In early 2011, Jason Goldberg shut down Fabulis, a social network for gay men that he had launched about fourteen months earlier. By one report at the time it had more than 140,000 registered users. Goldberg told the same outlet that he did not believe a non-sexual social network for gay men could pass a million members, and that without ten million it was hard to build a significant business in that space.

That is an awkward place to start an article in favor of niches, because Fabulis was a niche. It was narrow in exactly the way startup advice tells you to be. PayPal, by contrast, began with a group of eBay power sellers that Peter Thiel has put at 20,000 to 25,000 people, and Facebook began on a single campus. Same advice, opposite outcomes.

So the useful question is narrower than "should you niche down?" It is what a niche has to contain before it deserves the name. My argument, which I will test against the evidence as we go, is that a niche is a shared situation rather than a small market. The groups that worked were stuck with the same problem, could be found, and talked to each other. The one that failed had a clear identity but no urgent problem in common.

What "everyone" costs you

Targeting everyone rarely fails because the market is too big. It fails because nothing ever gets ruled out. When your customer is a crowd, every choice about message, channel, first feature and price lacks a tiebreaker.

Bill Aulet's Disciplined Entrepreneurship gives a workable test. Customers in a real segment buy the same product, buy it in the same way, and influence one another. His site adds a practical check: if your sales approach has to change from one customer to the next, you have not narrowed enough.

Paul Graham gave the Airbnb founders related advice: focus on 100 people who love you instead of a million who mildly like you. Brian Chesky has called it the best advice he ever received. That is advice about depth more than size, but depth is easier to reach when the group is small enough to meet.

Article #5 argued that a price selects your audience. A niche does something similar earlier. It decides whose words you borrow, where you show up, and which complaints you hear first.

What PayPal and Facebook actually had

Look at how Thiel described PayPal's first market in a 2014 talk at Harvard Business School. The power sellers were people who needed something much better than checks, and he says PayPal reached 30 to 35 percent of them within three months. That is a recollection from an investor with every reason to tell it well, so treat the percentage with caution. Notice the shape of it, though: a concentrated group with a daily, costly annoyance. Inc. reports Thiel's account of an earlier plan, payments between PalmPilots, as too large and too spread out for the company to dominate.

Facebook's early record is easier to check because a student newspaper wrote it down. After about three weeks of Harvard-only access, the Crimson reported that 5,055 Harvard students had joined, and that the site had just opened to Columbia and Stanford, with Yale next. Early profile counts at Columbia and Stanford were 37 and 26. The new campuses started tiny. What carried over was that students see each other every day, so word of mouth came built into the setting, and the same product worked at each school. Those line up with Aulet's three conditions, though that comparison is mine, not the Crimson's.

Both are survivor stories, and PayPal's is told partly by the people who profited. They show what a working niche can look like. They do not show that narrowness caused the result.

The niche that wasn't enough

Back to Fabulis. Goldberg gave several explanations when it closed. He said progress on gay rights had reduced demand for a gay version of existing services, and that the one real success in the space, Grindr, worked because it met a pressing need his company chose not to serve. He also said he could not make the numbers work. Fab then reinvented itself as a design flash-sale site.

My reading, which Goldberg's account partly supports but does not prove, is that Fabulis defined its audience by who they were, while PayPal defined its audience by what they were stuck with. An identity is a good way to find people. It is not a reason for them to leave the networks they already use. The power sellers had a daily annoyance with a price tag attached. The Fabulis audience had alternatives and no urgent job.

It also shows the ceiling problem. A hundred and forty thousand registered users sounds substantial, and Goldberg still judged it far short of what the business needed. Whether a niche is too small depends on the business you are trying to build.

Four checks before you commit

These are my synthesis of Aulet's criteria and the Fabulis account. I have not seen evidence that they predict success, so read them as questions and not as a score.

1. Can you describe the group by a situation instead of a label? "Small business owners" is a label. "Sellers who reconcile payouts from three marketplaces every Friday" is a situation. A situation tells you what to say, where to say it, and what they already do about the problem.

2. Can you list a hundred of them by name, and do they know each other? Reachability and word of mouth are why a small group can be won quickly. If you cannot name them, you will spend your time searching instead of selling.

3. Do they already spend time or money managing the problem? A workaround, a spreadsheet, a part-time hire, a monthly fee for something worse. This is the difference between identity and urgency, and it is the check Fabulis seems to have failed.

4. Does the ceiling fit the business you want? Work backwards. Say 3,000 people in your niche are reachable, 10 percent buy, and each pays $29 a month. That is 300 customers and $104,400 a year (invented numbers, for illustration). A solo founder might be happy with that. A team that needs ten times as much has found a beachhead, not a business, and should ask early where the next ring of customers is. Goldberg's ten million came from the company he was trying to build, not from a universal threshold.

Finding the group, and testing it before you build

Most founders choose a niche at a whiteboard. A better starting point is evidence you already have. Article #3 laid out stages from reached to committed. Look at the people who got furthest and sort them by situation, not by demographics. If your three best conversations share a situation you did not plan for, that is your candidate.

Then try a test of my own, which I have not seen validated anywhere. Write one sentence of offer aimed at that situation. Show it to five people inside the situation and a few just outside it. If only the insiders lean in, your edge is real. If everyone reacts the same way, you have described a label, and the niche is not doing any work.

Finally, ask for something with a cost. Article #2 ranked evidence from compliments to payments, and Article #5 covered paid pilots. Five people in one situation agreeing to a paid pilot tests the niche and the demand at once. That is cheaper than building for months and learning that your niche was a label.

When to stay narrow and when to widen

Facebook widened within weeks, while Harvard was still signing up, but each step added a group that already behaved like the first. Thiel describes the pattern as moving gradually into related, slightly broader markets. The sequence seems to matter more than the speed.

I would widen when people outside the niche arrive unprompted and use the product the same way, when customers inside it refer others without being asked, and when the next group would buy the same product the same way, which is Aulet's test applied to the neighbor. I would stay when customers still need you to explain what the product is for, or when serving a second group would make you say two different things on one page.

The evidence for niche-first is thinner than the advice suggests. A 1994 study of 123 new ventures in the Strategic Management Journal found the highest sales growth among ventures pursuing broad strategies in high-growth industries. An older study of mature manufacturing firms found that broad-scope firms did better on average, though some narrow strategies worked. That is a different era and industry, so it is no verdict on software startups. And a 2025 study of 988 AI ventures found that academic founders built technology applicable to more industries, while its author noted that serving many niches takes resources. That measures how widely a technology applies, not who you sell to first, and the difference is worth keeping. Your technology can be general while your first customer is specific.

I could not find a controlled comparison of niche-first and broad-first startups. What exists is a handful of survivor stories, a few failures, and reasoning about limited resources.

So

A niche is a bet that a small, specific group will teach you faster and pay you sooner than a crowd would. It is not a guarantee, and it is not a permanent identity. Fabulis shows that narrow is not enough. The group needs a problem it is already trying to solve, people who can be reached, and enough of them to carry the business you want.

Pick the smallest group that is stuck with your problem, then let them tell you when to stop being a niche.

Your turn: If you narrowed your audience, or deliberately refused to, what was the first real sign that you had picked the right group, or the wrong one?