Network Effects
You are trying to determine whether your product has a defensible moat — or just a head start. Network effects are the most powerful source of competitive defensibility in technology, but they are also the most misunderstood. Most products that claim network effects do not actually have them. The test is simple: does each new user make the product more valuable for every existing user? If yes, you have a network effect. If the product just gets more popular without becoming functionally better for existing users, you have scale — not a network effect.
The Core Idea
A network effect exists when the value of a product increases as more people use it. The telephone is the canonical example: one telephone is useless, two are marginally useful, and a network of millions is indispensable. The value scales non-linearly with the number of participants.
Hamilton Helmer, author of 7 Powers, draws a critical distinction that most founders miss: the difference between network effects and network economies. As he puts it, “There are lots of things that I would say have network effects but not network economies.” The difference is materiality — “whether the value benefit is large enough to engender a price delta significant enough to give you materially different margins into the future.” A network effect that does not clear this significance hurdle is not a source of power. Oji Udezue offers the clearest definition of the mechanism itself: “Network effects is when you create value for passive members by other people joining the network. I am by myself, I have done nothing. I’m at home, chilling, but one person joins the network and immediately I gain benefit.”
The Network Effects Taxonomy
| Type | Mechanism | Strength | Example |
|---|---|---|---|
| Direct | Same-side users increase value for each other | Strongest | WhatsApp, Slack (within a team) |
| Cross-side (marketplace) | More supply attracts more demand and vice versa | Strong | Airbnb, Uber, DoorDash |
| Data | More users generate more data, improving the product | Moderate | Google Search, Waze, Netflix recommendations |
| Platform/API | More developers build on the platform, attracting more users | Moderate-Strong | iOS, Salesforce, Shopify |
| Social | User-generated content or social graph increases value | Moderate | Instagram, TikTok, LinkedIn |
| Expertise | Community knowledge accumulates | Moderate | Stack Overflow, Reddit |
The distinction between direct and cross-side network effects is critical for strategy. Direct network effects (messaging, social) create winner-take-most dynamics because the product with the largest network is objectively better. Cross-side network effects (marketplaces) can support multiple winners because geographic, category, or quality segmentation allows parallel networks to coexist.
When Network Effects Actually Kick In
Yuriy Timen, who led growth at Grammarly and has advised Canva, Airtable, and others, is blunt about the manufacturing problem: “The first thing you look for is, is there inherent product network effects? It’s something that it’s either there, or isn’t from inception from my experience. It’s very hard to manufacture product network effects if they aren’t there from the get-go.” The products that have them must still solve the cold start problem — building from zero to a self-sustaining network.
The Cold Start Problem
Every networked product faces the same chicken-and-egg problem: the product is not valuable until it has users, but users will not join until it is valuable. Chen describes the progression:
-
Atomic Network — The smallest unit of the network that can sustain itself. For Slack, this is a single team. For Uber, this is a single city with enough drivers to provide acceptable wait times. For Tinder, this is enough users in a geographic area to produce matches.
-
Tipping Point — The moment when network effects become self-reinforcing. Before this point, growth requires investment (subsidies, manual effort, content seeding). After this point, the network grows organically.
-
Network Saturation — The network reaches a size where adding new users provides diminishing marginal value. Growth slows not from competition but from saturation.
The actionable insight: focus on winning one atomic network completely before expanding. Airbnb started with one neighborhood in San Francisco. Facebook started at Harvard. Uber started in San Francisco. The strategy is: go small, go deep, then expand.
The Traction Threshold
Sarah Tavel, partner at Benchmark and former first PM at Pinterest, provides a framework for evaluating when network effects are actually contributing to defensibility. Tavel has “an allergic reaction to vanity metrics” and argues that what matters is not user counts but whether users are completing the “core action” — the single behavior that proves they understand the product’s value. For Pinterest, it was pinning. For Facebook, friending. “If you’re not doing that action, you’re not really a user of the product. That’s why the MAU thing doesn’t really mean anything.”
| Signal | What It Means | How to Measure |
|---|---|---|
| Retention improves with network size | Users in larger networks stay longer | Cohort analysis by network size |
| Organic acquisition percentage grows | Less paid spend needed per user over time | % organic vs paid acquisition |
| Value delivered increases with usage | Power users get exponentially more value | Engagement metrics by user tenure |
| Switching costs compound | Users accumulate investment that is hard to move | Data, connections, content locked in |
Measuring Network Effect Strength
The presence of a network effect is not binary — it exists on a spectrum. Key dimensions to measure:
Hamilton Helmer warns against confusing the existence of a flywheel with the strength of one: “We laugh whenever we hear somebody say they have a flywheel, which gives you the idea of network economies. There are often flywheels. The ones that really are material are rare. The key thing here is materiality, not whether the flywheel exists, but whether the effect is strong enough to really tilt returns.”
Network density: What percentage of the possible connections in your network are actually active? A dense network (where users frequently interact with many other users) produces stronger effects than a sparse one.
Engagement curves by network size: Does user engagement increase as the network around them grows? Plot engagement metrics against network size. If the curve flattens early, the network effect is weak. If it continues to climb, the effect is strong.
Marginal user value: Does each additional user create meaningful value? In a messaging app, the 101st connection matters less than the 5th. The question is whether the value curve is steep enough to drive adoption past the cold start.
Network Effects vs. Scale Effects
The most common confusion in startup strategy. Not the same thing.
| Feature | Network Effect | Scale Effect |
|---|---|---|
| Value driver | More users = better product | More volume = lower costs |
| Defensibility | Very high | Moderate |
| Who benefits | Every user | The company (through margin) |
| Example | Amazon fulfillment | |
| Vulnerability | Multi-homing | Competitor with capital |
Amazon has scale effects (their warehousing and logistics become cheaper at volume), but the Amazon marketplace has network effects (more sellers attract more buyers attract more sellers). The marketplace is more defensible than the logistics.
Network Effects and Defensibility Over Time
Not all network effects persist. Several dynamics can erode them:
Multi-homing: When users can easily participate in multiple competing networks simultaneously (Uber and Lyft, DoorDash and Uber Eats), network effects provide less lock-in. The user does not have to choose one network; they can use all of them.
Disintermediation: When a marketplace’s supply and demand develop direct relationships that bypass the platform (a guest finds an Airbnb host on Instagram and books direct), the network effect weakens.
Clustering: When a large network fragments into local clusters that do not interact with each other, the whole-network effect weakens. A social network where different demographic groups never interact has weaker effects than one with cross-group interaction.
Building Network Effects Deliberately
Network effects are not discovered — they are designed. The product decisions that enable or prevent network effects:
Multiplayer by default: Products where collaboration or sharing is the primary use case (Figma, Slack, Notion shared workspaces) build network effects into the core experience. Products where sharing is an add-on (most productivity tools) do not.
User-generated content: Platforms where users create the value for other users (TikTok, YouTube, Stack Overflow) develop content network effects. The more creators, the more content, the more viewers, the more creators.
Invitations and referrals as features, not promotions: When inviting others makes the product better for the inviter (Slack — need teammates to communicate; Figma — need collaborators to design), invitations are organic. When invitations are incentivized with credits or discounts (Uber referral codes), they are promotional — growth without network effects.
Key Takeaway
- Network effects exist when each new user makes the product more valuable for existing users. Popularity is not a network effect.
- The type of network effect determines its defensibility. Direct effects (messaging) are strongest; data effects are moderate.
- Solve the cold start problem by winning one atomic network completely before expanding. Go small, go deep.
- Multi-homing is the primary threat to network-effect-based moats. If users can easily use competing products simultaneously, the lock-in is weaker.
- Measure network effect strength through engagement curves by network size, not just total user count.
Related
- Product-Led Growth — PLG companies often rely on network effects for acquisition
- Growth Loops — Network effects power viral and content growth loops
- Word of Mouth — Organic sharing that can bootstrap network formation
- Competitive Moats — Network effects as one of the 7 Powers
- Marketplace Dynamics — Cross-side network effects in action
Sources
- Hamilton Helmer on 7 Powers — Network economies vs network effects, materiality test, flywheel skepticism
- Oji Udezue on virality and network effects — Network effects definition, synthetic vs organic virality, Slack as case study
- Yuriy Timen on subscription growth — Inherent network effects, manufacturing impossibility, virality loops
- Sarah Tavel on the hierarchy of engagement — Core action framework, vanity metrics, Pinterest engagement model
- Dan Hockenmaier on marketplace growth — Winner-take-all dynamics, marketplace network effects