Where it sits
Which is why these businesses are so hard to start and so hard to dislodge.
Tell it apart
| Aspect | Network effect | Economic moat |
|---|---|---|
| Is a | Mechanism | Category |
| Requires users | Yes | No |
| Count | One of five | Five types |
| Named | 1908 / 1993 | 1990s |
| Types | One | Five |
| Can be built alone | Rarely | Sometimes |
The network effect is ONE of five named moat types, not a synonym for moat. Cost advantage, switching costs, intangible assets and efficient scale do the same defensive job with no network at all.
What it doesn't mean
Metcalfe framed it from 1980 in terms of COMPATIBLE COMMUNICATING DEVICES — and so, still, did George Gilder when he named the law in Forbes ASAP in September 1993: 'connect any number, n, of machines — whether computers, phones or even cars — and you get n squared potential value.' The USERS framing attached to it only afterwards, and the shift from devices to people quietly changed what the law claimed.
A 2006 IEEE Spectrum paper argued that Metcalfe's law is wrong, proposing value grows as n log n, because assuming equal value per connection ignores diminishing returns at scale. Competing models bracket it: Sarnoff's law (n) and Reed's law (2 to the n).
What it really means
The clearest early statement of the network effect was made in 1908, and it was an argument for a monopoly.
Theodore Vail, running a telephone company after its patents expired, wrote in the annual report that competing telephone networks were wasteful — because a phone is worth nothing unless the people you want to call are on the same system. Two competing networks means two half-useful networks. The logic is correct, and the conclusion he drew from it was that there should be one company, and it should be his.
That origin is worth keeping, because the concept has since been repackaged as a product-design insight and it began life as a competition argument.
The quantified version, Metcalfe's law, has two things wrong with how it is usually cited.
First, the subject changed. Metcalfe framed it from 1980 in terms of compatible communicating devices — a hardware argument about why buying one fax machine is pointless — and put it on a slide for 3Com's sales force in 1983. George Gilder named it after him in Forbes ASAP in 1993 — still, in his own words, about machines, "whether computers, phones or even cars." The users version attached to it afterwards, and it is the one everybody now cites. Devices are interchangeable and equally valuable. People are not, and the substitution smuggled in an assumption.
Second, the exponent is disputed. A 2006 paper in IEEE Spectrum argued directly that Metcalfe's law is wrong, proposing value grows as n log n rather than n², because treating every possible connection as equally valuable ignores that most of them are worthless to you. Competing models bracket the range widely: Sarnoff's law (value ∝ n) and Reed's law (value ∝ 2ⁿ).
So the honest summary: the effect is real and enormously important, and the arithmetic is contested by a factor that ranges from linear to exponential — which is a polite way of saying nobody can currently price it.
Where it comes from
English
English
The earliest articulation was a MONOPOLY ARGUMENT, not a product-design one — it was made by a telephone executive arguing that competition in telephony was wasteful.
- 1908
Theodore Vail, the first post-patent president of a telephone company, argues the case in the annual report while advocating for a telephone monopoly.
- 1980–1993
Robert Metcalfe frames the idea in terms of compatible communicating devices, from 1980 and in a 1983 presentation at 3Com; George Gilder attaches it to users and names it Metcalfe's law in Forbes ASAP, 13 September 1993.
In real life
- The first telephone in a town is nearly useless; every new subscriber makes everyone else's phone a little more worth having.
- On a marketplace app, more sellers attract more buyers, and more buyers attract more sellers — each side is the other side's reason to join.
- A language-exchange app with ten users is a ghost town; with ten thousand, you can find a Portuguese speaker who wants to practise English at three in the morning.
Questions
Q01Who first described network effects?
The account usually credited is Theodore Vail's: he set out the concept in a telephone company's 1908 annual report — while arguing that telephony should be a monopoly, because a network's value depends on everyone being on it.
Q02Is Metcalfe's law correct?
Challenged. A 2006 IEEE Spectrum paper argued for n log n rather than n squared, on the grounds that not all connections are equally valuable.
Q03What are the rival laws?
Sarnoff's law, where value is proportional to n, and Reed's law, where value is proportional to 2 to the n. The three bracket a wide range, which tells you how unsettled the quantification is.