Employee count is the easiest number to find about a company and one of the least informative. The US Census Bureau counted 30,427,808 nonemployer establishments in 2023, up from 29,811,495 the year before, and they generated close to $1.8 trillion — about 6.4% of GDP that year.
None of that shows up in a headcount table. The same distortion now runs through the other direction: a company with 25 people on payroll can operate infrastructure that would have needed several hundred a decade ago, because most of it is rented.
The two numbers rarely sit side by side
The Census figures make the scale of the gap plain. In 2023 there were 5.58 million US firms with at least one employee but fewer than 500 — and more than five times as many businesses with no employees at all.
Revenue per employee, the metric usually reached for next, inherits the problem. It rises when a company hires well and it rises when a company simply stops doing things in-house.
A payments team replaced by a processor, a support desk replaced by an outsourced tier one, and a data center replaced by three cloud accounts all move the ratio the same direction. The number goes up either way, which makes it useless for the comparison people want to make.
What actually left the payroll?
The functions that migrated out are consistent across industries:
- payment acceptance, settlement, and chargeback handling;
- identity verification and fraud screening;
- hosting, monitoring, and incident response;
- reporting and reconciliation tooling.
Labor markets shifted alongside this. As Maker Stations documented in its gig economy statistics, the Federal Reserve’s October 2024 survey found 20% of US adults doing some form of gig activity, with 70% of them spending under five hours a week on it. Work is being bought in smaller and smaller units, and so is capability.
Regulated sectors make the pattern legible
Where licensing exists, the split between headcount and capability is documented rather than inferred, because a regulator writes down who is answerable for what.
Take online gambling, which is licensed in most of the markets it operates in. The UK’s Gambling Commission assesses licence applicants on identity and ownership, finances, integrity, competence and criminal record, and requires a UK correspondence address maintained for the life of the licence.
Nothing in that list scales with staff numbers. A twelve-person company faces the same test as a listed one.
So operators split the work. Brand, marketing, and player relationships stay in-house; the platform underneath is bought.
A supplier such as the Agreegain white label platform provides the game library, payment integrations, and back-office reporting, which is why a company that looks small on LinkedIn can run a consumer product with tens of thousands of accounts.
The same structure appears in regulated fintech, telehealth, and insurance distribution, where the licensed entity is often a fraction of the size of the stack it operates.
Buying capability relocates scrutiny rather than removing it
This is the part that headcount analysis misses in the other direction. Outsourcing a function does not put it outside the regulatory perimeter — it moves the examination to the supplier.
New Jersey is explicit about this. Companies providing goods or services tied to internet gaming, including payment processing, identity checks, age verification, and geolocation, must be licensed as casino service industry enterprises by the Division of Gaming Enforcement, which publishes the list of applicants that have cleared preliminary review. A small operator’s compliance posture is therefore partly a description of its vendor list.
Anyone reading company statistics for signal should treat a low headcount in a regulated sector as a question rather than an answer: which functions were bought, and from whom.
A better set of numbers to ask for
Headcount survives because it is public and comparable. Vendor concentration, contract count, and the share of revenue-critical functions running on third-party systems are none of those things, which is why they rarely appear in a statistics roundup — and why the roundups keep describing companies that no longer exist in the shape they imply.
The practical version of this for anyone building: decide which functions have to be inspectable by an outsider, and staff or source those deliberately.
Everything else can sit wherever it is cheapest to run. That decision shapes what a company can actually do far more than the number of names on the payroll, and it is the number nobody publishes.
