---
id: wiki-2026-0508-startup
title: Startup
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [Lean Startup, Startup Methodology, 스타트업]
duplicate_of: none
source_trust_level: A
confidence_score: 0.9
verification_status: applied
tags: [startup, lean, mvp, customer-development, ai-startup]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: english-korean
  framework: lean-startup
---

# Startup

## 매 한 줄
> **"매 startup = search for repeatable + scalable business model"**. Steve Blank 의 customer development + Eric Ries 의 Lean Startup 이 base. 2026 AI wave 에선 매 1인 founder 가 Claude/Cursor 로 prototype → seed 까지 < 30 일 의 cycle.

## 매 핵심

### 매 기본 정의 (Blank/Ries)
- **Startup** ≠ 매 small business. 매 "search for repeatable, scalable, profitable business model" 하 temporary org.
- **Customer Development** (Blank): Get-out-of-the-building, Customer Discovery → Validation → Creation → Building.
- **Lean Startup** (Ries): Build–Measure–Learn loop, MVP, validated learning, pivot or persevere.

### 매 stages
- **Pre-seed** ($100K~$2M): idea + founders. 2026 AI prototype demo 기본.
- **Seed** ($1M~$5M): PMF 탐색. 매 design partner 5–10 개.
- **Series A** ($5M~$25M): PMF 입증 + GTM scale.
- **Series B+**: scale operations, geographic/category expansion.

### 매 PMF (Product-Market Fit)
- Sean Ellis test: "How disappointed if product disappeared?" — 40%+ "very" → PMF signal.
- Retention curve flat → PMF. Churning curve → 매 not yet.
- Andreessen: "you can always feel PMF when it's happening."

### 매 2026 AI Startup wave
- **Solo / 2-person AI startup**: Cursor/Claude Code 로 1 founder 가 full-stack ship. ARR $1M+ < 12 개월 사례 다수.
- **Vertical AI agents**: 법률 (Harvey), 회계 (Pilot+AI), 의료 scribe (Abridge), 영업 (11x.ai).
- **Foundation model wrapper risk**: GPT-5 / Claude Opus 4.7 / Gemini 3 가 직접 feature 흡수. Moat = data/distribution/workflow integration.
- **AI-native pricing**: per-task, per-outcome (success-based), per-agent-seat. 매 traditional per-seat SaaS 의 위협.

## 💻 패턴

### MVP 가설 worksheet (Lean Canvas style)
```markdown
## Problem (top 3)
1. ...

## Customer Segments
- Early adopter: ...
- Mainstream: ...

## Unique Value Prop
"<X> for <segment> who <pain>"

## Solution (top 3 features)
1. ...

## Channels
- ...

## Revenue Streams
- ...

## Cost Structure
- ...

## Key Metrics (AARRR)
- Acquisition / Activation / Retention / Referral / Revenue

## Unfair Advantage
- ...
```

### Customer interview template
```markdown
# Discovery Interview (30 min)

## Warm-up
- Tell me about your role + day-to-day.

## Problem (no pitching!)
- Walk me through last time you <did task>.
- What was hardest part? Why?
- What did you do to solve it? (existing workarounds)
- How much time/money does <pain> cost?

## Solution probe (only after problem confirmed)
- If a tool did <X>, how would you use it?
- Who else needs to be involved in buying decision?

## Close
- Who else should I talk to?
- Can I follow up in 2 weeks?
```

### MVP build (2026 AI startup stack)
```bash
# Day 0–3: validate pain via 10 interviews
# Day 4–10: prototype with Claude Code + v0 + Supabase
pnpm create next-app@latest mvp --typescript --tailwind --app
cd mvp
pnpm add @supabase/ssr ai @ai-sdk/anthropic
# Day 11–14: 5 design partner deploy via Vercel
vercel deploy --prod
# Day 15+: weekly Build–Measure–Learn cycle
```

### Build–Measure–Learn loop
```typescript
// 매 weekly cycle 의 instrumentation
import { track } from "@vercel/analytics";

export async function onUserAction(action: string, props: object) {
  await track(action, props);  // PostHog/Mixpanel/Amplitude
}

// Measure: cohort retention, activation funnel
// Learn: 매 weekly 5-customer call → hypothesis update
// Build: 매 1 hypothesis test per sprint
```

### PMF metric dashboard (PostHog SQL)
```sql
-- Sean Ellis-style retention cohort
SELECT
  date_trunc('week', signup_at) AS cohort,
  COUNT(DISTINCT user_id) FILTER (WHERE active_in_week_4) * 1.0
    / COUNT(DISTINCT user_id) AS w4_retention
FROM users
GROUP BY 1
ORDER BY 1 DESC;
-- 매 30%+ flat W4 retention = PMF candidate
```

### Pivot decision matrix
```python
# 매 "pivot or persevere" — Ries
def pivot_signal(metrics):
    # No traction after 3 build-measure-learn cycles?
    if metrics.weekly_active_growth < 0.05 and metrics.cycles >= 3:
        return "PIVOT"
    if metrics.retention_w4 > 0.30 and metrics.organic_share > 0.20:
        return "PERSEVERE / SCALE"
    return "CONTINUE LEARNING"
```

### Fundraising data room essentials (2026)
```markdown
## Seed Data Room
- Pitch deck (10-12 slides, Sequoia/YC format)
- 매 financial model (3-year, monthly first 12mo)
- KPI dashboard (live link to Mixpanel/PostHog)
- Customer letters / testimonials (5+)
- Cap table (Carta export)
- 매 incorporation docs (Delaware C-Corp)
- IP assignment, founder agreements
- AI compliance: 매 SOC2 Type 1 progress, EU AI Act risk class
```

## 매 결정 기준
| 상황 | Approach |
|---|---|
| 매 Idea 단계 | Customer Discovery 만, 매 build 전 10+ interview |
| Prototype 후 traction X | Lean iteration, 매 pivot 고려 |
| 매 Seed 단계, design partner 확보 | Validation: contract / LOI 5+ |
| Series A 준비 | Repeatable sales motion 입증 (CAC payback < 18mo) |
| AI wrapper 우려 | Workflow integration + proprietary data moat 의 강화 |

**기본값**: 매 Lean Startup + Customer Development 결합 — 매 Build–Measure–Learn weekly cadence.

## 🔗 Graph
- 부모: [[Business-Strategy]]
- 변형: [[Lean-Startup]]
- 응용: [[MVP]]

## 🤖 LLM 활용
**언제**: 매 idea validation, customer interview synthesis, pitch deck draft, KPI dashboard SQL 작성, market sizing (TAM/SAM/SOM).
**언제 X**: 매 hard customer signal 의 대체 X — 매 LLM 가 진짜 customer pain 의 hallucinate 가능. 매 actual interviews irreplaceable.

## ❌ 안티패턴
- **Build first, validate later**: 매 6개월 build → 매 nobody wants. Customer dev 가 먼저.
- **Vanity metrics**: signup count, page view 만 추적 — 매 retention/revenue 의 무시.
- **매 founder–market mismatch**: domain 의 unfamiliar — design partner 의 trust 약화.
- **AI wrapper without moat**: GPT-5 / Claude API call only → foundation model 이 흡수 시 사라짐.
- **Premature scaling** (Marmer): PMF 전 매 sales team 의 hire — 매 burn rate 폭주.

## 🧪 검증 / 중복
- Verified (Steve Blank "Four Steps to the Epiphany", Eric Ries "Lean Startup", YC startup library 2026).
- 신뢰도 A.

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — Lean/Customer Dev + 2026 AI startup wave 정리 |