Network effects: when each user makes it better
Direct, two-sided and data network effects, the economics of the loop, what it costs a rival to fight one, and the four ways network effects break.
Key takeaways
- A product has a network effect when each new user makes it more useful for the others.
- UPI, India's instant payment system, is interoperable: any UPI app can pay any account at any bank on the system.
- Common mistakes: Calling any popular app a network effect: the test is whether one user's presence makes the product better for others.
Key idea
A product has a network effect when each new user makes it more useful for the others. Then the biggest network attracts the most new users, which makes it bigger still. The effect protects a business only if users cannot easily use a rival network at the same time.
Three kinds, in plain words
- Direct: users gain from other users of the same kind. A payments or messaging app is more useful when the people you pay or message use it too.
- Two-sided (also called indirect): two different groups gain from each other. More riders bring more drivers, and more drivers mean shorter waits, which bring more riders. Buyers and sellers on a marketplace work the same way.
- Data: more use produces more data, which improves the product, for example better route times or better search results. This is usually the weakest kind, because each extra piece of data adds less than the one before.
Worked case
Why drivers go where the riders are, and what it costs to fight that
The prompt
Illustrative numbers, fictional apps. In a city in Singapore, drivers on the larger ride app, App B, carry a passenger for 45 minutes of each hour online, because riders are everywhere. Drivers on the smaller App A carry a passenger for only 30 minutes of each hour. On both apps a driver keeps SGD 0.80 for each minute with a passenger. How much more does a driver earn per hour on App B, and what would App A have to spend in bonuses to match it for 2,000 drivers working 40 hours a week?
The structure
- Driver earnings per hour = busy minutes per hour x earnings per busy minute
- Compare the two apps
- Cost to match = gap per hour x drivers x hours
Working it through
1. App A earnings
30 busy minutes at SGD 0.80.
App A driver earnings (SGD per hour):30 × 0.8 = 242. App B earnings
45 busy minutes at SGD 0.80.
App B driver earnings (SGD per hour):45 × 0.8 = 363. The gap
The extra a driver earns on App B every hour.
Earnings gap (SGD per hour):45 × 0.8 - 30 × 0.8 = 124. Weekly cost to match
SGD 12 an hour, for 2,000 drivers, for 40 hours.
Bonus cost (SGD per week):12 × 2,000 × 40 = 960,0005. Yearly cost to match
Fifty-two weeks.
Bonus cost (SGD per year):960,000 × 52 = 49,920,000
The recommendation
A driver earns SGD 36 an hour on the bigger app against SGD 24 on the smaller one, only because the bigger app keeps drivers busier. To match it, App A would have to pay about SGD 50 million a year in bonuses for just 2,000 drivers, and the moment it stops, drivers drift back. That is the economics of a network effect: the leader gets the same result for free that the follower must buy. App A's better options are to win one district or one niche where it can be the denser network, or to combine with a rival.
Risks: If drivers run both apps at once, the gap shrinks, because App A can borrow App B's drivers; Network effects are local: winning one city says little about the next one.
Real cases: network effects are local
Two exits that show the economics
- Southeast Asia: on 26 March 2018 Grab announced that it had acquired Uber's Southeast Asia operations in Cambodia, Indonesia, Malaysia, Myanmar, the Philippines, Singapore, Thailand and Vietnam, with Uber taking a 27.5 percent stake in Grab. A global brand did not automatically win each city; the denser local network did.
- Middle East: on 2 January 2020 Uber completed its USD 3.1 billion acquisition of Careem, taking Careem's mobility, delivery and payments businesses across the greater Middle East (approval in a few countries was still pending on that date). Here the global player bought the local network instead of paying to outgrow it.
- The economic reading: when two networks fight city by city, both pay to subsidise drivers and riders. Combining ends the subsidy. That is why regulators look closely at such deals.
UPI, India's instant payment system, is interoperable: any UPI app can pay any account at any bank on the system. A Government of India release from August 2026 reports that banks live on UPI rose from 44 in the financial year 2016-17 to 703 in 2025-26, and yearly transactions from 1.78 crore to more than 24,162 crore (one crore is 10 million). The network effect is huge, but it belongs to the shared system, not to any one app. So apps compete on features and offers, and a user can move between them easily. A network effect only becomes a company's advantage if the company controls who can join the network.
Banks live on UPI rose from 44 in 2016-17 to 703 in 2025-26. By how many times did the number of banks grow? Round to one decimal.
UPI handled more than 24,162 crore transactions in 2025-26. One crore is 10 million. How many billion transactions is that? (One billion is 1,000 million.)
Four ways network effects break
- Multi-homing: users and suppliers use two or more networks at once (drivers with two apps, shops on two marketplaces). Then the leader's size counts for less.
- Locality: the effect stops at a city or a country border. A leader in one place has no edge in the next.
- Congestion and quality: past a point, more sellers or more posts make the product worse, not better, and users leave for a cleaner network.
- Openness and rules: shared systems such as UPI, or regulators requiring big platforms to open up, let rivals tap the same network. (Lesson 7 has the European rules.)
Calling any popular app a network effect: the test is whether one user's presence makes the product better for others. Ignoring multi-homing. Assuming a global leader wins every city. Overrating data: more data often adds little after a point.
Two food delivery apps operate in one city. Almost every restaurant lists on both, and most riders work for both. What does this tell you about the leader's network effect?
Which is the clearest two-sided network effect?
Sources, checked on 2026-10-01. Company figures are from the company's own reports or filings; per-unit figures and comparisons are our arithmetic on them.
- Grab, press release "Grab merges with Uber in Southeast Asia", 26 March 2018: https://www.grab.com/kh/press/business/grab-merges-with-uber-in-southeast-asia/
- Uber Technologies, press release "Uber completes acquisition of Careem", 2 January 2020: https://investor.uber.com/news-events/news/press-release-details/2020/Uber-Completes-Acquisition-of-Careem/default.aspx
- Press Information Bureau, Government of India, release on ten years of UPI, 24 August 2026 (volumes, value and banks live, from NPCI data): https://www.pib.gov.in/PressReleasePage.aspx?PRID=2302657®=48&lang=2
Sources for this lesson (4)
- Recognized public explanations of case-interview concepts and terms
- Grab, press release "Grab merges with Uber in Southeast Asia", 26 March 2018
- Uber Technologies, press release "Uber completes acquisition of Careem", 2 January 2020
- Press Information Bureau, Government of India, release on ten years of UPI, 24 August 2026 (volumes, value and banks live, from NPCI data)
My notes on this lesson
0 of 5,000 characters. Saves automatically.
Try the 4 remaining checks and drills above to complete this lesson (0 of 4 done).
Spotted something wrong or out of date? Report a mistake. We check every report and correct the page.