Online Tutoring vs In-Person Training Centers in China: Which Education Delivery Approach?

Article ID: CG360-EDUCATION-COMP-026  |  Category: Education Sector Comparison  |  Industry: China Private Tutoring (K-12 & Post-Secondary)

1. Overview of China’s Education Delivery Landscape Post-Double-Reduction Policy

For decades, China’s private tutoring sector — known locally as buxi (补习) — was one of the most explosive growth stories in global education. Before 2021, the market was estimated at over ¥800 billion (≈$110 billion USD), fueled by relentless academic competition, the high-stakes gaokao university entrance exam, and a culture of intensive shadow education. Parents in every tier of city invested heavily in after-school tutoring, and both online and in-person delivery models flourished side by side.

The landscape was reshaped irrevocably in July 2021 when the Chinese central government unveiled the “Double Reduction” policy (双减, shuāng jiǎn) — an unprecedented regulatory intervention aimed at reducing the burdens of excessive homework and after-school tutoring on students in compulsory education (Grades 1–9). The policy effectively banned for-profit tutoring in core academic subjects (Chinese, math, English, physics, chemistry, etc.) during weekends, national holidays, and winter/summer breaks for K-9 students. It also capped class durations, forbade超前教学 (advanced teaching ahead of curriculum), and required all remaining tutoring to take place on weekdays before 8:30 PM for in-person and 9:00 PM for online.

This seismic shift forced a dramatic re-evaluation of both online and in-person tutoring models. Many of the largest players — including New Oriental (新东方), Gaotu TeChedu (高途), and Zuoyebang (作业帮) — saw their market capitalizations collapse by 90% or more. Tens of thousands of training centers shuttered, and millions of tutors lost their jobs. However, the policy did not eliminate the demand for educational support; rather, it forced the industry to transform into new forms: non-profit tutoring, quality-education (素质教育) courses in arts and sports, adult education, study abroad services, and intelligent learning devices.

Today, the debate between online and in-person delivery is more nuanced than a simple preference for convenience versus tradition. Regulatory treatment, cost structures, teacher qualification pathways, and geographic penetration all differ substantially between the two models. This article provides a comprehensive comparison of the two approaches across all critical dimensions, with up-to-date analysis through mid-2026.

2. Regulatory Landscape: How the Double-Reduction Policy Affects Both Models

2.1 Impact on In-Person Training Centers

In-person training centers bore the immediate brunt of the policy. All for-profit tutoring for K-9 academic subjects was banned outright. Existing centers that wished to continue operating academic tutoring were required to re-register as non-profit entities (非营利性机构) under the supervision of local education bureaus. They faced strict caps on pricing — many local governments imposed per-class price ceilings (e.g., ¥30–70 per hour depending on city tier and subject). Additionally, centers could not operate during weekends, holidays, or vacation periods, effectively reducing their available teaching hours by 70–80%.

Physical centers that pivoted to quality education (素质教育) — such as art, music, dance, coding, chess, and physical education — faced lighter regulation. These subjects are explicitly encouraged by the government. However, they still require business licenses, fire safety inspections, and teacher qualification documentation at the district level.

2.2 Impact on Online Tutoring Platforms

Online tutoring platforms were arguably hit even harder. The policy specifically restricted online tutoring more tightly than in-person in several respects:

• Time windows: Online tutoring for K-9 is limited to weekdays between 6:00 PM and 9:00 PM — a narrower window than in-person (which can run until 8:30 PM).
• Overseas listings: Many online tutoring companies had listed on U.S. stock exchanges (NYSE, NASDAQ). The government imposed a new regulation requiring all education-tech platforms holding student data to undergo cybersecurity review before listing abroad, effectively freezing IPOs.
• Content review: Online platforms must store all tutoring session recordings for at least one year and submit to random real-time inspections by government monitors.
• Foreign involvement: Foreign investors are prohibited from holding stakes in for-profit K-9 tutoring platforms, and foreign nationals cannot serve as teachers for K-9 academic subjects.

2.3 The Non-Pivot to Adult and Vocational Education

Both online and in-person providers have pivoted aggressively to adult education, vocational training, and study-abroad services, where regulations are far more relaxed. Adult English training, professional certification exam prep (CPA, CFA, legal exams), and digital skills training (coding, AI, data science) are all booming segments. Online platforms have an edge in this space because adult learners prefer asynchronous, on-demand content, while in-person centers still dominate in hands-on vocational skills (cooking, hairdressing, auto repair).

3. Licensing Requirements for Each Model

3.1 In-Person Center Licensing

Operating an in-person training center in China requires a multi-step licensing process:

• Education Business License (办学许可证): Issued by the local district Education Bureau (教育局). Requires proof of suitable premises (minimum floor area varies by city, typically 200–500 sqm), fire safety certification, and compliance with sanitary standards.
• Non-Profit Registration (民办非企业单位登记证书): For academic tutoring, the entity must register as a non-profit with the Civil Affairs Bureau (民政局).
• Commercial Business License (营业执照): For quality-education training, a standard for-profit business license suffices, plus a training-related business scope.
• Faculty review: All teaching staff must submit qualifications (教师资格证) and background checks to the local bureau. Foreign teachers require a valid work visa (Z-visa) and a Foreign Expert Certificate.
• Annual inspections: Centers must pass annual reviews by the local Education Bureau, including financial audits and classroom observation reports.

3.2 Online Platform Licensing

Online tutoring platforms face a different and arguably more complex licensing regime:

• ICP License (ICP备案/许可证): Required for any website or app operating in China. A commercial ICP license (ICP证) is needed if the platform charges fees.
• Online Education Filing (在线教育备案): Platforms must file with the Ministry of Education (MOE) and provide details on course content, teaching methods, data storage locations, and server infrastructure.
• Cybersecurity Multi-Level Protection Scheme (信息安全等级保护, MLPS 2.0): Online tutoring platforms handling student data must achieve MLPS Level 2 or Level 3 certification — a technical and administrative audit that can take 6–12 months and cost ¥200,000–¥1,000,000 depending on scale.
• Data localization: All student and teacher data must be stored on servers physically located within mainland China. Cross-border data transfer is severely restricted.
• App store approval: Mobile apps must be approved by the relevant Android app stores and Apple’s App Store China, which now require proof of education filings before listing.

4. Cost Structure Comparison

The cost divergence is stark. In-person centers are capital-intensive with high fixed costs (rent, facilities) but lower technology overhead. Online platforms have near-zero marginal delivery cost per student but require heavy upfront technology investment and ongoing customer acquisition spending. Post-double-reduction, the cost advantage has shifted somewhat: online platforms can no longer offset their high tech costs with massive student volumes because K-9 class size is capped at 30 students online (and 50 offline), limiting the scalability benefit.

5. Teacher Qualification Requirements

5.1 In-Person Center Teachers

Teachers at in-person training centers must hold a Teacher Qualification Certificate (教师资格证) issued by the Ministry of Education. Since the double-reduction policy, enforcement has tightened: all tutoring staff for academic subjects must display their certificate in the classroom and have it on file with the local education bureau. In practice, many in-person centers in lower-tier cities historically operated with uncertified teachers, but post-2021 compliance raids have led to substantial fines and forced closures.

Foreign teachers at in-person centers require:

• A valid Z-visa (work visa) with a residence permit.
• A Foreign Expert Certificate issued by the State Administration of Foreign Experts Affairs.
• A bachelor’s degree (minimum) and at least two years of teaching experience, or a teaching certification (TEFL, TESOL, CELTA — typically 120+ hours).
• A criminal background check from both China and their home country.

5.2 Online Platform Teachers

Online platform teachers face similar base requirements but with some additional constraints:

• Must hold the same Teacher Qualification Certificate for academic subjects. Since 2022, the MOE has required online platforms to submit a full roster of teachers with certificate numbers for cross-verification against a national database.
• Foreign teachers for online academic tutoring: Effectively banned for K-9. The 2021 regulations prohibit foreign nationals from providing online tutoring in compulsory education academic subjects. For non-academic subjects (e.g., spoken English, coding, art), foreign teachers are allowed but must be physically present in China and hold a work visa — purely overseas-based foreign tutors are no longer permitted for Chinese K-12 students.
• Qualification display: Online teachers must display their credentials on the platform and during the first minute of each recorded session.
• Background checks: Platforms must conduct criminal background checks on all teachers through official Chinese channels; many platforms now use AI-powered face recognition to verify teacher identity at log-in.

6. Student Recruitment and Marketing Channels

6.1 In-Person Center Marketing

Physical training centers in China rely heavily on local, community-based marketing:

• School proximity and partnerships: Centers located near key primary and middle schools benefit from foot traffic. Some form informal referral relationships with school teachers (though this is officially discouraged post-double-reduction).
• Open houses and trial classes (试听课): Free 30–60 minute demo classes are the most effective conversion tool. A strong trial class can yield conversion rates of 30–50%.
• Parent referral programs: “Bring a friend” discounts remain highly effective in Chinese communities, where word-of-mouth trust is paramount.
• WeChat groups: Nearly every in-person center operates a constellation of parent WeChat groups (often 10–50 groups of 500 parents each), used for announcements, homework sharing, and relationship building.
• Local advertising: Banners, billboards near schools, and ads on local community WeChat Official Accounts.

6.2 Online Platform Marketing

Online platforms operate at a national or regional scale with digital-first marketing:

• SEO/SEM (搜索引擎营销): Dominant spend on Baidu, Sogou, and 360 Search for keywords like “小学数学辅导” (elementary math tutoring) or “英语一对一外教” (one-on-one English with foreign teacher). CPC (cost-per-click) in competitive keywords ranges from ¥5–50.
• Short-video platforms (Douyin / TikTok China, Kuaishou): Massive investment in education KOLs and live-streaming sales. Top education influencers can generate 10,000+ leads in a single 3-hour live session.
• Information flow ads (信息流广告): On WeChat Moments, Toutiao, and Xiaohongshu (RED). Precise targeting by location, age of child, school type, and browsing behavior.
• Affiliate and referral programs: Multi-tier referral rewards where existing parents earn commission for successful sign-ups — similar to multi-level marketing structures, though subject to increasing regulatory scrutiny.
• Content marketing: Free study resources, exam tip videos, mini-courses on Bilibili and WeChat Video Channel to build brand authority.

7. Geographic Reach: National vs Local

Perhaps the single most important strategic difference between the two models is their geographic footprint.

In-Person Training Centers are inherently local. A single center serves a catchment area of 2–5 km radius in dense urban areas, or perhaps 10 km in suburban areas. To achieve national coverage, a brand must open dozens or hundreds of franchise or chain locations — a capital-intensive, slow process. This creates natural protection: a well-run local center faces limited competition from distant rivals. However, it also means that market size is capped per location. The largest in-person chain, New Oriental, operates around 700+ learning centers across China — but that took over 28 years to build.

Online Tutoring Platforms can achieve national reach from day one. A platform based in Beijing can serve a student in rural Gansu province as easily as one in Shanghai. This was especially transformative for students in Tier-4 and Tier-5 cities and rural areas, where quality in-person tutoring was scarce or unavailable. Before the crackdown, platforms like Zuoyebang and Yuanfudao (猿辅导) boasted over 400 million registered users across all of China’s 31 provinces.

8. Revenue and Scalability Differences

8.1 Revenue Model Comparison

8.2 The Scalability Trade-Off

Online platforms have inherently superior scalability because their marginal cost of serving an additional student approaches zero (especially for recorded/pre-recorded content). A single star teacher on an online platform can reach 10,000 students in a live session — something impossible in a physical setting. This is why venture capital poured over ¥100 billion into Chinese edtech between 2016 and 2021, chasing the hope of software-like margins.

However, the double-reduction policy fundamentally undercut this scalability story for K-9 academic tutoring by:

• Capping class sizes at 30 for live online sessions.
• Banning the sale of pre-recorded academic courses to K-9 students (all live must be interactive with real-time teacher presence).
• Prohibiting the use of foreign teachers for K-9 academic subjects, which had been a key value proposition for premium online platforms.

In response, surviving online platforms have pivoted to adult education, AI-powered adaptive learning tools, smart hardware (e.g., learning tablets with pre-loaded content), and subscription-based study apps — all of which escape the class-size and time-window restrictions. These new models have restored some scalability, though margins remain thinner than the pre-2021 heyday.

9. Technology Requirements and Platform Development

9.1 In-Person Center Technology

Traditional in-person centers have modest technology needs:

• Classroom management software (CRM) for scheduling, attendance, and billing.
• Smart boards or projectors for interactive teaching.
• WeChat mini-program or simple app for homework distribution and parent communication.
• Some centers now use AI-powered quizzes and adaptive homework platforms (e.g., Squirrel AI’s adaptive engine) as a hybrid enhancement.

Technology spend is typically 2–5% of revenue. Many centers outsource their CRM and mini-program development to local IT vendors for ¥50,000–200,000 upfront.

9.2 Online Platform Technology

Online platforms require a significantly more sophisticated technology stack:

• Video conferencing infrastructure: Low-latency, real-time interactive video supporting whiteboards, screen sharing, breakout rooms, and recording. Most platforms use WebRTC with custom signaling servers, or license technology from Agora (声网) — which itself was built on the back of China’s edtech boom.
• AI-powered features: Speech recognition for spoken English assessment, handwriting recognition for math problem solving, facial emotion tracking to measure student engagement, and adaptive learning path algorithms.
• Backend systems: User management, billing/subscription engine, content management system (CMS), analytics dashboards, and data warehousing — all compliant with China’s Personal Information Protection Law (PIPL, 个人信息保护法) and the Data Security Law (数据安全法).
• CDN & Edge Computing: To deliver smooth video across China’s fragmented ISP landscape, platforms must deploy content delivery nodes in at least 5–10 major cities. Latency requirements for real-time tutoring (under 200ms) make this non-trivial.
• Security & Compliance: MLPS 2.0 Level 2 or 3 certification requires investment in firewalls, intrusion detection systems, encryption (both at-rest and in-transit), access controls, and regular penetration testing.

Total technology investment for a production-grade online tutoring platform is ¥5–50 million upfront, with ongoing infrastructure costs of ¥1–10 million per month at scale. This is a massive barrier to entry that favors well-funded incumbents.

10. Parent/Student Preferences in Different Tiers of Cities

Parent preferences for online vs. in-person tutoring in China vary dramatically by city tier, income level, and the age of the child.

10.1 Tier-1 Cities (Beijing, Shanghai, Guangzhou, Shenzhen)

• Strong preference for in-person among high-income families. Parents view in-person as more rigorous, with better accountability and social development. Many premium in-person centers charge ¥500–1,000 per hour and operate at near-full capacity.
• Online tutoring is used as a supplement — for specialized subjects (e.g., advanced math Olympiad, spoken English with native speakers) or as a backup when schedules conflict.
• Parents in Tier-1 cities are highly educated and discerning; they demand credentialled teachers and transparent pricing.

10.2 Tier-2 Cities (Chengdu, Hangzhou, Wuhan, Nanjing, Xi’an, etc.)

• Split preference. Upper-middle-class parents increasingly prefer in-person, mirroring Tier-1 trends. However, middle-income and commuting families embrace online for its convenience and lower cost.
• Online tutoring penetration is high in subjects where local teachers are perceived as weak, such as spoken English or advanced STEM topics.
• Price sensitivity is higher than Tier-1; parents actively compare per-class costs between online and in-person options.

10.3 Tier-3 and Tier-4 Cities (Luoyang, Linyi, Mianyang, etc.)

• Online tutoring was the dominant growth story before 2021 because the supply of quality in-person tutors was limited. Platforms like Yuanfudao and Zuoyebang gained explosive traction here.
• Post-policy, many online options have withdrawn or shifted to adult/VET, leaving a vacuum. Local in-person centers have partially filled the gap with smaller-scale operations.
• Parents are highly price-sensitive (¥30–80/class is the typical acceptable range) and rely heavily on WeChat-based word-of-mouth for recommendations.

10.4 Rural Areas

• In-person tutoring is virtually non-existent outside of county seats. Online tutoring — via smartphone — was the only realistic option, and its curtailment has been deeply felt.
• The government has attempted to fill the gap with free online public education resources through the National Smart Education Platform (国家中小学智慧教育平台), but adoption remains uneven due to internet connectivity issues and lack of parental guidance.

11. Case Examples of Successful Models

11.1 Successful In-Person Model: New Oriental’s Quality Education Pivot (新东方素质教育转型)

New Oriental (NYSE: EDU), China’s oldest and largest private education provider, faced an existential crisis in 2021. Its core K-12 after-school tutoring business — which generated over 60% of its ¥4.2 billion quarterly revenue — was effectively outlawed. Under founder Michael Yu Minhong (俞敏洪), the company executed a dramatic pivot. It closed hundreds of tutoring centers but repurposed many into quality education centers offering courses in: Chinese calligraphy, traditional painting, programming for kids (少儿编程), public speaking, and physical fitness. It also leaned heavily into its well-established study-abroad consulting and adult English training businesses.

By 2025, New Oriental had stabilized revenue at approximately 60% of pre-policy levels, with quality education contributing 35% of total revenue. The key lesson: in-person centers with strong brand equity, diversified offerings, and physical assets (real estate) can survive a regulatory catastrophe by re-deploying their facilities toward encouraged subjects.

11.2 Successful Online Model: Gaotu TeChedu’s Pivot to Adult and Vocational (高途转型成人教育)

Gaotu TeChedu (NYSE: GOTU), formerly GSX Techedu, was a pure-play online K-12 tutoring platform that grew to over ¥7 billion in quarterly revenue before the double-reduction policy. When the ban hit, Gaotu’s stock lost 95% of its value. The company slashed its K-12 team from 20,000+ employees to under 2,000 and made a decisive pivot to adult online education: professional certification courses (CPA, CFA, civil service exam preparation), digital skills (Python, data analytics, AI), and university-level supplementary courses.

By mid-2025, Gaotu had reached cash-flow positive operations with quarterly revenues of ¥800–900 million — roughly 12% of its pre-ban peak but achieved with a lean cost structure. The company’s online platform, technology stack, and data-driven marketing engine were directly transferable to the adult education market. The key lesson: online platforms that invested in their technology infrastructure and data capabilities can pivot their delivery model to serve a different demographic without rebuilding from scratch.

11.3 Hybrid Model: Xueersi / TAL Education’s AI Learning Devices (好未来智能学习硬件)

TAL Education (好未来, parent of the Xueersi / 学而思 brand) took a different path. It developed AI-powered learning tablets (学而思智能学习机) that combine pre-loaded curriculum content, adaptive practice problems, and AI-driven personalized study plans — all in a device that does not require a live human tutor. These devices are sold through e-commerce channels (JD.com, Taobao) and in physical retail stores. The hardware bypasses the double-reduction restrictions because it is classified as a learning tool, not a tutoring service.

TAL’s smart device business generated over ¥2 billion in revenue in fiscal 2025, with gross margins exceeding 50%. This hybrid approach — leveraging the company’s curriculum expertise and brand name, delivered through technology rather than live instruction — represents the most innovative survival strategy in the post-double-reduction era.

12. Future Trends and Regulatory Outlook

12.1 Continued Regulatory Tightening

The central government has shown no sign of loosening the double-reduction policy. To the contrary, enforcement has steadily expanded: in 2023, the MOE launched a nationwide crackdown on “disguised tutoring” (变相补课) — including one-on-one home tutoring, study groups disguised as “study rooms,” and tutoring bundled with travel or summer camps. In 2024–2025, several provinces began requiring all private tutoring advertisements (online and offline) to be pre-approved by education authorities. Expect further tightening of data privacy requirements and teacher qualification enforcement through 2027.

12.2 The Rise of AI-Native Education

The most disruptive trend is the acceleration of AI-native education tools. Since 2023, the rapid adoption of large language models (LLMs) — including Chinese-developed models like Baidu’s Ernie Bot (文心一言), Alibaba’s Tongyi Qianwen (通义千问), and startups like Zhipu AI (智谱AI) and Minimax — has enabled a new generation of AI tutoring systems that can hold natural-language conversations, explain concepts adaptively, and grade open-ended assignments. These systems are not subject to the same regulations as human-led tutoring and represent the primary growth vector for the entire sector.

12.3 Convergence of Online and In-Person

A growing number of successful operators are adopting a blended or OMO (Online-Merge-Offline) model. A typical OMO center might have a physical storefront (for trust-building and local marketing) but deliver most instruction online via a centralized pool of teachers. This captures the best of both worlds: the local trust and low CAC of a physical presence, with the scalability and teacher utilization of an online model. Regulatory treatment of OMO models remains ambiguous, but early adopters in quality education and adult training are showing promising unit economics.

12.4 International Expansion

Chinese education companies are increasingly looking overseas. New Oriental, TAL, and several online platforms are investing in Southeast Asian markets (Vietnam, Indonesia, Thailand) where regulatory environments are more favorable and demand for Chinese-language education and STEM tutoring is growing. This “going out” (出海) strategy is expected to accelerate, with Chinese edtech companies exporting both their curriculum expertise and their technology platforms.

12.5 Focus on Non-Academic and Life-Long Learning

The government’s explicit encouragement of quality education, sports, arts, and vocational training will continue to shape the market. Both online and in-person providers are building offerings in coding for kids, robotics, Chinese traditional arts, mental wellness coaching, and financial literacy for teenagers. The total addressable market for non-academic education in China is estimated at ¥500–800 billion by 2028, and both delivery models will compete for share — with online favored for information-centric skills (coding, language) and in-person favored for hands-on skills (sports, music, painting).

Conclusion: Which Delivery Approach Wins?

There is no universal answer to the question of whether online tutoring or in-person training centers are superior in China’s post-double-reduction landscape. The answer depends on three factors: target demographic, subject matter, and regulatory risk tolerance.

References and Further Reading

• Opinions on Further Reducing the Burden of Homework and After-School Tutoring for Compulsory Education Students (中共中央办公厅 国务院办公厅印发《关于进一步减轻义务教育阶段学生作业负担和校外培训负担的意见》), July 2021.
• Ministry of Education, “Notice on Regulating Online Training for Compulsory Education” (教育部关于规范义务教育阶段在线培训的通知), 2021–2022 series.
• China Private Education Industry Report, iResearch / 艾瑞咨询, 2025 Edition.
• New Oriental Education & Technology Group, Annual Report FY2025 (SEC Filing 20-F).
• Gaotu TeChedu Inc., Quarterly Earnings Release, Q2 2025.
• TAL Education Group, Smart Device Business Update, Investor Presentation, 2025.
• National Bureau of Statistics of China, “Education Expenditure by Urban and Rural Residents” (中国统计年鉴 教育支出篇), 2024.