17zuoye China

17zuoye (literally '17 Homework') was China's largest K-12 online homework and adaptive learning platform, serving over 50 million students, 4 million teachers, and 40 million parents at its peak. Founded in 2011 by Liu Chang (former Sina executive), the platform digitized homework submission, automated grading, and provided AI-powered personalized learning paths. The value proposition was compelling: reduce teacher workload through automated grading, give students instant feedback with gamified learning, and provide parents real-time visibility into academic progress. The 'Why Now' was perfect timing with China's mobile internet explosion (2011-2015), rising middle-class anxiety about education, and government push for education technology. With $585M from tier-1 investors like DST Global and Temasek, 17zuoye became the poster child for Chinese EdTech, achieving unicorn status by 2018. The platform combined homework management, adaptive question banks, live tutoring, and a freemium-to-premium conversion funnel. However, the business model was fundamentally vulnerable: it relied on converting free school users to paid after-school tutoring services, creating a regulatory time bomb in a sector where the Chinese government maintains tight ideological control.

SECTOR Communication Services

PRODUCT TYPE EdTech

TOTAL CASH BURNED $585.0M

FOUNDING YEAR 2011

END YEAR 2021

Discover the reason behind the shutdown and the market before & today

Failure Analysis

17zuoye's death was a regulatory execution, not a market failure. On July 24, 2021, China's State Council issued the 'Double Reduction' policy, banning for-profit...

Market Analysis

The global K-12 EdTech market today is $180B+ and growing at 16% CAGR, but it's fragmented by regulatory environment. In China, the 2021 Double...

Startup Learnings

Regulatory risk is unhedgeable in authoritarian markets: 17zuoye had perfect execution (product, growth, unit economics) but died because it operated in a sector where...

Market Potential

In 2011-2021, the Chinese K-12 after-school tutoring market was $100B+ annually, driven by gaokao (college entrance exam) pressure and one-child policy families investing heavily...

Difficulty

The core technical challenge - adaptive learning algorithms, automated grading, and homework management - is significantly easier today than 2011. Modern LLMs (GPT-4, Claude...

Scalability

17zuoye demonstrated exceptional scalability mechanics before regulatory intervention. The freemium model had near-zero marginal cost for digital homework (pure software), with network effects as...

Rebuild & monetization strategy: Resurrect the company

✓

Pivot Concept

A pure B2B SaaS platform for K-12 schools that automates homework grading and provides teachers with actionable insights, monetized via per-teacher or per-school subscriptions (not parent-paid tutoring). The core insight: 17zuoye's free homework tool was loved by teachers but never monetized directly. Modern rebuild focuses on selling to schools/districts with institutional budgets, avoiding the regulatory risk of parent-paid tutoring. The product uses LLMs (GPT-4, Claude) for automated grading of open-ended responses (essays, math word problems, science explanations), provides real-time dashboards for teachers to identify struggling students, and integrates with existing LMS platforms (Google Classroom, Canvas, Schoology). The wedge is offering a free tier for individual teachers (viral bottom-up adoption), then upselling to school-wide licenses. The moat is integration depth (becoming the system of record for homework data) and switching costs (teachers build their question banks and workflows in the platform). Revenue model: $15/month per teacher (freemium) or $5,000-$50,000 per school per year (enterprise). Target market: US/EU schools first (stable regulation, institutional budgets), then expand to India/Southeast Asia. The key difference from 17zuoye: we never touch parent-paid tutoring, so we avoid the regulatory trap. We're pure infrastructure for schools, not a consumer EdTech play.

Suggested Technologies

Next.js 14 + Vercel (web platform, edge functions for low-latency grading)Supabase (PostgreSQL + real-time subscriptions for teacher dashboards)OpenAI GPT-4 + Anthropic Claude 3.5 (automated grading, feedback generation)LangChain (orchestration for multi-step grading workflows, RAG for curriculum-specific rubrics)Stripe (subscription billing, school invoicing)Google Classroom + Canvas LTI APIs (LMS integrations for viral adoption)Resend (transactional emails for teacher notifications)Vercel AI SDK (streaming responses for real-time grading feedback)PostHog (product analytics to track teacher engagement and feature usage)

•

Execution Plan

Phase 1

Step 1 - Teacher Wedge (Months 1-3): Build a free Chrome extension that integrates with Google Classroom and auto-grades multiple-choice and short-answer homework using GPT-4. Target math and science teachers (objective grading, clear rubrics). Growth loop: teachers share with colleagues when they save 5+ hours/week. Metric: 1,000 active teachers, 80% weekly retention.

Phase 2

Step 2 - Validation and Feedback Loop (Months 4-6): Add essay grading (English, history) using Claude 3.5 with custom rubrics. Launch a freemium web app (free for individual teachers, $15/month for advanced features like custom question banks and parent reports). Conduct 50+ teacher interviews to identify the killer feature for school-wide adoption. Metric: 5,000 teachers, 100 paying subscribers, NPS > 50.

Phase 3

Step 3 - School Sales and LMS Integration (Months 7-12): Build deep integrations with Canvas and Schoology (not just Google Classroom). Launch school-wide licenses ($5,000-$20,000/year based on school size). Hire 2 inside sales reps to close deals with district administrators. Create ROI calculator showing time savings (teachers save 10 hours/week = $15,000/year in labor costs per teacher). Metric: 50 school contracts, $500K ARR.

Phase 4

Step 4 - Moat and Expansion (Year 2+): Become the system of record for homework data by adding features schools can't live without: predictive analytics (identify at-risk students before they fail), curriculum alignment (map homework to state standards), and parent portals (read-only access to student progress). Expand to India and Southeast Asia with localized content. Build a marketplace for teacher-created question banks (take 20% commission). Metric: $5M ARR, 500 schools, 50,000 teachers, path to $50M ARR within 5 years.

•

Monetization Strategy

Three-tier model: (1) Free Tier - Individual teachers get basic auto-grading for up to 100 students, limited to multiple-choice and short-answer questions. This is the viral wedge. (2) Teacher Pro ($15/month or $150/year) - Unlimited students, essay grading, custom rubrics, parent reports, priority support. Target: 10-20% conversion from free tier. (3) School/District Enterprise ($5,000-$50,000/year) - School-wide licenses with SSO, LMS integration, admin dashboards, dedicated onboarding, and SLA guarantees. Pricing scales with school size (small schools $5K, large districts $50K+). Additional revenue streams: (a) Marketplace - Teachers sell question banks, we take 20% commission (similar to Teachers Pay Teachers model). (b) API Access - Sell grading API to other EdTech companies at $0.10 per graded response. (c) Data Licensing - Aggregate anonymized homework data to sell insights to curriculum publishers (with strict privacy controls and opt-in consent). Target unit economics: CAC $500 (bottom-up viral + inside sales), LTV $3,000 (3-year average retention for schools), LTV:CAC ratio of 6:1. Gross margins of 80%+ (pure software, LLM costs are $0.01-$0.05 per graded response). Path to $50M ARR within 5 years with 500-1,000 schools.

Descargo de responsabilidad: Esta entrada es un resumen y análisis asistido por IA derivado únicamente de fuentes disponibles públicamente (noticias, declaraciones de fundadores, datos de financiación, etc.). Representa patrones, opiniones e interpretaciones con fines educativos, no hechos verificados, acusaciones o asesoramiento profesional. La IA puede contener errores o 'alucinaciones'; todo el contenido es revisado por humanos, pero se proporciona 'tal cual' sin garantías de exactitud, integridad o fiabilidad. Renunciamos a toda responsabilidad por la confianza o el uso de esta información. Si usted es un representante de esta empresa y cree que alguna información es inexacta o desea solicitar una corrección, haga clic en el Descargo de responsabilidad botón para enviar una solicitud.