A hundred people who all know each other beat ten thousand who don't.

If your product gets better as more people use it — a messaging app, a marketplace, a team tool, a social anything — there’s a number you’ll be tempted to chase, and it’s usually the wrong one. Total users. It’s the number that looks good on a slide, so it’s the number the whole org orients around. And for a networked product it can climb steadily while the thing quietly fails to get more useful to anyone.

The reason is that network value doesn’t come from how many people are in the product. It comes from how many people you care about are in it. A messaging app with ten million users is worthless to you if none of your six closest people are on it, and precious if all six are, even if it has no one else. The unit that matters isn’t headcount. It’s density within the small world each user actually lives in.

The honest version of the network-effects story

The famous line — a network’s value grows with the square of its users — is more of a rallying cry than a law, and it’s been argued down hard. The square assumes everyone values a connection to everyone else equally, which is plainly false; most of the possible connections in any large network are worth nothing to you. The more careful critiques suggest something tamer than n² and, for our purposes, something more useful: value is lumpy. It concentrates in clusters. A connection inside your cluster is worth a great deal; a connection to a random stranger three clusters over is worth roughly nothing.

Which means the same thousand new users produce wildly different value depending on where they land. Spread one each across a thousand different friend groups, cities, or companies, and almost no one crosses the threshold where the product becomes useful — a thousand lonely first-arrivers, each the only person they know here, each likely to leave. Put all thousand inside one friend group, campus, or company, and that one world tips: now it’s the place everyone there already is, and it holds.

What this changes about how you grow

It reframes the goal. You are not trying to raise a global user count. You are trying to get one small world after another across its own tipping point, then move to the next. This is why the products that look like they conquered the world almost always started absurdly narrow — one university, one city’s drivers, one niche community — and why a broad, shallow launch across everyone at once so often dies: it spends its users as a thin film instead of pooling them deep enough to ignite anything.

Two practical consequences. First, your metric should be density inside a defined world — what fraction of a team, a campus, a category already uses it — not the global total. A chart that’s up and to the right on total users while every individual cluster stays below its threshold is a chart of a slow death. Second, build a single-player mode that’s worth it alone. The cruel part of needing density is that the first person in any new world has none — so they need a reason to stay while they wait for the second and third. The products that cross threshold after threshold usually give the lone arrival real standalone value, so the empty network is a promise rather than a disappointment.

So before you celebrate the headcount, ask the quieter question: of the people already here, how many are in a room together? That’s the number the product actually runs on.