Summary

Andrew Chen (growth @ Uber, partner @ a16z) gives a crash course on all things network effects, which can be broken down into the engagement, acquisition, and economic effects. To solve the cold start problem, attract the “hard side” of the network by any means necessary through growth hacks. Network effects are inherently asymmetric, favoring David over Goliath.

“Cherry picking is an enormously powerful move because it exposes the asymmetry inherent in the David and Goliath nature of networks”

Key Takeaways

• Network effect: A product that gets more valuable as people use them
• 5 stages of network effect: (S curve)
  1. Cold start problem: attract “hard side” of network, growth hacks
  2. Tipping point:
  3. Escape velocity:
  4. Hitting the ceiling: market + network saturation; adjacent users
  5. The moat: winner-take-all
• Network effect: trio of underlying forces:
  1. Engagement effect: how a denser network creates higher stickiness (use cases, loops)
  2. Acquisition effect: ability to tap into its network to acquire customers (viral growth)
     1. Viral factor: ratio of cohort that drives next
  3. Economic effect: ability to improve business model with growth (ARPU)
     1. Data network effect: eg. Credit bureaus
• Density of network > size of network

Notes

1: Network effects

1: Network effects

• A product that gets more valuable as people use them
• Low technical risk → low defensibility

2: A brief history

• Metcalf’s law: value of network = n^2
• ↑ population dynamics are more accurate
• Ali threshold: carrying capacity
  • = market saturation

3: Cold start theory

• 5 stages of network effect: (S curve)
  1. Cold start problem
  2. Tipping point
  3. Escape velocity
  4. Hitting the ceiling
  5. The moat

2: The cold start problem

4: Tiny speck

• Have to attract “hard side” of the network

5: Anti-network effects

• Churn caused by lack of network
• Marketplace products: top 4 → 76% gross
• Slack: 3 people, 2,000+ messages is sticky
• Facebook: 10 friends in 7 days
• Zoom: 2 people
• Airbnb: 300 listings, 100 reviewed listings
• Uber: 15-20 concurrent rides
• Size of initial market determines launch strategy
• Density, interconnectedness

6: The atomic network

• Credit cards: Fresno, CA BofA
• Growth hacks: do whatever it takes
• Opposite of disruption theory
• Have a hypothesis about end state
• Niche down more than you want to
• Determine size of atomic network

7: The hard side

• Eg. Wikipedia: 0.02% contributors

8: Solve a hard problem

• Tinder: men: 50% yes, women: 5% yes
• Supply side is generally hardest

9: The killer product

• Zoom: do one thing well
• Viral, easy to use
• Basically PLG: free tier

10: Magic moments

• Clubhouse: take advantage of shifts
• Zero: opposite of magic moment
• Problem to be continually solved

3: The tipping point

11: Tinder

• Dating platforms are notoriously difficult
  • Geography, demographics, high churn
• Tinder USC party (95% conversion)
• Top-down marketing: popular, influential ppl
• Repeatable strategy

12: Invite only

• LinkedIn, Gmail, Netflix
• LinkedIn: key was mid-tier professionals
• Not FOMO, but carefully curating invites to copy/paste to similar networks

13: Come for the tool, stay for the network

• Instagram: retro filters → social media
• ↑ has to be a pivot
• OpenTable: management → reservations
• Google: search → ads

14: Paying up for launch

• Coupons for Coca Cola
• Eg. subsidizing content creators (hard side)
• Partnerships → distribution

15: Flintstoning

• Missing features → manual human effort
• Eg. contacting developers
• Reddit: dummy accounts, scrapers
• First-party content

16: Always be hustling

• Uber: ops > product company
  • Eg. Uber ice cream
• Stunts, hacks, highly manual tactics
• B2B GTM 3 sources:
  1. Personal network
  2. Go to where they are
  3. Press
• PH: density of early adopters

4: Escape velocity

17: Dropbox

• Growth team: focused on growth tactics
• Low vs high-value users

18: Trio of forces

• Network effect: trio of underlying forces

19: Engagement effect

• Engagement effect: how a denser network creates higher stickiness (use cases, loops)
• Cohort retention curve: eg. scurvy study
• Study: we spend 80% phone time w/ 3 apps
• Min baseline: 60% 1D, 30% 7D, 15% 30D
• For each cohort, analyze what differentiates low, high-value users

20: Acquisition effect

• Acquisition effect: ability to tap into its network to acquire customers (viral growth)
• PayPal: birth place of PLG growth hacks
• Network-driven viral growth: embedded into product experience itself
• Optimize viral loop for conversion rate
• ↑ often unique to product / category
• Viral factor: ratio of cohort that drives next
  • Eg. 0.95 → 20x cohort size

21: Economic effect

• Economic effect: ability to improve business model with growth (ARPU)
• Data network effect: eg. Credit bureaus
• Subsidies: imitate larger market
• Network → higher conversion rate
  • Eg. more friends to impress

5: The ceiling

22: Twitch

• Saturation, churn, identifying ICP

23: Rocket ship growth

• Only 5% of investments 10x+
• Stock market: 57% venture-backed comps
• T2D3: PMF → $2m ARR → 3x 3x 2x 2x 2x
  • Land at $100M+ ARR, $1B+ valuation
• Establish leading metrics, set goals
• Ceiling → brain drain

24: Saturation

• Market saturation:
• Solution to saturation: ↑ ARPU, upsell
  • Ie. enshitification
• Network saturation: diminishing returns of larger network (eg. 100 v 1000 results)
• Adjacent users: users whose experience is sub-par (vs power users)
  • Focus on this when saturated

25: The law of shitty clickthroughs

• The law of shitty clickthroughs: every marketing channel degrades over time
• Banner ads: 100x drop, email: 30% → 13%
• Leads to ballooning marketing costs
• Solutions:
  • Growth teams, deep user analysis
  • New channels: influencers, memes, etc

26: When the network revolts

• Power law of high-value users
• Feedback loop creates concentration ↑
• Professionalizing the hard side

27: Eternal September

• Context collapse: too many networks simultaneously brought together
  • Unsure how to act
• Provide networks within networks
• Network self-governance: moderation
• Dunbar’s number

28: Overcrowding

• Preferential attachment: ↓
  • Old vs new money / social capital
  • Upstarts will try other platforms
• Signals of user’s interaction (for you feed)
• Fight against overcrowding never ends

6: The moat

29: Wimdu vs Airbnb

• Rocket internet: copy US businesses
• Windu’s quantity vs Airbnb’s quality

30: Vicious cycle, virtuous cycle

• Winner-take-all
• Not features, but leveraging network
• Experiment, incubate, pivot

31: Cherry picking

• Unbundling of Craigslist: COFACTORY IDEA
• Enshitification: inevitability of incumbents over-serving customers
• “Cherry picking is an enormously powerful move because it exposes the asymmetry inherent in the David and Goliath nature of networks”
• ^ especially for consumer

32: Big bang failures

• Google+
• Bottom-up > big bang

33: Competing over the hard side

• Uber: live competitor analysis

34: Bundling

• Microsoft: eg. Office, Visual Basic
• Super-app, upsell, cross-sell
• Not assured

The future of network effects

• Crypto
• Alumni from past network successes
• Viral growth, launching new markets, accelerating engagement