# Benchmarks — B2B Marketplaces & AI Verticals (Strategic Roadmap §13)

## Direct B2B Marketplace References

### Faire ($12.4B valuation, 5 years)
- Cold-start playbook: **buyers first via outbound sales**, then supply onboards to meet demand
- Supply-side retention: "net terms" (60-day payment for retailers) as moat
- Take rate: 15-25% on gross (high — we plan 3-5%)
- Key metric: GMV, not users

### Ankorstore ($2B, 3 years)
- European Faire clone
- Localization (language + currency + logistics) critical
- Partnered with shipping providers — doesn't own logistics
- Key metric: producer NPS (they overinvest in supplier experience)

### Flexport ($8B, 7 years)
- B2B logistics + trade data
- **Data = moat**: every shipment trains their ML predictions
- Network effects: more shippers → more carriers → better pricing
- Lesson: proprietary data flywheel is everything

### Alibaba 1688 ($500B+ parent)
- Trust layer: **escrow + verification** mandatory
- 20 years to build
- Lesson: Trust infrastructure = switching cost

### Thomasnet (~$500M, declining)
- Old-school industrial B2B directory
- Didn't add AI/modern UX
- **Lesson: Data without AI = death in 2026**

## AI × B2B (2024-2026 Reference Cases)

### Clay.com ($1.5B, 3 years)
- AI sales enrichment — vertical not horizontal
- **Vertical AI > horizontal AI** for B2B
- Pricing: $149-800/mo per user

### Harvey AI ($1.5B, 2 years)
- Vertical AI for legal industry
- Deep domain expertise as differentiator
- Ex-lawyers on team, not just ML engineers
- **Lesson: Domain expertise + AI = defensibility**

### Gong ($7.25B)
- Revenue intelligence — records + analyzes sales calls
- **Data about deals = very valuable**
- Enterprise plays big, PLG for SMB

### Perplexity ($9B, 2 years)
- AI search
- **Lesson: 10x better product can scale fast**
- But: high burn, uncertain moat

## Ukrainian Cases (for inspiration)

### MacPaw (~$50M+ revenue, bootstrapped)
- Never took VC
- **Lesson: If unit economics work, VC not required**

### Grammarly ($13B)
- B2C first, then B2B Enterprise
- **Enterprise expansion = 10x ARR lever**

### Preply ($560M)
- Two-sided marketplace (tutors + students)
- **Network effects in marketplaces are real — but take 5+ years**

### Ajax Systems ($1B+)
- Hardware + SaaS for global market
- Made-in-Ukraine sells internationally
- **Quality + brand > low price**

### Reface (viral AI)
- 100M+ downloads in months
- **One viral mechanic can define a company**
- But: hard to monetize virality → now pivoting

## Frameworks to Know (ORACLE must remember)

### The Cold Start Problem (Andrew Chen)
- Marketplaces need 1 side to exist first
- **Hard side (supply) first is the pattern for B2B**
- Atomic network: smallest subset that provides value to itself

### Blitzscaling (Reid Hoffman)
- Growth > efficiency in early stages
- Only works if market is winner-takes-most
- **B2B Export is more "winner-takes-some" → don't blitzscale blindly**

### Crossing the Chasm (Geoffrey Moore)
- Early adopters ≠ mainstream
- Need different messaging for each group
- **Applies to us at M6-M9**: early producers ≠ late majority

### The Mom Test (Rob Fitzpatrick)
- Don't ask "would you use this?"
- Ask "what do you currently do about X problem?"
- **Apply in every customer call**

### Network Effects Manual (NFX)
- 13 types of network effects
- Our type: 2-sided marketplace
- **Key metric: "density" — enough producers in same category for a buyer**

## Key Benchmarks for Our Investor Deck

| Metric | Our target M12 | Comparable at that stage |
|---|---|---|
| ARR | $10M | Faire M18 = $8M |
| Producers | 300 | Ankorstore M24 = 250 |
| GMV/month | $15M | Faire M18 = $20M |
| Gross Margin | 70% | Marketplace median = 65% |
| LTV:CAC | 3:1 | Good B2B SaaS = 3-5:1 |
| Match rate | 35% | Ankorstore published = 28% |

## Lessons for Corevia GEMALLI

1. **Supply-side first** — producers before buyers (Faire pattern)
2. **Data flywheel** — every match must improve next match (Flexport)
3. **Trust layer obsolete without** — escrow + KYB (Alibaba)
4. **Vertical AI** — narrow focus > horizontal AI (Harvey)
5. **NPS from producers** critical — overinvest in supplier UX (Ankorstore)
6. **Bootstrap-capable unit economics** — don't need VC if numbers work (MacPaw)