Distribution Strategy
Frameworks, analysis, and tactics for building distribution. Key insight: you don't need a moat to win, you need a wedge. The moat is what you build while winning.
The Wedge → Lock → Expand Framework
Stage 1: WIN
- Product advantage (taste, speed, DX, flexibility)
- Goal: Get users, establish position
Stage 2: LOCK
- Convert usage into dependency
- User's data/workflow/context lives in your product
- Integrations that make you central to their stack
Stage 3: EXPAND
- Adjacent products that increase surface area
- Ecosystem/platform that others build on
- Become infrastructure, not just a tool
Failure mode: Staying in Stage 1 too long. Product stays great but you never convert to lock-in.
"Better Product" is a Wedge, Not a Moat
"Better product" alone is not durable. But it's a wedge that buys time to build durable moats.
Better product → initial traction → [build network effects / switching costs / brand] → durable moat
The question isn't "is your product better?" It's "what are you building while your product is better?"
TikTok: Algorithm was the wedge
Why the algorithm was better: Format enabled cleaner signals, cold start solved differently, content density (60+ data points/hour).
But Instagram Reels and YouTube Shorts caught up algorithmically. TikTok is still dominant because of downstream moats built while they had the advantage:
- Creator network effects
- Cultural position ("TikTok" as verb)
- Content library
- Personalized FYP (switching = retraining)
Is data a moat?
Data fits into multiple moat categories—it's an input, not a moat itself:
| Data type | Moat category | Example |
|---|---|---|
| More users → more data → better product | Network effect | Waze, Google clicks |
| Your content/history lives in product | Switching cost | Slack messages, Notion docs |
| Proprietary dataset no one else has | Cornered resource | Bloomberg terminals |
| Data to train better models | Contested | See below |
The "better algorithm" logic is leaky:
- 10x data ≠ 10x better model (diminishing returns)
- Better model ≠ better product (OpenAI beat Google without more data)
- Better product ≠ winning (distribution matters more)
Possession vs Flow:
- Static possession ("we have a big dataset") — not durable
- Dynamic flow ("every interaction improves product in ways users notice") — real moat
Honest take for most AI products: Data is not their moat. Their moat (if any) is distribution, switching costs, brand, or speed of iteration.
Lollapalooza Effect: Compounding Moats
Munger's concept: multiple forces combining for outsized results. Each moat type covers the weakness of another.
Pattern 1: Network effects + Switching costs
The network brings you in, switching costs keep you.
- Slack: Team adoption + message history/integrations
- Salesforce: Partner ecosystem + data/workflows
- LinkedIn: Professional graph + profile/connections
Pattern 2: Content + Distribution + Advertising
Flywheel: good product → users → attention → ad revenue → fund better product.
- Google: Search + Chrome/Android + AdWords
- YouTube: Video platform + Google integration + ads
- TikTok: Content + viral sharing + in-feed ads
Pattern 3: Platform + Ecosystem + Data
- iOS: iPhone + App Store + usage data
- Salesforce: CRM + AppExchange + customer data
- Shopify: E-commerce + app ecosystem + merchant data
Pattern 4: Infrastructure + Scale + Switching costs
- AWS: Cloud + volume discounts + migration pain
- Stripe: Payments + fraud detection + deep embedding
- Twilio: Communications API + volume pricing + rewriting logic
Pattern 5: Brand + Premium + Reinvestment
- Apple: Aspirational + 30-40% margins + R&D/retail
- Tesla: Visionary + higher margins + Gigafactories/Superchargers
- LVMH: Luxury + 60%+ margins + craftsmanship
Pattern 6: Hardware + Software + Services (Apple)
- Hardware alone is commoditizable
- Software alone can be copied
- Services alone have low switching costs
- Together: Photos + apps + watch + AirPods + iMessage = impossible to leave
Pattern 7: Marketplace + Logistics + Subscription (Amazon)
- Marketplace: Sellers ↔ Buyers
- FBA: Scale economies, speed
- Prime: Pre-paid commitment
- Data: Informs inventory, recommendations
The meta-pattern
| Moat type | Weakness | What covers it |
|---|---|---|
| Network effects | Niche unbundling | Switching costs |
| Switching costs | Enough incentive | Brand loyalty |
| Brand | Can erode | Premium pricing → reinvestment |
| Scale economies | Capital can replicate | Network effects |
| Ecosystem | Platform can change rules | Standards (you ARE the rules) |
Weak lollapaloozas
Netflix: Content + Algorithm + Brand + Scale. Missing: ecosystem, switching cost, network effects.
Uber: Marketplace + Brand. Missing: switching cost (multi-homing easy), ecosystem.
Viral Distribution as Built-in Engine
Viral distribution is a genuine edge when built into product nature — usage = exposure.
Types
| Type | Mechanism | Examples | Strength |
|---|---|---|---|
| Transactional | Product mediates transaction | Calendly, DocuSign, PayPal | Strongest |
| Collaborative | Requires multiple people | Slack, Figma, Notion | Strong |
| Output | Output naturally shared | Loom, Canva, TikTok | Wide reach |
| Communication | Product IS the medium | WhatsApp, iMessage | Constant |
Why it's an edge
- Zero marginal acquisition cost
- Qualified leads (already in use case)
- Trust transfer ("colleague uses this")
- Compounding (each user = distribution channel)
Viral ≠ Network effects
| Viral Distribution | Network Effects | |
|---|---|---|
| What it does | Acquires users | Makes product more valuable |
| Mechanism | Usage creates exposure | More users = better for each |
- Loom: Viral (share video), NO network effects
- LinkedIn: Network effects, WEAK viral
- Slack, WhatsApp: BOTH
Inherent vs Bolted-on
Inherent: Calendly links, DocuSign signatures — happens through normal usage, sustainable.
Bolted-on: "Invite friends, get $10" — works when inherent isn't possible or as acceleration.
K-factor math
K = invitations × conversion rate
- K = 0.2 typical for B2B SaaS
- K = 0.5-0.7 excellent
- K > 1.0 rare, temporary
Growth multiplier: 1/(1-K)
- K = 0.5 → 2x multiplier
- K = 0.7 → 3.3x multiplier
Building vs Using Channels
The distinction
| Building the channel | Using for distribution | |
|---|---|---|
| Example | You ARE eBay | Seller ON eBay |
| Who owns network effect? | You | The platform |
| Your moat | Strong | Weak (renting) |
Creating your own network effects through channels
1. Product is inherently viral — WhatsApp on App Store has its own network effects.
2. Usage creates exposure — Calendly, DocuSign, Loom.
3. Launchpad strategy ⭐ [Go deeper]
Airbnb on Craigslist: Used their distribution, but built own network (reviews on Airbnb, trust transferred to brand, users graduated).
Why brilliant:
- Distribution without building audience from scratch
- Build YOUR asset on THEIR traffic
- Platform can't easily stop you
- When ready, you own the network
Execution:
- Find platform with your users but poor UX
- Provide better experience routing through you
- Accumulate value (reviews, reputation) on YOUR side
- Graduate users to your platform
Other examples: PayPal/eBay, YouTube/MySpace, Zynga/Facebook (ended badly)
Risk: Platform dependency during launchpad. If banned before network built, lose everything.
4. UGC/Community compounds — Stack Overflow, Reddit. Network effect in CONTENT, not channel.
The ownership test
"If this channel disappeared tomorrow, would my users still recruit other users?"
- Yes → You have your own network effect
- No → You're renting
Product Analysis
| Product | Wedge | What they built | Durability |
|---|---|---|---|
| Stripe | DX when alternatives painful | Switching cost, ecosystem (Atlas, Billing, Tax), brand | Very high |
| Notion | Flexibility ("everything tool") | Switching cost (company lives there), network effects, brand | High |
| Vercel | DX ("git push magic") | Switching cost, ecosystem (owns Next.js), distribution | Medium-high |
| Linear | Product taste (speed, design) | Switching cost (issues, roadmap), brand | Medium |
| Superhuman | Product feel (speed, keyboard-first) | Brand, weak switching cost | Fragile |
Stripe: The Gold Standard
- Wedge: "Payments API that doesn't make you want to die"
- Lock: Switching is terrifying (critical path, compliance)
- Expand: Billing, Tax, Atlas, Identity
- Become infrastructure: "The financial infrastructure layer"
Superhuman: Cautionary Tale
- Wedge: Beautiful, fast email
- Built: Mostly brand, weak lock-in
- Vulnerability: Gmail improves, no deep switching cost
Case Study: Cursor and Commoditization
When models commoditize, switching costs are low
Code lives in git, not Cursor. Compare:
- Notion: entire company's docs → HIGH
- Stripe: embedded payment flows → VERY HIGH
- Cursor: code in git → LOW
Competitors: Not just editors, but AI labs
| Layer | Players |
|---|---|
| Editor/IDE | Cursor, Windsurf, VS Code |
| AI Lab | OpenAI (Codex), Anthropic (Claude Code), Google |
Why labs care about agentic coding:
- Training signal for better agents
- Wedge into "computer-use agent" vision
- Direct distribution for models
Cursor's vulnerability
Platform vs application tension. Cursor built on OpenAI/Anthropic. Platform has incentive to absorb application layer.
Worse than competing with Microsoft:
- Microsoft wants to bundle (might acquire)
- AI labs want training signal and distribution (building competing products)
What would make Cursor defensible
- Codebase memory (context takes time to rebuild)
- Team features (shared knowledge)
- Workflow integrations (central to deployment, PR review)
- Build ecosystem (Cursor + hosting?)
Scenario matrix
| Scenario | Outcome |
|---|---|
| Models commoditize, Cursor standalone | Vulnerable |
| Cursor builds ecosystem | Possible — hard to out-platform Microsoft |
| AI labs win agentic coding | Squeezed by suppliers |
| Cursor acquired | Most likely |
| Cursor wins before labs catch up | If UX gap persists, build lock-in fast |
Narrow path
- Stay model-agnostic
- Build switching costs fast
- Win on UX (labs historically bad at this)
- Get acquired before window closes
Historical Patterns
Compounding in action
PayPal: Viral referral with $20 two-sided rewards, 7-10% daily growth. Reduced rewards as network effects kicked in ($20 → $10 → $5).
Dropbox: 100K to 4M users in 15 months. 35% of signups from referrals at peak. Previous CAC from Google Ads was $233-388 for $99 product — impossible.
Slack: 70% of $100K+ customers started team-level. 10,000 message limit triggers at maximum lock-in.
Amazon flywheel: Lower prices → customers → sellers → selection → scale → lower prices. Prime added switching cost compounding.
WhatsApp: 400M users with zero marketing. Phone contact list = pre-built social graph.
Behavioral Foundations
Cialdini's principles
Reciprocity: Freemium, lead magnets, content marketing.
Social proof: Testimonials, user counts, "X people viewing." 70% trust peer opinions vs 14% ads.
Scarcity: Waitlists, limited editions, exclusive access.
Commitment/consistency: Micro-conversions, free trials, foot-in-door.
Unity: "Join 8,000 marketers like you."
Hooked model
Trigger → Action → Variable Reward → Investment
Variable rewards create dopamine surges. Investment phase creates switching friction.
Counterintuitive Strategies
Friction as advantage: Duncan Hines cake mixes with eggs outsold "just add water." Waitlists, qualification processes, premium pricing.
Easy cancellation increases retention: Olipop, SparkToro found 8%+ of monthly customers previously subscribed.
Zero-click content: Give complete value without requiring clicks. 93% of word-of-mouth is offline.
Brand during downturns: 99-1 Rule: only 1% of B2B buyers in-market. Counter-cyclical investment builds mental availability cheaper.
Entering "dead" channels: When everyone abandons email/podcasts, competition drops, costs plummet.
Under-discussed Mechanisms
Bundling as distribution: Rippling, Deel, Workday — reduces wedge opportunities, procurement advantages.
M&A as distribution: Acquire customers, channels, talent. Netflix content acquisition.
Standards-setting: Once you're the standard (PDF, USB), competitors must conform.
Infrastructure/API: Stripe, Twilio, AWS — every integration increases switching costs.
Regulatory capture: Shape rules to favor your model. Healthcare, finance, telecom.
Mental Models
Second-order effects
- First-order: "This channel will bring users"
- Second-order: "Will those users refer others? Will success attract competitors?"
- Third-order: "How does market-wide adoption change economics?"
Inversion
Instead of "How do I succeed?" ask "What would cause failure?"
Compounding vs linear
Does this appreciate or depreciate over time?
- Compounding: SEO, network effects, brand
- Linear: most paid advertising
- Depreciating: growth hacks
Opportunity cost
- What are we NOT pursuing by focusing here?
- Is this the highest-leverage investment?
Reversibility
Highly reversible: Paid ads, short partnerships Hard to reverse: Brand positioning, platform dependencies
Make reversible bets when uncertain.
Time Horizon
Not every company needs to be Stripe. Winning a window and exiting is a valid strategy.
The strategic question: Are you building for dynasty or decisive window + exit?
Related
- Distribution Reference - Definitions, taxonomies, and inventories
- Clear ROI is a Must for Sales-Led Products
- How Granola Grows - Wedge + lock example
- Expertise Enables Ambition