Learn it → Apply it → Save as Skill
When a network lets members form their own groups, its potential value can grow far faster than its size. Reed's Law is a reminder to ask whether a product's real value comes from connecting pairs of people, or from letting any subset of people organize around something together.
The Dr. Wallace's Favorite Mental Models library is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Gemini and other AI assistants apply the mental models you value.
A network built for one-to-one connections (think phone calls) grows roughly with the square of its members. A network built for letting any subset of members form a group (a forum, a community app, a shared workspace) has vastly more possible groups than pairs, so its potential value can climb much faster as it grows.
Important: 2ⁿ is a theoretical ceiling on possible groups, not a prediction of actual dollar value. Treat it as a direction, not a forecast.
Reed's Law is most useful when you are deciding whether to invest in features that let users form, name and manage their own groups, rather than just connect one-to-one or broadcast to everyone.
Reed's Law does not replace Metcalfe's Law. It tells you which growth curve actually applies to the product you are evaluating.
Does the product let members form subgroups (teams, channels, circles, shared docs), or mainly connect pairs, or mainly broadcast one-to-many?
Match the product to a growth pattern: broadcast (n), pairwise (n²), or group-forming (up to 2ⁿ possible groups).
Look at how many groups actually form and stay active, not the theoretical count of every possible subset.
Invest in group-forming features only where real groups will form and get real value from forming.
Open a use case to see what the model changes.
If customers would value creating their own channels, projects or shared views, group-forming features can compound engagement faster than adding more one-to-one messaging or sharing options.
A tool that lets any department spin up its own workspace can become load-bearing infrastructure quickly, because the number of possible team configurations grows much faster than headcount.
Sub-communities, topic channels and interest groups are what make a community sticky. Measure how many stay active, not how many could theoretically exist.
Weigh a pairwise feature (direct messaging) against a group-forming feature (shared group spaces) by asking which one more members will actually use to organize themselves.
A platform where users or agents can assemble their own combinations of tools, skills or sub-agents can gain value from the number of useful combinations that get formed, not just the number of tools available.
When a pitch deck cites exponential network value, ask which law actually describes the product's real usage pattern before accepting the growth story.
This is an illustrative example designed to teach the comparison, not a measured case study.
| Members (n) | Broadcast value (n) | Pairwise value (n²) | Group-forming ceiling (2ⁿ) |
|---|---|---|---|
| 10 | 10 | 100 | 1,024 |
| 100 | 100 | 10,000 | astronomically large |
| 1,000 | 1,000 | 1,000,000 | astronomically large |
Expand for the origin of the law and the real debate about how far it should be trusted.
David P. Reed, a computer scientist and former chief scientist at Lotus, described it in a widely circulated 1999 essay and a 2001 Harvard Business Review piece, arguing that networks which let members form groups ("group-forming networks") have a potential value tied to the number of possible subsets of members, which grows as 2 to the power of n, far outpacing the n-squared value Robert Metcalfe had associated with simple pairwise connections.
Sarnoff's Law describes one-to-many broadcast value growing roughly linearly with audience size (n). Metcalfe's Law describes pairwise connection value growing with the number of possible pairs (close to n squared). Reed's Law describes group-forming value growing with the number of possible subsets (2 to the power of n). Each law assumes a different kind of network; none of them is "more correct" in general, they describe different structures.
No, and this is the most important caveat. 2ⁿ counts every mathematically possible subset of members, including huge numbers of subsets that would never form or have no value if they did. Reed's own point was directional: that the ceiling on potential value rises faster than people assumed, not that realized value matches that ceiling. Economists Andrew Odlyzko and Benjamin Tilly published a widely cited 2005 critique arguing that for most real networks, including the internet, actual value grows closer to n log n, well below both Metcalfe's and Reed's curves, because attention, trust and coordination costs limit how many connections or groups people actually use.
Treat Reed's Law as a prompt to check which growth pattern fits your product's real behavior, and as a caution against both underestimating group-forming features and over-trusting an exponential-sounding pitch. The useful output is a better question, not a precise valuation formula.
The exponential count is a ceiling on possible groups, not a dollar figure or a usage forecast.
Metcalfe's Law is about pairwise connections. Reed's Law is specifically about group formation. A product can fit one and not the other.
More possible groups also means more spaces to moderate, more fragmentation risk, and more coordination overhead that can offset raw growth.
A product with no real group-forming behavior gets no benefit from citing this law. Check actual usage patterns first.
"Exponential network effects" is sometimes used as a pitch-deck shortcut. Ask what specific group-forming behavior is actually driving the claim.
Humans can only meaningfully track and participate in a limited number of active groups at once, which caps realized value well below the mathematical ceiling.
the product is mainly one-to-many broadcast, like a newsletter or a traditional media channel, with little peer-to-peer or group structure.
value genuinely comes from pairwise connections, like a phone network, a simple marketplace, or direct messaging without group features.
the product is too new or too small for any network-effect law to be more than a guess; measure actual retention and engagement first.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply Reed's Law, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
With n members, the number of possible subsets (groups of any size, including the empty set) is 2ⁿ. Reed used this count as a proxy for the number of ways members could organize into groups, arguing it was a more realistic upper bound on a group-forming network's potential than counting only pairs.
No. The ceiling tells you the model's upper bound, not what real users will do. A product decision should rest on evidence that users want to form groups and get value from doing so, not on the size of the mathematical ceiling alone.
Platforms that let third parties combine tools, agents or plugins into custom bundles are, in effect, group-forming networks for capabilities rather than people. The same caution applies: count the combinations that actually get used and deliver value, not every combination that is technically possible.