had co-founded three years earlier, had reached roughly $15 million in annual recurring revenue. Net dollar retention stood at 110 percent. New leaders had arrived to run engineering and customer success. The scrappy service that began with three friends and a collection of no-code tools had become, by ordinary startup standards, a real business.

A month later, ordinary standards became useless. Amazon, Google, Meta and other technology companies began cutting jobs. Interest rates rose. Hiring slowed, then froze. The market filled with laid-off engineers and recruiters. Dover's original promise had been built for employers with the opposite problem: I need more candidates, quickly. Suddenly, candidates were abundant and urgency had evaporated.

Revenue eventually fell to $1.8 million. The arithmetic was simple enough to fit in a cell. Living inside it took two and a half years.

A founder trained to count what others overlook

Carollo came to software by way of systems that are considerably less fashionable and much harder to move. At Stanford, he studied economics and urban studies, then completed a master's degree in civil and environmental engineering. As a research assistant, he helped produce a 102-page study of public-private partnerships for infrastructure - a field occupied by procurement rules, toll roads, risk allocation and the sober question of who pays when a grand plan meets the physical world.

The work after Stanford kept the same analytical spine. At Parsons Brinckerhoff, he built financial models for infrastructure projects. At TrueBridge Capital Partners, he conducted analysis for the Forbes Midas List. At the fintech startup then known as LendingHome, he led strategy and finance. Long before he wrote about recruiter marketplaces, he was studying arrangements in which incentives, risk and reward rarely travel together by accident.

Recruiting gave those ideas a human cast. In 2016 he co-founded TopFunnel, software intended to help employers find and engage candidates. The company lasted two years. When his friends Max Kolysh and Anvisha Pai later invited him to work on another recruiting startup, Carollo did not leap theatrically at destiny. He hesitated. He had already learned the space once.

“After a couple months of working with them, I was like, well, it's kind of very good to work with two really good friends that you respect a lot.”George Carollo, speaking on Mixergy

The friendship mattered more than a thunderbolt idea. Carollo had known both founders for years. He and Kolysh had worked on side projects and invested small checks with a group of friends. Pai was part of the same Stanford orbit and married to Carollo's best friend from college. Dover began less like a corporate merger than a trusted group deciding to take a difficult problem personally.

The software company that began by being a service

In its first months, Dover's founders became recruiters. They ran the process manually, using Airtable and other lightweight tools before committing much engineering. The point was not to produce a handsome demo. It was to stand close enough to the work that its irritations could no longer hide.

They saw a hiring stack full of capable pieces and exhausted people. A job board might deliver 500 applicants, but someone still had to find the two plausible ones. A scheduler could arrange an interview, but only after someone had chosen whom to meet. Candidate sources, outreach, interviews and feedback lived in different places. Eighty percent of the founders' recruiting time, they concluded, was operational rather than strategic.

Dover automated that busywork. It helped source candidates, filter applicants, coordinate outreach and manage interviews. In 2019 the company joined Y Combinator. In July 2021 it raised a $20 million Series A led by Tiger Global, with Founders Fund, Abstract Ventures and YC participating. More than 100 customers were using the platform by then. The improvised service had acquired software, capital and momentum.

When the customer's problem disappears

Founders are told to listen to customers. They are less often told what to do when the customers stop needing the thing they loved last quarter. Dover tried to revive the old engine through self-serve pricing, per-candidate billing and six free micro-products. Each experiment was a sensible bridge back to the main product. None reached the other bank.

The hard part, Carollo has said, was not technical. It was emotional. Four years of code carried four years of decisions, effort and identity. Software may be an asset on a balance sheet, but in a founder's head it can become an argument: surely this much work must still be worth something.

“Sometimes you have to delete four years of code and start over.”George Carollo

Carollo eventually stepped away from daily operations to run a four-month experiment. No team. No internal meetings. He talked to customers and iterated quickly. The constraint returned him to Dover's earliest method: before building a system, perform enough of the work to discover where the real friction has gone.

The old obsession returns in a new form

Carollo had been writing about recruiter incentives since Dover's first year. In a 2019 essay called “The $30,000 Intro,” he unpacked the contingency model: a recruiter paid only upon a hire appears perfectly aligned with the client, yet may rationally spread attention across many companies, favor easier searches and send the same strong candidate into several processes. The invoice rewards the placement. It does not necessarily reward patience, fit or ownership.

The marketplace was a different answer to the same concern. Startups often need experienced recruiting help, but hiring comes in bursts and seldom justifies a permanent recruiter. Independent recruiters want durable relationships without pretending every client has identical needs. Dover now matches the two sides, provides recruiting software, and lets companies see pricing and performance. The company's Y Combinator page says more than 500 companies use its products; its marketplace, launched in mid-2024, grew to serve hundreds of early-stage technology companies.

By Carollo's account, Dover returned to profitability and 100 percent year-over-year growth. The comeback was not a restoration. The old product had organized recruiting operations. The new one organized trust between people.

A life arranged in spreadsheets, with room for cake

There is a temptation to turn Carollo's story into a hymn to relentless optimization. Even his private life offers suspiciously tidy material. He and Sachi Takahashi-Rial met in middle school in Sacramento, grew close through high school, conducted a Berkeley-Stanford romance, lived in San Francisco group houses and later moved to New York. When they decided to marry, there was no ceremonial proposal. There was a shared Google spreadsheet titled “Life Planning.”

Their wedding invitation styled the celebration as an investor offsite, promising insights, innovation, vision and, mercifully, cake. The three-day gathering included breakout sessions for basketball and cake decorating. One friend reasonably asked whether they were getting married or raising a venture fund.

It is charming because it reveals the limit of the spreadsheet as well as the instinct behind it. A model can clarify a decision. It cannot make the market obey, make a marriage meaningful, or make a candidate choose your company. At some point the cells end and judgment begins.

In early 2026, Kolysh stepped down as Dover's CEO and said Carollo had been the mastermind behind many of the company's innovations, especially the recruiter marketplace. He called his co-founder relentless and fast-moving. Carollo took the role with a smaller, profitable company whose second product-market fit had been earned in public.

He now writes frequently about fake applications, ghost jobs, AI screening, startup hiring and the changing work of founders. One recent reflection mourned a skill he had once considered a personal advantage: duct-taping Airtable, Google Sheets, Zapier and Framer into a quick MVP. AI coding tools, he observed, have made it easier to build the real version. Even his superpower had lost its market.

The response was characteristically unsentimental. Learn the new tool. Keep talking to customers. Move toward the work that remains. Carollo calls the period between products “the lost era.” His more useful observation is the sentence that follows: you can come out the other side.