Betting on the small company

Earlier this year I started building a few small services instead of hunting for one big idea. Unovote, Duogenda, Trefine. Three products, one person, no plan to raise money or hire a team for any of them. That is a bet on where software is going, and this is the reasoning behind it.

The decoupling

The link between headcount and output has come apart. In 2021 a SaaS company at roughly $250,000 of revenue per employee was considered efficient. By 2026 the median across private SaaS has climbed past $190,000 while headcount in every revenue band has fallen, and AI-native startups are reporting closer to $3 million per employee with teams about 40 percent smaller than their peers.

Midjourney reached around $200 million in annual revenue with roughly 40 people. More than fifty AI-native companies are expected to cross $10 million in revenue this year with fewer than ten employees. Sam Altman has a standing bet with other founders about the year the first one-person billion-dollar company shows up, and the people who would know think it is close.

New business formation says the same thing from the other end. Americans filed around 5.9 million business applications in 2025, up eight percent on the year before and roughly a third above the pre-pandemic rate. Applications from one-person firms rose more than twenty percent in a year. The tech layoffs of the last three years were not only cost-cutting. A good share of them were companies deciding that a 500-person org could be a 50-person org.

So the premise holds. You do not need a large company to build something people pay for. Fifty people can. Sometimes five. Sometimes one.

Priced for a company you are not

The trouble starts when you try to operate one of these companies. The software you run it on is priced for the company you are not. The performance and 1:1 tools in Duogenda's category will sell you a seat for around $11 a month and then attach a $4,000 minimum to the annual contract. The product-feedback platforms Unovote sits next to mostly do not publish a price at all: you fill in a contact-sales form and, if you are small, you get a five-figure annual quote for software three people would open twice a week. Seat minimums that start at 25. Single sign-on locked to the enterprise tier. None of it shrinks when you are three people. It just takes a bigger bite, because most of the cost is fixed and the vendor set the floor with someone else in mind.

That gap is the interesting part. The 5-to-50 band is going to be a much larger slice of the economy than it is now, and the tooling market still treats small as a phase you either grow out of or die in. Plenty of these companies are going to stay small on purpose and be very profitable doing it. Someone builds the operational layer for that size, priced honestly, and it is a real market rather than a cheaper tier of an enterprise product.

What has to change

Some of this is already moving. Per-seat pricing as a primary model dropped from 21 to 15 percent of SaaS companies in a single year, and usage-based plans are becoming the default. But usage-based often just means the small company pays less this month and gets a surprise next month. It is not a product designed for a small permanent team.

The big platforms are a separate question. Microsoft and Google built enterprise-grade toolchains on enterprise budgets, and small companies have had a cheap ride on that for a decade: the single sign-on, the audit trails, the security work that someone else paid to build, folded into a modest per-seat price. As the small-company segment turns into the larger market, that arrangement inverts. The same per-seat model starts asking a ten-person company to cover its share of an investment scoped for an organisation a hundred times its size. Fair enough, someone will say, they rode on it cheaply for years. Maybe. It still means the pricing has to change: less per head, more per company, more shaped to the customer in front of them.

I do not know how fast this happens or which categories move first. The signal I am watching for is a pricing page with a plan that assumes you will never call sales, sized for a company that means to stay ten people, sold without an asterisk.