## The Deep Dive
_Practical insights for evaluating opportunities and avoiding mistakes._
---
## 1. WHY NOW?
Infrastructure has been falling apart for decades. What changed in 2023-2024?
**Five things converged:**
1. **Edge AI works now** - GPT-4 and edge models make real-time predictions viable. Cost dropped from dollars per inference to <$0.01. Can run AI on $50 edge devices.
2. **Sensors got cheap** - Industrial IoT sensors went from $100+ (2017) to <$10 (2023) at scale. LIDAR costs down 90%. Deployment now pays back in months, not years.
3. **Mobile penetration crossed 70%** in target markets - Inspectors use smartphones instead of clipboards. No legacy IT needed.
4. **Regulations kicked in** - US PFAS standards (2024), lead pipe deadlines (2024-27), infrastructure cybersecurity rules (2023), climate mandates across APAC/MENA. Created budgets and urgency.
5. **Climate forcing the issue** - Insurance costs up 30-40% annually. Extreme weather accelerating degradation 2-3x in hot/humid regions.
**The window is 2024-2026.** Couldn't do this in 2020 (too expensive). Shouldn't wait until 2027 (competition will arrive).
---
## 2. THE WEDGE STRATEGY
Don't try to do everything. Pick one thing, dominate it, expand systematically.
### Phase 1: Beachhead (12-18 months)
- **One** specific pain point
- **One** customer type
- **One** geography
- Prove 10x better, not 10% better
**Good wedges:**
- Water leak detection for Malaysian utilities
- Desalination monitoring in UAE
- Mining equipment in Zambian copper operations
- Bridge inspection in ASEAN corridors
**Bad wedges:**
- "Infrastructure maintenance in Asia" (too broad)
- "AI for utilities" (too vague)
### Phase 2: Expand within customer (18-30 months)
Same customer, adjacent assets. Water utility goes from pipes → wastewater → quality sensors. ACV increases 3-5x.
### Phase 3: Horizontal scale (30-48 months)
Prove it works across geography. Malaysia → Thailand → Indonesia. Use case studies and partnerships.
### Phase 4: Platform (36+ months)
Become the infrastructure OS. APIs, marketplace, data monetization.
**Example progression:**
```
Year 1: Water pipes in Malaysia - $100K ACV
Year 2: + Wastewater + Quality - $250K ACV
Year 3: + Treatment + Pumps - $400K ACV
```
---
## 3. UNIT ECONOMICS THAT WORK
**Revenue model:** Hardware + SaaS hybrid
Year 1: $250-400K (hardware $150-250K + software $100-150K)
Year 2+: $150-250K (software only + expansion)
**Key metrics:**
- Gross margin: 65-75% blended (40% Year 1, 85% Year 2+)
- CAC payback: <12 months (target 6-9)
- LTV/CAC: >5:1 (target 8-12:1)
- Logo retention: 95-100% (utilities don't churn)
- Net retention: 110-130% (they buy more over time)
**Quick example:**
```
Water utility customer:
- Pays $300K Year 1, $175K Years 2-6
- Total LTV: $1.2M
- Your CAC: $75K
- LTV/CAC: 16:1 ✓
- They save $500K/year in water loss
- Their ROI: 2.5x what they pay you
```
**Path to scale:**
- Year 1: $0.5M ARR (10 customers)
- Year 3: $5M ARR (50 customers) - Break even here
- Year 5: $40M ARR (200 customers)
---
## 4. DEFENSIBLE MOATS
Technology commoditizes. Here's what actually defends your position:
### 1. Data network effects (the main one)
Every sensor you deploy trains your AI. After 1,000 deployments, you have patterns competitors need 3-5 years to replicate. Your accuracy improves → customers stay → more data → better AI. This compounds.
### 2. Regulatory capture
Work with government to write the standards. Join committees. Provide reference implementations. Get your methodology written into procurement specs. Suddenly competitors need to be "XYZ-compliant" and you have an 18-month head start.
### 3. Operational lock-in
Embedded in daily workflows. 3-5 years of historical data. Staff trained. API integrations everywhere. Switching costs are $500K-2M and 12-18 months of pain.
### 4. Geographic density
Own 70-80% of utilities in Malaysia before competitors arrive. Local partnerships, language localization, regulatory knowledge. Hard to displace.
### 5. Channel control
Exclusive partnerships with infrastructure contractors. They sell your tech, you leverage their relationships. Takes time to replicate.
**Timeline:** Moats take 24-36 months to build, not Day 1. Plan for this.
---
## 5. HOW TO PICK OPPORTUNITIES
Score each on 7 criteria (2 points = strong, 1 = okay, 0 = weak):
**1. Non-discretionary?** Regulatory mandate or prevents asset failure (not "nice to have")
**2. ROI <12 months?** Can they measure savings clearly and quickly?
**3. Budget exists?** Dedicated line item or creating new category?
**4. Fast validation?** Prove it in 12-18 months with 10 customers, or 5-year development cycle?
**5. Expansion path?** Can you 3-5x revenue per customer over time?
**6. Moat possible?** Data network effects or regulatory capture achievable?
**7. Founder-market fit?** Domain expertise + technical chops + sales ability?
**Scoring:**
- 12-14 points: Do it
- 9-11 points: Maybe
- <9 points: Probably not
---
## 6. WHAT WILL GO WRONG
**Long sales cycles:** Utilities take 18-24 months to buy.
→ Fix: $25K pilot programs (90 days), partner channels, target smaller customers first
**Hardware fails in harsh conditions:** Sensors die in extreme heat/humidity.
→ Fix: Industrial-grade components, redundancy, performance-based contracts, 3-year warranty
**AI misses failures:** Miss a prediction → pipe bursts → liability.
→ Fix: Position as "decision support" not autonomous, humans review recommendations, insurance
**Well-funded competitor:** Someone raises $50M and undercuts pricing.
→ Fix: Focus on outcomes not price, lock-in through data, move fast, partnership model harder to copy
**Regulatory changes:** New rules favor incumbents.
→ Fix: Engage regulators early, join standards committees, get certified, partner with certified contractors
**Budget cuts:** Economic downturn freezes spending.
→ Fix: ROI so good they can't cut it, compliance budget more stable, diversify customers
---
## 7. COMPETITION
**vs. Status quo (70-80% of market):** Manual processes, Excel tracking
→ You win: 10x faster, cheaper at scale, better accuracy
**vs. Legacy tech (15-20%):** Old monitoring companies
→ You win: Modern AI vs rules, mobile-first, faster iteration
**vs. Other startups (5%):** Well-funded new entrants
→ You win: Geographic focus, data head start, better economics
**vs. Big tech (5%):** GE, Siemens adding IoT
→ You win: Faster, better business model, they sell equipment not outcomes
**vs. Tech giants (negligible):** Oracle, Microsoft, Google
→ You win: Not worth their time until $1B+ market (you have years)
**Strategy by phase:**
- Year 1-2: Beat status quo (prove ROI)
- Year 2-3: Defend vs startups (build data moat)
- Year 3-5: Fend off big tech (platform effects)
- Year 5+: Exit to them or IPO
---
## 8. TEAM REQUIREMENTS
**You need three capabilities:**
**Domain expert (CEO):** 7-10 years in the industry. Knows customers, has relationships, credible with buyers. Buyers don't trust tech outsiders.
**Technical leader (CTO):** Sensors/IoT + AI/ML, shipped hardware before. Reliability matters more than innovation in infrastructure.
**GTM expert (CRO):** Sold to utilities/government/industrial before. Understands procurement, RFPs, budget cycles. Can close first 10 customers.
**Red flags:**
- All tech team learning industry from scratch
- No enterprise sales experience
- No hardware experience building IoT
- Remote team selling into foreign markets
**If you have gaps:**
- Missing domain: Recruit advisor with equity, hire from industry, spend 6 months in customer discovery
- Missing sales: Hire experienced VP Sales early, don't DIY
- Missing tech: Partner for hardware, focus on software
- Missing local knowledge: Establish local subsidiary, hire local team, co-locate for 6-12 months
---
## BOTTOM LINE
**Winners will:**
- Pick narrow wedges and dominate (12-18 months)
- Prove 10x better, not incremental (speed, cost, or accuracy)
- Build data moats (24-36 month barrier)
- Target emerging markets (higher growth, less competition)
- Show ROI <12 months (measurable outcomes)
- Have founder-market fit (domain + tech + sales)
- Expand systematically (wedge → platform)
- Hit strong unit economics (LTV/CAC >5:1, margins >65%)
**That's it. Everything else is noise.**