Reed's Law
If users can form their own groups inside your product, the value grows exponentially — and the groups become the moat, not the chat itself.
Why PMs should care
If users of your product can form their own groups — Slack channels, Facebook groups, Discord servers, WhatsApp group chats, GitHub organisations — the number of possible groups is 2^n. That's a much bigger number than Metcalfe's already-steep n², and it changes the picture completely.
This is why group-forming platforms have the stickiest moats in software. The switching cost doesn't grow with the number of users — it grows with the number of groups those users have formed, and the content living inside those groups.
For a PM, the strategic point is this: if you have a chance to ship 'users can make groups of their own here' as a feature, it's almost always under-prioritised compared to its long-term value. The short-term metrics look worse (groups split attention) while the long-term moat builds quietly.
Example in product work
Slack's channel moat. Slack's moat isn't that the 500 people at a mid-size company can message each other — that's the Metcalfe layer, and any competitor with a decent chat app could replicate it. The moat is that those 500 people have formed 1,200 channels: #eng-backend, #design-critique, #brussels-office-coffee-chat, #project-atlas-2024, #pets-of-slack. Each channel has its own membership, history, pinned docs, and norms. Replacing Slack means replacing those 1,200 mini-communities, not just the chat. This is why Slack can be technically worse than every competitor for years and still win — the 2^n group structure is the product, and the chat is just the substrate. Microsoft Teams didn't win on being better at chat; it won where it won because of pre-existing licensing leverage, not because it solved the group problem more cleanly.
Figma vs Sketch. Figma's multiplayer mode is the same idea in design tooling. Sketch had a better static design canvas for years. Figma let teams form arbitrary subgroups of shared files, live cursors, comment threads per frame, and design-system libraries shared across projects. Every team that adopted Figma didn't just adopt a tool — they spun up hundreds of shared files with permissions, comment histories, and linked components. The switching cost wasn't the software; it was the accumulated group structure inside it. Sketch couldn't be caught up to by shipping multiplayer later, because the moat was already compounded.
What to do when you see it
- Metcalfe's n² counts connections between individuals. Reed's 2^n counts possible groups, which is a much bigger number.
- When users can create their own Slack channels, Facebook groups, Discord servers, or GitHub orgs, the switching cost grows with the number of groups, not just users.
- Group-forming products have the stickiest moats in software — because users would be leaving their groups, not just their chat.
- 'Let users make groups here' is almost always under-prioritised relative to its long-term value.
Sources & further reading
- The Law of the Pack — David P. Reed, HBR, 2001Reed's own Harvard Business Review article introducing his law of group-forming networks.
- That Sneaky Exponential—Reed's Law — David P. ReedReed's original 1999 essay explaining why group-forming networks scale as 2^N.