Back to all topics
economics

Metcalfe vs Zipf

How network value really scales—not quadratically like you think.

Tiny Summary

Metcalfe's Law (n²) overstates network value. Zipf's Law (n·log n) is more accurate—connections aren't equally valuable, so growth is logarithmic, not quadratic.


Metcalfe's Law

Formula: V = n²

Claim: Network value grows with square of users

100 users → Value = 10,000
1,000 users → Value = 1,000,000 (100x!)

The Problem

Assumes all connections equally valuable:

  • 1st friend: Very valuable
  • 10th friend: Still valuable
  • 100th friend: Somewhat valuable
  • 1,000th friend: Barely know them

Reality: Diminishing returns, not all connections matter, active ≠ total users


Zipf's Law

Formula: V = n · log(n)

Claim: Network value grows logarithmically (power law distribution, 80/20 applies)

100 users → Value ≈ 460
1,000 users → Value ≈ 6,900 (15x, not 100x)

The Comparison

Users | Metcalfe (n²) | Zipf (n·log n) | Ratio
------|---------------|----------------|-------
100   | 10,000        | 461            | 21.7x
1,000 | 1,000,000     | 6,908          | 144.8x
10,000| 100,000,000   | 92,103         | 1,086x

As networks grow, Metcalfe increasingly overstates value.


Real-World Evidence

Facebook/Social Networks: Actual growth logarithmic (Zipf), not quadratic. You don't interact with all 500 friends—90% of interactions with top 10-20.

Tencent Study (2015): WeChat and QQ data validated n·log(n), not n²


Why It Matters

Valuations: Don't say "1M users = Metcalfe value X". Say "100k active users with engagement Y"

Projections: 10x users → 30-40x value (logarithmic), not 100x

Focus: Engagement and retention (power law users), not pure acquisition


When Each Applies

Metcalfe works: Small fully-connected networks (Slack team < 50 people, family chat), high-engagement platforms (trading, marketplaces)

Zipf wins: Large social networks (most are lurkers), communication platforms (message same 10 people 90% of time), content platforms (watch same creators repeatedly)


Key Insights

Metcalfe oversells—assumes all connections equal, leads to bubble valuations. Zipf is realistic—matches empirical data, accounts for power laws. 1,000 active users beats 10,000 inactive. Quality of network beats size of network. Focus on retention and engagement over acquisition.

Use the simulation to see how the models diverge!