Vertical Integration
Bundle or unbundle products and services?
Vertical integration is a strategy where a company expands its operations to control multiple stages of the supply chain, from production to distribution.
The Model
Why integrate vertically? Control creates compounding advantages.
| What You Control | Advantage |
|---|---|
| Data source | Proprietary insights |
| Workflow | Switching costs |
| Payments | Revenue per customer 10x |
| Distribution | Customer acquisition |
The thesis: Vertical integration from sensor to workflow creates compounding moats.
Three Expressions
Vertical integration appears in different forms:
| Expression | What It Is | Dig Deeper |
|---|---|---|
| Vertical AI | AI trained on industry-specific data | Below |
| Vertical SaaS | Software tailored to industry workflows | VSaaS Playbook |
| Full-Stack Startups | Own the entire value chain | a16z |
Vertical AI
Tremendous disruptive potential as language models cross capability thresholds for products that solve customer problems end-to-end.
Building successful vertical AI requires:
- Deep iteration on prompts and workflows
- Deep integration of business domain expertise
- Rigorous testing for 100% reliability
Data sources:
- Domain-specific integrations
- Proprietary datasets
- Customer-specific connections
Impact: CoCounsel
AI legal assistant that can:
- Engage in dialogue
- Analyze millions of documents for evidence
- Produce well-researched memos
Tasks that took lawyers days now take minutes. Customers had existential crises seeing demos—this fundamentally changes legal work.
The Insight
The future of prompting is not just about examples of good outputs, but teaching models expert domain-specific reasoning processes.
Vertical SaaS
The playbook for building industry-specific software lives at VSaaS.
Quick reference:
| Component | What To Learn | Link |
|---|---|---|
| Strategy | 9 Keys to Monopolizing | Principles |
| Tech Stack | DePIN + Embedded Finance | Platform |
| Implementation | Development Playbook | Protocols |
| Metrics | ACV Tiers, Focus Matrix | Performance |
| Team | Industry Engineers | Players |
Full-Stack Verticals
Industries where vertical integration matters most:
| Vertical | Why | Link |
|---|---|---|
| Construction | Fragmented, physical + digital | High opportunity |
| Real Estate | High transaction value, data-rich | High opportunity |
| Healthcare | Compliance, integration complexity | Medium opportunity |
| Manufacturing | Supply chain, automation | Medium opportunity |
| Education | Fragmented, outcome measurement | Growing opportunity |
See Full-Stack Startups for more verticals.
Market Examples
Who's winning through vertical integration:
| Category | Dominant Players | Market Share |
|---|---|---|
| Restaurant Management | Toast, Square | ~70% combined |
| Construction Management | Procore, Autodesk | ~80% combined |
| Real Estate Brokerage | CoStar Group | ~80% in CRE data |
| Auto Dealership | CDK Global, Reynolds | ~90% combined |
| Legal Practice | Clio, MyCase | ~75% combined |
| Dental Practice | Dentrix, Eaglesoft | ~80% combined |
| Fitness/Spa | Mindbody | ~70% (post-Booker acquisition) |
The pattern: In every vertical, 2-3 players capture 70-90% of the market.
The Vertical Discovery Matrix
Use Matrix Thinking to find which verticals have the highest integration opportunity:
| Low Competition | High Competition | |
|---|---|---|
| High Data Value | Build here — first-mover advantage | Differentiate on AI/workflow |
| Low Data Value | Avoid — no moat | Race to bottom |
The process:
- Idea Discovery — Identify forces creating friction
- Industries — Deep-dive 5P analysis per vertical
- Vertical Integration (you are here) — Strategy to capture the value chain
- Matrix Thinking — Find gaps others miss
The Dual Play
Every vertical presents two plays. Sell RaaS — package the platform as software for operators in that vertical. Operate — run a venture in the vertical using the platform as in-house competitive advantage. The right play depends on three scores.
| Score | What It Measures | Source |
|---|---|---|
| Data Value | How valuable is the vertical's data footprint for AI training and predictions? | Industry Matrix |
| Feature Coverage | How many of the vertical's required functions already exist in the Mycelium? | RaaS Catalog |
| Capital Signal | Is money flowing into this vertical? Are exits happening? | 2026 Vertical Report |
Vertical Scoring
Each vertical scored against data footprint value, platform readiness, and capital momentum. Data and AI scores from the industry matrix. Feature coverage from the RaaS vertical requirements. Capital signal from the 2026 Vertical Report.
| Vertical | Data | AI | Feature Coverage | Capital Signal | Best Play | Why |
|---|---|---|---|---|---|---|
| Healthcare | 5 | 5 | Low (compliance gap) | Strong (597 deals, $11.8B, led exits) | Sell RaaS | Regulatory moat too deep to operate without domain team. Sell compliance-ready tooling. |
| Manufacturing | 4 | 3 | Low (IoT gap) | Accelerating (41% deal growth, 57% AI-native) | Sell RaaS | Capital pouring in, operators need workflow + sensor tools. |
| Construction | 3 | 2 | Medium (project mgt partial) | Moderate | Sell RaaS | Fragmented SMBs buy tools. Data moat + payments = Levelset pattern ($500M exit). |
| Real Estate | 4 | 3 | Medium (6 PropTech tracks mapped) | Strong ($300T+ market) | Both | Sell property management RaaS AND operate tokenized RE ventures. |
| Energy | 5 | 4 | Low (DePIN gap) | Growing | Operate | DePIN device deployment = data moat. Own the sensors, own the predictions. |
| Education | 3 | 4 | Low | Emerging (highest early-stage step-ups 2.5x) | Operate | Low capital intensity ($200K/employee). Build credentialing on the platform. |
| Legal | 3 | 4 | Medium (doc intelligence exists) | Hot (GC AI $555M, 50x multiples) | Sell RaaS | PLG possible. Fastest path to $1M ARR. |
| Advertising | 4 | 5 | Low | Saturated | Operate | Data IS the product. Use AI tooling in-house for agency ventures. |
Play Logic
Sell RaaS when:
- High regulatory burden (healthcare, legal) — operators pay for compliance
- Fragmented market of SMBs (construction, field services) — volume play
- Feature coverage is already medium+ — shorter build cycle to sellable product
- Capital is flowing in — buyers have budget
Operate when:
- Data footprint is the competitive advantage — you need to own the sensors and the data flywheel
- Low capital intensity — you can bootstrap with the platform
- AI automation is high — the platform does the work, fewer humans needed
- The vertical's moat is operational execution, not software features
Both when:
- The vertical has sub-segments with different dynamics (RE: commercial vs residential vs tokenized)
- RaaS revenue funds the operating venture
Key Features by Play
What each play demands from the RaaS catalog:
| Play | Critical Features | Why |
|---|---|---|
| Sell RaaS | Embedded payments (BILL-002), multi-tenant auth (AUTH-001–008), compliance framework (SEC-003–005), webhook system (INTG-003) | Customers need to pay through you, isolate their data, meet regulations, and integrate with existing tools |
| Operate | ETL pipelines (DATA-007), AI orchestration (AI-008), DePIN telemetry (IOT-001–002), analytics (ANAL-001–002) | You need data capture, AI-driven operations, sensor networks, and performance dashboards |
| Both | All of the above + workflow engine (WORK-001), task management (WORK-002), document intelligence (AI-001) | Full stack: sell the software AND run on it |
The 2025 Signal
The 2026 Vertical Report confirms the thesis. Vertical AI is no longer a hypothesis — 53% of deal volume, $186B deployed. The data:
- Capital is vertical-first. Quarterly share grew from 19% to 42% through 2025.
- Manufacturing is AI-native. 57% of funded companies founded post-2022. Fastest deal acceleration (41%).
- Healthcare exits lead. 43 exit transactions. Proven liquidity path.
- Aerospace leads capital intensity. $476K per employee — conviction despite unproven VC returns.
- Education has the best step-ups. 2.5x early-stage valuation despite lowest capital velocity. Underpriced.
- Geography matters. Non-coastal US states show 60-71% verticality. The opportunity isn't where the VCs sit.
The verticals where the platform already has feature coverage are the fastest path to RaaS revenue. The verticals where data footprint value is highest but coverage is lowest are the operate-first opportunities — deploy DePIN devices, capture the data, build the moat before selling the tools.
Bundle Economics
The Mycelium is a bundle. Every venture composes from shared capabilities. The question is not whether to bundle — it's which capabilities to package together, for whom, and at what price. Shishir Mehrotra's bundling framework provides the decision logic.
Audience Structure
Start with audience, not price. Every capability in the RaaS catalog has two audience types:
| Audience | Definition | Example |
|---|---|---|
| SuperFans | Would buy this capability standalone at full price | CRM power users, compliance-heavy operators |
| CasualFans | Would use it if access were easier or cheaper — bundled in | Operators who need basic invoicing but won't buy a standalone billing tool |
The best bundle expands the CasualFan market without destroying the standalone willingness-to-pay of SuperFans. If the same loyal users would already buy every component separately, the bundle cannibalizes value instead of creating it.
Bundle Design
| Principle | Rule | Mycelium Application |
|---|---|---|
| Minimize SuperFan overlap | Different core audiences per component | CRM SuperFans and ETL SuperFans are distinct — good bundle candidates |
| Maximize CasualFan overlap | Broad interest across components | Most operators casually need auth, notifications, analytics — bundle these |
| Mixed products beat similar products | Distinct loyal audiences, shared casual appeal | Workflow engine + document intelligence + payments = three different SuperFan bases, one coherent operator experience |
| Make components legible | Users see standalone worth of each part | Show what each capability costs individually so the bundle feels like a deal |
Revenue Allocation
Marginal Churn Contribution — not usage — is the fairness test inside the bundle. The question: "what would customers cancel if this piece disappeared?" A capability that gets low clicks but prevents cancellation (auth, compliance) is worth more than a high-usage feature customers could replace.
| Metric | What It Measures | Why It Matters |
|---|---|---|
| Marginal Churn Contribution | How much churn increases if this component is removed | Determines revenue share inside the bundle |
| Retention impact | Does the bundle reduce overall churn vs standalone? | The primary success signal — not raw usage |
| Standalone reference price | What each component costs on its own | Keeps the bundle legible and prevents black-box pricing |
AI Bundle Logic
AI changes bundling economics. When users buy outcomes, coordination, and convenience rather than isolated compute, subscription logic beats usage logic.
- Bundle AI capabilities as a coordinated system. Most operators will not buy agent orchestration, document intelligence, and predictive analytics separately. They buy "AI-powered operations."
- Existing subscriptions are part of the product. AI authenticates and personalizes through services a user already pays for — the bundle includes what they already own.
- AI increases bundle coherence. If the AI layer makes combined products feel more unified, bundling gets stronger. If it doesn't, the components should stay standalone.
Bundle Tests
| Test | Pass | Fail |
|---|---|---|
| Good bundle | More users adopt, more users stay, at least one product becomes easier to justify | — |
| Bad bundle | — | Users can't tell what they're paying for, heavy users would have bought everything anyway, removing one component barely affects churn |
| Pricing | Standalone reference prices visible, bundle feels like a deal | Black box — no legibility |
| Portfolio | Organized around jobs to be done, not feature silos or org charts | Organized by internal team structure |
Applied: Mycelium Bundles
The ventures are the proving ground for bundle design. Each venture is a bundle composed from the Mycelium. The long-term play: dogfood the platform by running ventures on it, then package the same capabilities as RaaS for any business in any vertical.
| Bundle Tier | What's Included | Target | Pricing Model |
|---|---|---|---|
| Foundation | Auth, notifications, analytics, file management | Every venture and every RaaS customer | Subscription (low per-seat) |
| Operations | CRM, workflow engine, task management, embedded payments | Operators running a business in any vertical | Subscription + payments % |
| Intelligence | AI orchestration, document intelligence, ETL pipelines, predictive analytics | Operators who need AI-powered decision support | Subscription + usage hybrid |
| Infrastructure | DePIN telemetry, IoT device management, geospatial, blockchain | Operators deploying physical infrastructure | Outcome-based (per device/per asset) |
Each tier has distinct SuperFans. Foundation SuperFans are developers. Operations SuperFans are business operators. Intelligence SuperFans are data teams. Infrastructure SuperFans are hardware deployers. CasualFan overlap is high across the first two tiers — most businesses casually need auth and CRM even if neither is their core job.
Context
- VSaaS Playbook — The 5P framework for building vertical software
- RaaS Catalog — The demand map: every feature scored by ROI
- Vertical RaaS Functions — Vertical opportunity matrix with capability gap analysis
- Commissioning Dashboard — Current platform coverage by feature
- Products — The V × R framing (VSaaS × RaaS)
- DePIN — Physical infrastructure layer
- Industries — Data footprint value × AI readiness × evolution phase
- Business Models — Other model patterns
- Data Flow — Clean, fast, open data principles
- Idea Discovery — Forces → opportunities framework
Links
- Luke Sophinos — The 2026 Vertical Report — 2025 vertical AI capital share grew from 19% (Q1) to 42% (Q4). Manufacturing 57% AI-native (highest). PE-backed acquisitions dominated exits. Top 10 exits = 60% of $132.7B total exit value.
- Shishir Mehrotra — Rethinking Bundling in the Age of AI — SuperFan/CasualFan framework, Marginal Churn Contribution, AI bundle coherence
- Shishir Mehrotra — Four Myths of Bundling — Retention over usage, churn contribution as revenue allocation fairness test
- Shishir Mehrotra — Bundling Masterclass — Start with audience structure, not price
- Shishir Mehrotra — Practica — Minimize SuperFan overlap, maximize CasualFan overlap, keep components legible
- lukesophinos.com — VSaaS research
- Vertical SaaS Bible
- a16z Services Led Growth
Questions
If you run the Mycelium as in-house software for your own ventures first, which capabilities become the most defensible RaaS products — the ones you use most, or the ones that would cause the most churn if removed?
- Which Mycelium capabilities have distinct SuperFan bases — and which ones are everyone's CasualFan feature bundled in for free?
- When AI makes the combined Mycelium feel more unified than standalone tools, does that strengthen the bundle or make individual capabilities harder to price?
- If 2-3 players capture 70-90% of every vertical, what determines whether the Mycelium becomes one of those players or the substrate they all run on?
- What's the Marginal Churn Contribution of embedded payments vs AI orchestration — which one would operators cancel over?