Network Effects

Thinking Tools3.3 Mapping the System Around the ProblemAlso known as Metcalfe's Law

First stated by Theodore N. Vail (AT&T Annual Report, 1908), modeled by Jeffrey Rohlfs (1974), popularized as Metcalfe's law after Robert Metcalfe (early 1980s), formalized by Michael L. Katz and Carl Shapiro (Network Externalities, Competition, and Compatibility, 1985), and brought to strategy by Carl Shapiro and Hal R. Varian (Information Rules, 1998).

Explains markets where each user gains as more users join, which is why scale compounds and many platform markets settle on one or two winners.

Two freight marketplaces race for members, and a small early gap tips the whole region

Each month new members pick the network where a match is likelier, and that is the bigger one. At rest the rival starts 20 members ahead and ends with most of the region. Sign up a few extra carriers a month and Play: the tipping runs the other way. Raise the share who use both networks and the race stops tipping. Illustrative.

Select a loop to trace it, or a box to read it.R1, Our pull. More members raise our chance of a match, which draws more newcomers to us. No negative (−) links: an even count, so the loop reinforces.R2, The rival's pull. Every member the rival adds lowers our relative pull and raises theirs. Two negative links: this loop reinforces too, in the rival's favor. Two negative (−) links: an even count, so the loop reinforces.Our members. Shippers and carriers active on our marketplace: 500 at the start.Chance of a match. Why size matters: the more members on a network, the likelier a shipper finds a truck and a carrier finds a load. Newcomers weigh it heavily, so the bigger network wins more than its share.Rival's members. Active on the rival's marketplace: 520 at the start, a head start of three months.Carriers we sign up. Extra carriers brought on each month with a fee waiver.Members using both. Newcomers who join both networks. They count on each, so neither gains an edge from them.
0
0%

Our share of the region's members

0: 49%1: 49%2: 48%3: 48%4: 47%5: 46%6: 46%7: 45%8: 43%9: 42%10: 40%11: 39%12: 37%13: 35%14: 33%15: 31%16: 29%17: 27%18: 25%19: 23%20: 21%21: 19%22: 18%23: 17%24: 15%0%25%50%75%100%01234567891011121314151617181920212223240: 49%: 49%2: 48%: 48%4: 47%: 46%6: 46%: 45%8: 43%: 42%10: 40%: 39%12: 37%: 35%14: 33%: 31%16: 29%: 27%18: 25%: 23%20: 21%: 19%22: 18%: 17%24: 15%0%25%50%75%100%024681012141618202224

Month 24 of 24

Our members
705
Rival's members
3,915
Our share of the region's members
15%

R1, Our pull
More members raise our chance of a match, which draws more newcomers to us. No negative (−) links: an even count, so the loop reinforces.
R2, The rival's pull
Every member the rival adds lowers our relative pull and raises theirs. Two negative links: this loop reinforces too, in the rival's favor. Two negative (−) links: an even count, so the loop reinforces.
Our members
Shippers and carriers active on our marketplace: 500 at the start.
Chance of a match
Why size matters: the more members on a network, the likelier a shipper finds a truck and a carrier finds a load. Newcomers weigh it heavily, so the bigger network wins more than its share.
Rival's members
Active on the rival's marketplace: 520 at the start, a head start of three months.
Carriers we sign up
Extra carriers brought on each month with a fee waiver.
Members using both
Newcomers who join both networks. They count on each, so neither gains an edge from them.

An example

Two online marketplaces for short-haul freight launched in the same region a few months apart. The rival had 520 active members, shippers and carriers together, and ours had 500. The founders planned to win on product quality.

The head of strategy modeled how newcomers choose. A shipper joins the network where a truck is likelier to be free, a carrier the one where a load is likelier to be waiting, and both odds rise with size. With a 20-member gap and nothing else changed, newcomers would favor the rival a little more each month, and in two years we would hold about 15 percent of the region. Signing 20 extra carriers a month with a fee waiver reversed it, to about 86 percent. The model also showed what to watch: if newcomers began joining both networks, the winner's pull would stall near half. The company funded the waiver for a year, added consolidated invoicing to make single-network use cheaper, and tracked the share of members active on both.

Why it works

When each member gains as the network grows, the choice of network stops being independent of everyone else's. A newcomer picks the network where a match is likelier, usually the larger one, and that choice makes it larger still. A small lead compounds, a small deficit compounds the other way, and the market can tip to one winner on a difference nobody would call decisive at the start.

Naming the effect and its mechanism shows what decides the race. Early growth is worth more than its revenue, because it buys later growth, so pricing for adoption can be rational. It also shows what stops the tipping: when members can use both networks at little cost, size matters less and the lead stops compounding. How defensible a lead is depends on how costly it is to use both.

The thinking behind it

Jeffrey Rohlfs Jeffrey Rohlfs gave an early formal model in A Theory of Interdependent Demand for a Communications Service (Bell Journal of Economics and Management Science, 1974). A communications service is worth more to each subscriber as more people subscribe, so demand depends on the size of the user base itself, and a new service must reach a critical mass before its growth can sustain itself. Theodore N. Vail had put the idea plainly in AT&T's annual report for 1908: a telephone's value rises with the number of others it can reach.

Carl Shapiro Carl Shapiro, a Berkeley industrial organization economist, formalized the effect with Michael L. Katz in Network Externalities, Competition, and Compatibility (American Economic Review, 1985). Their model treats the gain each user gets as more people use a product, and shows how that gain shapes firms' choices to make their products compatible with rivals' or not. Compatibility is the question behind the figure's second slider: when members can be on both networks at once, the pull toward a single winner weakens.

Hal R. Varian Hal R. Varian, founding dean of the School of Information at the University of California, Berkeley, and later chief economist at Google, wrote Information Rules: A Strategic Guide to the Network Economy (Harvard Business School Press, 1998) with Carl Shapiro. The book applies standard economics to information goods, including versioning, lock-in and switching costs, standards, and network effects with positive feedback, and it carried those ideas to the managers deciding how to price and launch such products.

How to run it

  1. Ask whether each user gains as others join. Check with users whether the product gets better for them when the network grows.
  2. Separate same-side from cross-side effects. Shippers may care about carriers far more than about other shippers; mark which side pulls which.
  3. Name the mechanism. Compatibility, liquidity such as faster matches, or social proof; each one erodes in its own way.
  4. Measure the gap where users choose. Compare active members in the region, lane, or segment where the choice is made.
  5. Buy early growth if the effect is strong. Subsidize the side that attracts the other, and stop once the lead sustains itself.
  6. Watch use of both networks. Track the share of members active on both; a rising share weakens the leader's pull.
  7. Recheck for erosion each year. Regulation, a new standard, or a new technology can reset the race.

Where it goes wrong

The first failure is assuming a network effect where there is none. Many products have users without having a network: a buyer of accounting software gains nothing from other buyers. Spot it when nobody can say how a new user makes the product better for an existing one, and drop the subsidy plan.

The second is measuring the wrong network. The effects are often local, and a freight marketplace that leads nationally can trail in the one region where a shipper books, so measure where the choice is made. The third is treating a lead as permanent. Multi-homing, a compatibility rule, or a cheaper way to switch can stop a market from tipping, and a firm that priced for a monopoly is left with the subsidy bill.

When to use something else

  • Tipping Points. When the question is the threshold at which a market or a behavior flips, beyond two named rivals.
  • Technology S-Curve. When the question is how fast adoption will grow and level off, with no network to win.
  • Porter's Five Forces. When the question is the industry's overall profitability, with network effects as one force among several.

Where it sits in SOLVE eX

Go to the source

  • A Theory of Interdependent Demand for a Communications Service, Bell Journal of Economics and Management Science, 1974
  • Network Externalities, Competition, and Compatibility, American Economic Review, 1985
  • Information Rules: A Strategic Guide to the Network Economy, Harvard Business School Press, 1998
  • Metcalfe's law, from Ethernet sales presentations, early 1980s
  • AT&T Annual Report, 1908

Back to the catalog