Moat types
Network Effects: What They Are, the Main Types, and Real Examples
Updated 21 October 2026
A network effect is when a product becomes more valuable to each user as more people use it. A messaging app is useless with no friends on it and essential once everyone is. Network effects are one of the strongest moats in technology because a competitor cannot copy them by copying the product: it has to recreate the network too.
NFX estimates that network effects account for about 70% of the value created in technology since 1994. That number is an estimate, but the pattern behind it is easy to see. The largest consumer and fintech companies of the last three decades are mostly networks: social apps, marketplaces, payment rails and operating systems.
Why network effects matter for startups
A big company can defend itself with money, brand and distribution. A startup usually has none of those. What it can have is a network that gets better with every user and that a better-funded rival would have to rebuild from zero.
That is the point of the Moat Check question. Products are easy to copy, as we explain in what is a moat. A rival can rebuild your app in months. It cannot rebuild the fact that every buyer in a category is already talking to every seller on your platform.
For incumbents, network effects work in the other direction. They are the reason a large company can launch a weaker copy of a startup's product and still win: if the incumbent already owns the network (contacts, files, payments, developers), the startup's better features may not be enough. This is why the test pairs naturally with switching costs.
Types of network effects
NFX's network effects manual lists 16 types. For a founder, five distinctions cover most cases.
Direct network effects
Every user benefits directly from other users of the same kind. Phones, messaging apps and social networks work this way. The more of your contacts are on the product, the more useful it is. NFX lists WhatsApp, Facebook and LinkedIn as direct, identity-based networks, which are among the strongest because they are tied to who you are and who you know.
Two-sided (marketplace) network effects
There are two groups, and each attracts the other. More buyers bring more sellers; more sellers bring more buyers. eBay, Etsy, Airbnb and Uber are the textbook cases. These are powerful but often weaker than direct effects, because both sides can "multi-tenant," meaning use several competing marketplaces at the same time.
Platform network effects
Developers or suppliers build products specifically for a platform, and users choose the platform because of what has been built for it. iOS, Android, Windows and game consoles are examples. Because suppliers invest real engineering effort to integrate, they are slow to leave.
Data network effects
Usage generates data that makes the product better for every user. Waze is the classic case. NFX is direct about the limits here: data network effects "tend to be weaker than many people often want to believe," and more data does not automatically mean more value. Our separate data flywheel page covers how to tell a real one from a slide-deck one.
Local vs global network effects
This distinction matters more than most founders expect. In a global network, a user in New York benefits from a user in London. In a local network, they do not. Uber's network is mostly local: a rider in Chicago gets nothing from more drivers in Dallas. Andrew Chen describes Uber as "a vast collection of hundreds of hyperlocal marketplaces", each with its own network effects. That means a competitor can attack one city at a time.
Many local networks are also asymptotic. NFX classifies ride-hailing as an asymptotic marketplace: once pickup times are short enough, extra drivers add little value for riders. The network still helps, but the advantage stops growing.
Network effects vs virality
Virality and network effects are often confused because the biggest consumer apps had both. They are different things. NFX puts it simply: viral effects are when existing customers bring you new customers, while a network effect is when every customer adds value to all the other customers.
Virality is about acquisition. A network effect is about retention and defensibility. A quiz app can go viral and have no network effect: once the trend passes, users leave because nobody else on the app makes it better for them. A B2B marketplace can have a strong network effect and grow mainly through a sales team. If your growth comes from sharing but the product is just as good for a solo user, you have virality, not a moat.
Network effects examples
| Company | Type | How the network effect works |
|---|---|---|
| Direct | The app is valuable because the people you message are already on it. | |
| Direct | Your feed and groups depend on your friends and communities being there. | |
| Direct | Recruiters come because professionals are there, and the reverse. Microsoft paid $26.2 billion for it in 2016, when it had more than 433 million members. | |
| Slack | Direct (within and across companies) | Shared channels let separate companies collaborate in one workspace, so each connected company makes Slack more useful to its partners. |
| Figma | Direct (team) | Multiplayer editing lets designers, engineers and PMs work in one live file, so the tool is more useful as more of a team joins. |
| Microsoft Office | Direct (file formats and skills) | Documents are shared in Office formats, and NFX cites Excel as an expertise network effect: employers hire for the skill, so workers learn it. |
| eBay | Marketplace | More buyers attract more sellers, which attracts more buyers. |
| Etsy | Marketplace | Its marketplaces connected 7.5 million active sellers to 96.3 million active buyers at the end of 2021. |
| Airbnb | Marketplace | Guests turn into hosts: 36% of hosts in Q4 2022 had been guests first. |
| Uber | Local marketplace | In its S-1, Uber describes a "liquidity network effect" in each city where it operates. |
| Visa | Two-sided payments | More cardholders attract more merchants. Visa reported 4.6 billion cards and over 150 million merchant locations in fiscal 2024. |
| Waze | Data | Driving with the app open shares real-time speed and road data that improves routes for every other driver. |
| Stripe | Data | Radar fraud models are trained on data across the Stripe network, so each business benefits from fraud patterns seen at others. |
Notice that these effects differ a lot in strength. WhatsApp's is global and tied to identity. Uber's is local and asymptotic. Stripe's is a data effect layered on top of a product people already chose for other reasons.
How a startup builds a network effect from zero
- Pick the unit of value. Decide who must be on the product for it to work: a team, a buyer and seller in one category, two friends, drivers in one city.
- Win a small network first. Almost every durable network started by dominating a tight group (one campus, one city, one team) before expanding. This is the atomic network test.
- Make the product useful before the network exists. A design tool that is good for one designer, or a chat app that is good for one team, gives people a reason to show up early. A single-player hook gets users in so the network can form.
- Measure the effect directly. Track whether retention, match rate, time to fill or messages per user improves as a network gets denser. If the numbers do not move, the network effect is a hope, not a fact.
- Connect the networks. Once you win several local networks, look for ways users in one benefit from users in another (cross-company channels, cross-city travel, shared reviews).
Common mistakes and fake network effects
- Counting users instead of connections. A large user base where users never interact is scale, not a network effect.
- Calling virality a moat. Referrals bring users in; they do not stop users from leaving.
- Assuming the marketplace is global. If value is per city or per category, a rival only needs to win one segment at a time.
- Ignoring multi-tenanting. Drivers on two ride apps and sellers on three marketplaces cut the effect sharply.
- Overstating data effects. More data that does not measurably improve the product is just storage.
- "Network effects at scale." If the effect only appears after millions of users, it is not a moat today.
How Moat Check scores it
Moat Check asks one question for this test: "Does each new user make the product better for other users?" The reason it matters: a later entrant must recreate the whole network, not just the product. The test is worth 10 of the 100 points. See the methodology for the full framework.
- Pass (10 points): Users get more value as more people join, through a marketplace, messaging, collaboration or shared data, and this is true today or built into the product by design.
- Partial (5 points): The effect is plausible but weak, local or asymptotic, depends on users the product does not yet have, or only appears "at scale."
- Fail (0 points): Each user gets the same value regardless of how many others use it. Most single-user tools, including many AI apps, land here.
Moat Check uses AI and web research to apply this framework to your idea. It is a starting point for thinking, not investment advice. To see how network effects fit with the other nine tests, browse the moats hub.
Questions
What is a network effect in simple terms?
A network effect is when a product becomes more valuable to each user as more people use it. A phone network, a marketplace or a messaging app is worth more when the people you want to reach are already on it.
What are examples of network effects?
Common examples include WhatsApp and Facebook (more friends make the app more useful), Uber and Airbnb (more supply attracts more demand and the reverse), Visa (more cardholders attract more merchants), and Waze (more drivers produce better traffic data).
Is virality the same as a network effect?
No. Virality is how users bring in new users, which is a growth effect. A network effect is how each new user makes the product more valuable for existing users, which is a defensibility effect. A product can have one without the other.
What are the main types of network effects?
The main types are direct (users benefit from other users like them), two-sided or marketplace (buyers and sellers attract each other), platform (developers build for the users and users come for the apps), and data (usage produces data that improves the product). Each can be local or global.
Do all marketplaces have strong network effects?
No. Many marketplaces have local or asymptotic network effects, where value stops growing once there is enough supply in one city or category. Users and suppliers can also use several competing marketplaces at once, which weakens the effect.
Keep reading
Who will copy your startup?
Real competitors, the likely copycat and a moat score. Free, about a minute.