Network Effects: Why Some Products Get Stronger Every Time Someone Uses Them

On February 14, 1876, Alexander Graham Bell filed a patent for a device that transmitted sound through electrical wire. Six weeks later, on March 10, he spoke the first words ever carried by telephone: "Mr. Watson, come here. I want to see you." The technology worked. The product, by any measure, was a breakthrough. And for years, almost nobody wanted one.

The problem was simple and circular. A telephone is useless unless the person you want to talk to also has a telephone. In 1877, Bell's company had 778 phones in operation. To make a call, you needed to hope the one person you wanted to reach happened to be among those 778. The product wasn't failing on quality. It was failing on math. Theodore Vail, who would later build Bell into a monopoly, captured the paradox in his 1908 AT&T annual report: "A telephone, without a connection at the other end of the line, is not even a toy or a scientific instrument. Its value depends on the connection with the other telephone, and increases with the number of connections." Network effects, the force that eventually made the telephone indispensable, are the same force that nearly killed it at birth.

By 1880, there were over 130,000 telephones in the United States. By 1910, there were 5.8 million. Each new phone made every existing phone more valuable, because it added one more person you could reach. The product didn't improve. The network around it did. And the bigger the network grew, the harder it became for any competitor to offer a credible alternative.

What Are Network Effects and Why Do They Matter?

A network effect exists when a product or service becomes more valuable as more people use it. The telephone gets more useful as more people own telephones. A social platform gets more engaging as more of your friends join. A marketplace gets more attractive to buyers as more sellers list, and more attractive to sellers as more buyers browse.

Robert Metcalfe, who co-invented Ethernet in 1973, formalized this into what became known as Metcalfe's Law: the value of a network is proportional to the square of its number of users. Ten users create a hundred possible connections. A hundred users create ten thousand. A thousand users create a million. The growth in value is exponential while the growth in users is linear, which is why network-effect businesses tend to either dominate completely or fail entirely.

A 2015 study by Zhang, Liu, and Xu empirically validated Metcalfe's Law using data from Facebook and Tencent, finding that the square-of-users model fit observed network value better than competing formulas. The math isn't perfect for every network, but the principle holds: each new user doesn't just add their own value. They increase the value of every user already there.

This is what makes network effects the most powerful competitive moat in business. Brand equity can be matched over time. Proprietary technology can be leapfrogged. Speed of execution can be hired. But a network effect compounds. The bigger it gets, the harder it is to compete with, because the competitor doesn't just need a better product. They need to recreate the entire network.

How Craigslist Defeated Billion-Dollar Competitors With Almost No Effort

In 1995, Craig Newmark started an email list in San Francisco. He shared local events with friends. People started forwarding it. The list grew. Newmark added categories: housing, jobs, for sale. By 1996, it was a website. By 2000, it was in a dozen cities. He never ran an ad. He never raised venture capital. He employed roughly 50 people.

Craigslist destroyed the newspaper classified advertising industry. Between 2000 and 2012, newspaper classified revenue in the United States fell from approximately $19.6 billion to $4.6 billion. A 2014 study published in Management Science by Robert Seamans and Feng Zhu estimated that Craigslist's entry into local markets saved classified-ad buyers roughly $5 billion over the 2000 to 2007 period.

The mechanism was a network effect so powerful that it was functionally permanent. Sellers posted on Craigslist because that's where the buyers were. Buyers checked Craigslist because that's where the sellers were. Every new listing made the platform more valuable for buyers. Every new buyer made it more valuable for sellers. eBay spent billions trying to compete with Kijiji. Newspapers launched their own classified sites. None of them could overcome the network-effect moat that Craigslist had built accidentally, with no marketing budget, no sales team, and a website that looked like it was designed in 1997, because it was.

Craigslist still generates roughly $694 million in annual revenue. With 50 employees, that's approximately $14 million per employee, a number that makes most Silicon Valley startups look inefficient. The product didn't win on features. It won on network density.

The Data Network Effect: How Waze Got Smarter Every Time Someone Drove

Not all network effects work the same way. Craigslist runs on a two-sided marketplace effect: more buyers attract more sellers and vice versa. Waze, the navigation app founded in 2006 in Israel, runs on a different variant called a data network effect.

Every Waze user generates real-time location and speed data as they drive. That data feeds Waze's routing algorithm, which calculates the fastest path by analyzing actual traffic conditions across millions of simultaneous drivers. More drivers produce more data. More data produces better routes. Better routes attract more drivers. The loop compounds.

By 2013, Waze had roughly 50 million users. Google acquired it for $966 million. Google wasn't buying the code. Routing algorithms can be written by any competent engineering team. Google was buying the data network effect, the self-reinforcing loop that made Waze's product improve every time someone turned the app on. A competitor launching a rival navigation app would face a cold start: no drivers means no data, no data means bad routes, bad routes means no drivers.

When Network Effects Fail: The Homejoy Lesson

Network effects are not automatic. A platform that connects people does not necessarily generate a network effect, and the difference between platforms that do and platforms that don't is usually one structural flaw.

Homejoy, a house cleaning marketplace, launched in 2012 out of Y Combinator. The pitch was clean: connect homeowners with professional cleaners through an easy-to-use platform. The company raised approximately $40 million in venture capital and expanded to more than 30 cities. By any growth metric, it was working.

By July 2015, Homejoy was dead. Worker misclassification lawsuits prevented the next funding round, but the underlying problem was structural. Marketplace leakage had been draining the business from the start. Once a homeowner found a cleaner they liked through Homejoy, they exchanged phone numbers and booked directly. The platform had connected them, and then the connection made the platform unnecessary. Homejoy was paying to acquire both sides of a marketplace, and then watching them walk around it.

This is the test that separates real network effects from one-time matchmaking. A genuine network effect means the platform becomes more valuable to you the more people use it, permanently. Craigslist passes this test because you return to browse new listings from new sellers. Waze passes because every driver on the road improves your route in real time. Homejoy failed because the value transfer happened once, and then the network was no longer needed.

For entrepreneurs building platforms, the question isn't whether you can connect supply and demand. It's whether the connection creates ongoing value that keeps both sides on the platform, or whether it creates a relationship that makes the platform obsolete.

How Do You Build a Network Effect?

The hardest part of building a network-effect business is the beginning. Bell had 778 phones. Craigslist had an email list. Waze had a handful of drivers in Tel Aviv. Every network-effect business starts with a phase where the product is too small to be useful, and the only path forward is to get enough users into the system that the effect kicks in.

The strategies that work share a common principle: start small and dense. Craigslist started in one city. Facebook started at one university. Uber started in one neighborhood in San Francisco. Density matters more than breadth because a network effect requires that users encounter other users. A hundred users spread across fifty cities is a dead network. A hundred users in one zip code is a community.

The second principle is to deliver standalone value before the network effect materializes. Waze was a free GPS navigation app even with zero community data. Instagram was a photo editing tool even before your friends joined. The standalone value gets users in the door. The network effect keeps them from leaving.

The third principle is to make the contribution automatic. Waze users don't consciously share traffic data. They just drive, and the app collects. This is a data network effect at its most elegant: the act of using the product improves the product for everyone else, without requiring any additional effort from the user.