Skip to main content

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 ControlAdvantage
Data sourceProprietary insights
WorkflowSwitching costs
PaymentsRevenue per customer 10x
DistributionCustomer acquisition

The thesis: Vertical integration from sensor to workflow creates compounding moats.


Three Expressions

Vertical integration appears in different forms:

ExpressionWhat It IsDig Deeper
Vertical AIAI trained on industry-specific dataBelow
Vertical SaaSSoftware tailored to industry workflowsVSaaS Playbook
Full-Stack StartupsOwn the entire value chaina16z

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:

  1. Deep iteration on prompts and workflows
  2. Deep integration of business domain expertise
  3. 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:

ComponentWhat To LearnLink
Strategy9 Keys to MonopolizingPrinciples
Tech StackDePIN + Embedded FinancePlatform
ImplementationDevelopment PlaybookProtocols
MetricsACV Tiers, Focus MatrixPerformance
TeamIndustry EngineersPlayers

Full-Stack Verticals

Industries where vertical integration matters most:

VerticalWhyLink
ConstructionFragmented, physical + digitalHigh opportunity
Real EstateHigh transaction value, data-richHigh opportunity
HealthcareCompliance, integration complexityMedium opportunity
ManufacturingSupply chain, automationMedium opportunity
EducationFragmented, outcome measurementGrowing opportunity

See Full-Stack Startups for more verticals.


Market Examples

Who's winning through vertical integration:

CategoryDominant PlayersMarket Share
Restaurant ManagementToast, Square~70% combined
Construction ManagementProcore, Autodesk~80% combined
Real Estate BrokerageCoStar Group~80% in CRE data
Auto DealershipCDK Global, Reynolds~90% combined
Legal PracticeClio, MyCase~75% combined
Dental PracticeDentrix, Eaglesoft~80% combined
Fitness/SpaMindbody~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 CompetitionHigh Competition
High Data ValueBuild here — first-mover advantageDifferentiate on AI/workflow
Low Data ValueAvoid — no moatRace to bottom

The process:

  1. Idea Discovery — Identify forces creating friction
  2. Industries — Deep-dive 5P analysis per vertical
  3. Vertical Integration (you are here) — Strategy to capture the value chain
  4. 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.

ScoreWhat It MeasuresSource
Data ValueHow valuable is the vertical's data footprint for AI training and predictions?Industry Matrix
Feature CoverageHow many of the vertical's required functions already exist in the Mycelium?RaaS Catalog
Capital SignalIs 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.

VerticalDataAIFeature CoverageCapital SignalBest PlayWhy
Healthcare55Low (compliance gap)Strong (597 deals, $11.8B, led exits)Sell RaaSRegulatory moat too deep to operate without domain team. Sell compliance-ready tooling.
Manufacturing43Low (IoT gap)Accelerating (41% deal growth, 57% AI-native)Sell RaaSCapital pouring in, operators need workflow + sensor tools.
Construction32Medium (project mgt partial)ModerateSell RaaSFragmented SMBs buy tools. Data moat + payments = Levelset pattern ($500M exit).
Real Estate43Medium (6 PropTech tracks mapped)Strong ($300T+ market)BothSell property management RaaS AND operate tokenized RE ventures.
Energy54Low (DePIN gap)GrowingOperateDePIN device deployment = data moat. Own the sensors, own the predictions.
Education34LowEmerging (highest early-stage step-ups 2.5x)OperateLow capital intensity ($200K/employee). Build credentialing on the platform.
Legal34Medium (doc intelligence exists)Hot (GC AI $555M, 50x multiples)Sell RaaSPLG possible. Fastest path to $1M ARR.
Advertising45LowSaturatedOperateData 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:

PlayCritical FeaturesWhy
Sell RaaSEmbedded 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
OperateETL 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
BothAll 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:

AudienceDefinitionExample
SuperFansWould buy this capability standalone at full priceCRM power users, compliance-heavy operators
CasualFansWould use it if access were easier or cheaper — bundled inOperators 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

PrincipleRuleMycelium Application
Minimize SuperFan overlapDifferent core audiences per componentCRM SuperFans and ETL SuperFans are distinct — good bundle candidates
Maximize CasualFan overlapBroad interest across componentsMost operators casually need auth, notifications, analytics — bundle these
Mixed products beat similar productsDistinct loyal audiences, shared casual appealWorkflow engine + document intelligence + payments = three different SuperFan bases, one coherent operator experience
Make components legibleUsers see standalone worth of each partShow 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.

MetricWhat It MeasuresWhy It Matters
Marginal Churn ContributionHow much churn increases if this component is removedDetermines revenue share inside the bundle
Retention impactDoes the bundle reduce overall churn vs standalone?The primary success signal — not raw usage
Standalone reference priceWhat each component costs on its ownKeeps 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

TestPassFail
Good bundleMore users adopt, more users stay, at least one product becomes easier to justify
Bad bundleUsers can't tell what they're paying for, heavy users would have bought everything anyway, removing one component barely affects churn
PricingStandalone reference prices visible, bundle feels like a dealBlack box — no legibility
PortfolioOrganized around jobs to be done, not feature silos or org chartsOrganized 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 TierWhat's IncludedTargetPricing Model
FoundationAuth, notifications, analytics, file managementEvery venture and every RaaS customerSubscription (low per-seat)
OperationsCRM, workflow engine, task management, embedded paymentsOperators running a business in any verticalSubscription + payments %
IntelligenceAI orchestration, document intelligence, ETL pipelines, predictive analyticsOperators who need AI-powered decision supportSubscription + usage hybrid
InfrastructureDePIN telemetry, IoT device management, geospatial, blockchainOperators deploying physical infrastructureOutcome-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

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?