Metcalfe’s Law
In a network, adding participants can create value beyond the participants themselves, because each new member can also form new connections with the members already there, so the number of potentially valuable connections grows much faster than the number of people does.
What Is It?
Robert Metcalfe, inventor of Ethernet, proposed in the 1980s that a communications network’s value tracks the number of possible connections between its users, not just the number of users itself. With n people on a network, the number of possible pairwise connections grows roughly with n squared, so a phone network with two users has exactly one possible connection between them, while one with ten users has forty-five possible connections. Metcalfe’s formula, value proportional to n squared, treats every one of those possible connections as roughly equally valuable, which is best understood as a provocative heuristic rather than a literal valuation law: most possible connections in a real network are never used, and even among the ones that are, not all carry equal value. Economists including Andrew Odlyzko and Bob Briscoe have argued that counting every potential link as equally valuable substantially overstates real network value, and have proposed slower-growing alternatives, but the underlying shape they don’t dispute is that potential connectivity, not headcount, is what should drive how a network’s value is assessed, and that this potential connectivity grows faster than the user count does.
Why Does It Matter?
Organizations building anything with network effects, a professional community, a platform that gets more useful as more people use it, routinely underestimate how disproportionately weak early adoption feels relative to the value the same tool could eventually deliver at scale, because the number of potentially valuable connections grows faster than the user count does. This makes the early period of any network-effect product genuinely hard to evaluate fairly using raw usage numbers alone. But the lesson isn’t that user count itself is what matters, it’s that user count is a poor proxy for value in these systems, because not every possible connection between users is one that actually gets used or matters. An internal tool used by every employee doesn’t automatically become more valuable as headcount grows, more users can just as easily add noise, irrelevant channels, and coordination overhead instead of useful connections.
Matching feature quality doesn’t automatically match the connection value an established network-effect product has already built, since a new entrant with a technically superior product still has to overcome however much of that incumbent value is actually being used, though how large and how durable that gap is depends on the specific network’s structure, not on Metcalfe’s formula itself. Concepts like critical mass and tipping points are important companion ideas here, but they describe adoption dynamics more general than the n-squared relationship, worth holding separately rather than treating as something Metcalfe’s Law itself establishes.
What Changes Once You See It?
You stop evaluating a network-effect product only by how many participants it has and start asking how many useful connections that participation actually makes possible.
You start looking for density in the part of the network that actually needs to interact, rather than maximizing adoption indiscriminately, since a smaller network with high relevant connectivity can outperform a larger network whose members have little reason to connect with each other.
You also get more cautious about using raw user count as a proxy for network value, since Metcalfe’s n-squared formulation assumes potential connections are valuable in a way real networks rarely satisfy in full, which means a slow-looking early user count doesn’t automatically mean the underlying value is accumulating slowly, and a large user count doesn’t automatically mean it’s accumulating quickly either.
Common Misunderstandings
- It isn’t a reliable general-purpose valuation formula. The n-squared relationship counts every possible pairwise connection as roughly equally valuable, which real networks rarely satisfy, economists have proposed slower-growing alternatives for exactly this reason, treat it as a heuristic about the shape of network value, not an equation to plug numbers into.
- It isn’t the same as simple economies of scale, which are about unit cost declining as production volume increases. Metcalfe’s Law is about a network’s value potentially increasing disproportionately as its user base grows, a demand-side effect rather than a cost-side one.
- It doesn’t apply to products where users don’t meaningfully connect with or benefit from each other. A tool that each person uses in isolation, without any interaction or shared value between users, doesn’t generate the same network-value dynamics no matter how many people adopt it.
- It doesn’t map neatly onto every network structure. Metcalfe’s original formulation assumes simple pairwise connections, but marketplaces are often two-sided, social networks cluster locally, and some network effects can even be negative, congestion, spam, noise, so the specific n-squared shape is a starting intuition, not a description that fits every network-effect business.
- It isn’t a guarantee that any growing network will eventually become valuable. The accelerating value only materializes if the connections being formed are actually useful ones, a large network of low-quality or irrelevant connections doesn’t produce the same effect.
Diagnostic Question
Is the value of what we’re building coming primarily from the product itself or from the connections between the people using it, and if it’s the latter, which connections actually create value, and how many of them become possible or easier as the network grows?
Explore Further
Field Notes
- None yet.
Related Field Guide
Origin
Attributed to Robert Metcalfe in the 1980s, describing the potential connectivity value of Ethernet networks; the formal “Metcalfe’s Law” label and its broader popularization came later, notably through George Gilder’s 1993 Forbes writing. Economists including Andrew Odlyzko and Bob Briscoe have since argued the quadratic formula overstates real network value and proposed more conservative alternatives, treating it as a useful heuristic rather than a settled empirical law.