Headcount as Cost, Not Capability

"I could run that business from my house"

The day after he offered $125 a share for eBay, Ryan Cohen did the arithmetic out loud. eBay spends "$5.5 billion on operating expenses on an $11 billion business that has no inventory and is asset light," he said, and it carries "11,500 employees, and it doesn't make sense. I could run that business from my house."¹ The boast is doing less work in that sentence than the accounting is. His objection is not that eBay employs people. It is that nothing about a marketplace holding no inventory explains 11,500 of them.

Cohen's general rule is a claim about speed rather than cost, which is what separates it from ordinary cost-cutting: "the more people you add, the more you slow things down, and the fewer people you have, the more it's like a startup. You've got to always be in startup mode."¹ Headcount, read this way, is not a stock of capability a company draws down. It is friction the company pays to carry. He offers GameStop as his own demonstration, where SG&A came down 47 percent, some $800 million, largely by making marketing more efficient and "almost turning off marketing," and where he describes his method as "running the business like a family business. It's really not that complicated."¹

For evidence that a service survives losing most of the people who ran it, he points at the Twitter take-private. "The service kept working," he says, and the engineering teams are "much smaller and they're innovating a lot faster."¹ It is worth being exact about what he is claiming there, because he is careful about it himself. He argues that the damage to that business came from advertisers rather than from the cuts, and he flags the reading as his own: "My understanding is it really had nothing to do with the cost cuts."¹ That is one operator's interpretation of a contested episode, offered as an understanding rather than a finding, and it is the weakest link in his case.

The same discipline with a decade of P&L behind it

What Cohen's eBay argument lacks is a company that has actually been run this way, and Adam Foroughi supplies one. AppLovin's main business was throwing off roughly $1.3 billion of cash in a single quarter, a run rate better than $5 billion a year, on about 350 to 400 people, down from close to double that.² The reduction was not a response to trouble. He took roughly 40 percent of the team out across 2023 and 2024 while the business was growing about 100 percent year over year.²

Two mechanisms did the work. The first is structural refusal. AppLovin has no chief revenue officer and no chief operating officer; the executive team is a chief executive, a chief financial officer, a chief technology officer, and a general counsel, and the human resources function went from seventy or eighty people to fifteen.³ The standing test Foroughi applies to any role is that "every person in the organization should be doing something that creates direct value."³ The second is a piece of plumbing he found only after taking the HR function over himself: every departure automatically opened a backfill posting, so headcount could rise and could never fall. Once he made people convince him a role was needed, new hires fell from hundreds a year to tens, because only the genuinely desperate bothered to ask.²

His forward-looking version is the sharpest statement of the pattern on the record here: "With AI tools, one person who's really good can be as powerful as 10, 20, 50, 100 people."² The conclusion he draws from it is not to hire ahead of the work but to find those individuals, on the argument that hiring into an opportunity dilutes the average and leaves a tangle that cannot later be unwound.²

What the arithmetic does not show

Neither case is a business without employees, and both men are describing a ratio rather than a count. The 350 to 400 figure covers AppLovin's main business and excludes the game studios, more than 1,500 people distributed worldwide, that the company sold to Tripledot, so it is not a company-wide number and not a like-for-like measure of efficiency.² Foroughi's cut was also a re-sorting rather than a flat reduction: he moved lower-paid staff onto cash compensation and reserved equity for the critical product and engineering group, which is a decision about who the company is in it with for the long term, not simply about how many people it employs.²

Cohen's line about running eBay from his house remains a claim rather than a demonstration. He has not run eBay, and whether the value produced by an 11,500-person operation survives a cut to startup size is the question the pattern raises and does not answer. Both men narrate the cut; neither publishes the counterfactual. The honest reading is that they have proved something narrower and still useful, which is that the org chart a business has inherited is poor evidence for the org chart it needs.

The discipline is the negative image of the owner-operated growth lever, which asks what a founder must never hand off; this one asks what nobody needs to be hired to do at all. It also sits on top of the argument in owner-operator vs professional management, because both men trace the bloat to the same source. Cohen blames boards and executives who are "not operating like owners."¹ Foroughi credits AppLovin's escape from it to never having let professional managers empire-build.²