---
title: "Network effects"
description: "The product gets better because other people use it: social graphs, communities, respondent pools, audience reach. The defining property is that value grows non-linearly with the number of participants, while an empty product is worthless — which is why the cold start is the hardest problem for the builder and the attacker alike."
canonical: https://moa.giglabo.com/moats/1/
locale: en
---

# Network effects

> Markdown twin of https://moa.giglabo.com/moats/1/ — the same content as the HTML page, a fraction of the bytes.
> Every page of this site has one: append `.md` to any URL, or `index.md` to a directory URL.
> Site structure and the full page list for agents: https://moa.giglabo.com/llms.txt

- Sheet 1 / 35

## Passport

- **Rock**: Human networks
- **Depth**: 4 · Mine
- **Time to dig**: 10+ years
- **Capital**: ◐–● · high
- **Solo**: ~ partly
- **AI**: ↑ AI-resistant
- **Rent**: ✗ not for sale

## Sample

- **Share of apps**: 8.2%
- **No-rate**: 67%
- **Median price**: $22.97

Figures from the canivibecodeit sample. No-rate is the share of apps carrying this tag that cannot be vibe-coded — a proxy for structural strength. Only the first thirteen mechanics were measured.

## Essence

The product gets better because other people use it: social graphs, communities, respondent pools, audience reach. The defining property is that value grows non-linearly with the number of participants, while an empty product is worthless — which is why the cold start is the hardest problem for the builder and the attacker alike.

## How it is built

### Facebook — density beats size

Facebook did not try to be a network "for everyone" from day one: it opened at Harvard alone, reached near-total coverage there within weeks, and only then went campus by campus. The logic: a network holding 90% of your immediate circle is worth more than one holding 1% of the world. Density inside a closed group creates value immediately, and the groups are stitched together afterwards. This is the canonical answer to the cold start — narrow the market until the network effect switches on quickly.

### Slack — a network at the scale of one organisation

Slack never needed a global graph: its network is a single company. One enthusiast dragged their own team onto the free tier, value appeared at five or ten participants, the product then spread sideways through the organisation, and at some point the company woke up already hooked. Freemium here is not a monetisation model but a seeding mechanism: the barrier to entry is zero and the network effect is local, so it is reached fast.

### Discord — a federation of small networks

Discord grew not as one large network but as millions of independent servers, each with its own community. It started with gamers — a niche in acute pain (TeamSpeak and Skype were dreadful) whose ready-made communities moved across whole. Every server is a self-sufficient cell of value; the global graph assembled itself as a side effect. The lesson: a network effect can be collected out of autonomous clusters, without ever solving "lure everyone at once".

## How it is bypassed

### TikTok against Instagram and Facebook — change the kind of graph

Facebook held an unbreakable social graph: leaving meant losing your friends. TikTok made the graph irrelevant — the feed is assembled by an algorithm reading behaviour, not subscriptions, so a new user gets the full value in the first minute without a single friend on the app. The incumbent's network effect was not defeated, it was devalued: the contest moved to a plane where it did not exist. Instagram had to chase (Reels) on someone else's terms.

### Zoom against Skype and WebEx — remove the requirement to be on the network

Skype was a classic network: you could only call people who had Skype and an account. Zoom dropped the requirement — the host needs an account, the guest joins from a browser link. The product became useful at a single meeting, and every call with outside participants turned into a demo for new users. The bypass was to make the value non-networked and the growth viral through the use case itself.

### Multi-tenanting — win the use case, not the user

People sit in several networks at once, and that is the structural weakness of the moat: the attacker does not have to take the user, only one of their use cases. Snapchat took private teenage messaging from Facebook, LinkedIn never had to displace it, Discord took "hang out on voice with your people". The incumbent notices the loss once the use case is already gone. In practice: attack the underserved case where the incumbent's large graph gives it no advantage.

## Verdict

A network moat is dug by density in a narrow place, and it is bypassed sideways rather than head-on: a different kind of graph, the removal of the network requirement, the capture of one use case out of many.

## Related

- HTML version of this page: https://moa.giglabo.com/moats/1/
- Русский: https://moa.giglabo.com/ru/moats/1/
- Next: https://moa.giglabo.com/moats/2/
- Atlas: https://moa.giglabo.com/
- Catalogue: https://moa.giglabo.com/moats/
- Calculator: https://moa.giglabo.com/calculator/
- Site map for agents: https://moa.giglabo.com/llms.txt