Diligence report Online education, learning apps, and AI-enabled educational content Private late-stage unicorn under post-2021 regulatory transformation

Yuanfudao

Yuanfudao Startup Diligence Report

The core diligence question is whether Yuanfudao successfully transformed a historically high-growth, live-course-heavy K-12 tutoring business into a compliant, durable, AI-led learning platform with defendable economics. Public evidence supports the transformation narrative but not its financial success.

Company profile

Yuanfudao Startup Diligence Report

Yuanfudao was one of the world's most highly valued edtech unicorns at the 2020 peak, with clear public evidence of massive user scale, aggressive funding, a broad AI education product stack, and a large workforce. The current business remains active, but its post-2021 earnings quality and value are obscured by China's tutoring crackdown and by the company's private status.

Website: www.yuanfudao.com
Sector: Online education, learning apps, and AI-enabled educational content
Geography: China
Stage: Private late-stage unicorn under post-2021 regulatory transformation
Known aliases: Yuanfudao, Yuanfudao Group, Beijing Yuanfudao Online Education Organisation, Beijing Yuanli Science and Technology Co., Ltd.
Report version: 1.0
Timezone: UTC

Executive summary

Strengths

Current public sources support an active AI-centric product suite spanning courses, Xiaoyuan AI, Zebra AI Study, programming, and related digital-learning products.
Independent 2018-2020 reporting supports hundreds of millions of users and more than 1 million paid or full-price users at peak scale.
Public evidence strongly supports Yuanfudao's long-standing claim that AI research and personalization are central to the product story.

Risks

China's Double Reduction policy structurally undermined the historical tutoring model.
The best public valuation anchor is still 2020, leaving current value and financial quality opaque.
AI-driven education claims are material to the thesis but lack public evidence on model governance, child-data controls, and learning outcomes.

Gaps

Current revenue, gross margin, cash, debt, refunds, and audited profitability by product line are not public.
Current cap table, employee-equity programs, debt, and any post-2020 priced financing or secondary transactions are not public.
Customer cohort quality, retention, refund behavior, and product-level revenue concentration are not public.
Current headcount, severance, and post-crackdown organizational redesign are not public.
Litigation, insurance, and contract schedules require counsel-led diligence.

Recommended next steps

Obtain audited historicals and a post-2021 revenue bridge by product, channel, and legal entity before relying on the 2020 valuation story.
Run focused regulatory diligence covering product classification, non-profit / for-profit entity boundaries, and all correspondence with Chinese education regulators.
Review AI governance, child-data protection, model training, content rights, and learning-outcome validation for the current product suite.
Verify current headcount, teaching-center footprint, and restructuring costs after the 2021 sector reset.
Rebuild the financing story from the cap table forward, including preferences, debt, and any employee or investor liquidity since 2020.

Risk register

critical high likelihood

R-001: Regulatory business-model reset after Double Reduction

The 2021 Chinese policy framework severely restricted for-profit K-12 tutoring, forcing legal-entity, product-mix, and capital-market changes across the sector.

Diligence request: Review every current business line against current Chinese education regulations, entity structure, licenses, and revenue dependence on any restricted activities.

high high likelihood

R-002: Financial opacity and stale valuation anchors

The best public valuation evidence is from 2020; post-2021 financial statements, margins, burn, debt, and current fair value are not public.

Diligence request: Obtain audited statements, cap table, debt schedule, tax positions, and any post-2020 priced financing or secondary data before underwriting the equity story.

high medium likelihood

R-003: Revenue and product-mix transition risk

Historic revenue depended heavily on live courses, while the current public portfolio emphasizes non-subject and AI-enabled products after the crackdown.

Diligence request: Request revenue and gross margin by product, cohort retention, refunds, and evidence that replacement products can offset the loss of restricted K-12 tutoring economics.

high medium likelihood

R-004: AI quality, safety, privacy, and educational-outcome risk

Yuanfudao heavily markets AI personalization and automation, but public evidence does not provide model-governance detail, child-data controls, or outcome-validation methods.

Diligence request: Review model training, human-review loops, privacy controls, explainability, hallucination testing, and customer redress processes for minors.

medium high likelihood

R-005: Consumer-cohort concentration and retention opacity

The business publishes large user counts and app ratings but no current paying-user concentration, retention, refund, or cohort economics.

Diligence request: Request cohort retention, refund rates, CAC, payback, top product concentrations, and revenue by region and user segment.

medium high likelihood

R-006: Intense competitive pressure in Chinese edtech and AI learning tools

Yuanfudao competes with large, well-capitalized rivals and with broad free/low-cost education tools in a post-crackdown market that is structurally different from 2020.

Diligence request: Benchmark learning outcomes, user acquisition, pricing, teacher quality, and AI functionality against Zuoyebang, VIPKid, Yiqizuoye, and adjacent AI-study products.

medium medium likelihood

R-007: Advertising and compliance enforcement risk

The public record shows app-filing obligations and false-advertising / deceptive-pricing penalties, indicating ongoing regulatory sensitivity.

Diligence request: Review current advertising claims, price-promotion controls, app-compliance filings, and any open regulator correspondence or remediation plans.

medium medium likelihood

R-008: Organizational-complexity and workforce-transition risk

Public sources describe a 30,000-employee footprint in 2020 and teaching centers across China; the post-2021 workforce reset is not public.

Diligence request: Obtain current headcount, attrition, teaching-center utilization, severance obligations, and organizational redesign materials after the crackdown.

Chapter 01

01Financial Information

Public financing disclosures show an aggressive 2018-2020 scaling story culminating in a US$15.5B valuation, but the post-2021 financial model, current valuation, cap table, debt, and profitability are not publicly verifiable.

EC-003 EC-004 EC-005 EC-006 EC-007 EC-008 EC-011 EC-018

I.A Annual and quarterly financial information for the past three years

partially verified confidence: medium

Public sources provide historical user, revenue, and paid-user anchors, but no audited financial statements, AR aging, backlog, or management accounts.

EC-003 EC-005 EC-006 EC-018

Evidence gaps

Audited income statements, balance sheets, cash flows, tax workpapers, refund reserves, and cohort gross-margin data were not public.
No public source reviewed disclosed accounts receivable aging, backlog, or current cash and debt balances.

Hidden risks

Historical public revenue signals predate the 2021 regulatory reset and may not resemble the current earnings mix.
A large private-company operating footprint can hide deferred-revenue, refund, or teacher-cost volatility.

Follow-up questions

Provide audited annual and quarterly financials for the last three years plus year-to-date management accounts and refund reserve schedules.

Public revenue, user, and unit-economic signals
metric	public value	period	caveat
Recognized revenue	RMB 1.5B	2018	Historical figure from venture media; no audited statements or current equivalent disclosed.
Paid / long-term users	>1M paid users in 2018; >1M long-term full-price users in 2020	2018-2020	Definitions differ and do not reveal churn, refunds, or product mix after 2021.
Renewal / conversion	70%-80% renewal and ~42% marketing conversion reported in 2018	2018	Likely not representative of the post-crackdown product and regulatory environment.
Current margins / cash / debt	not publicly verifiable	current	Requires audited financials, balance sheet, debt schedule, and cash-flow detail.

I.B Financial Projections

not publicly verifiable confidence: low

No public forecast package was reviewed, and the 2021 regulatory reset materially weakens the predictive value of pre-crackdown operating metrics.

EC-011 EC-018

Evidence gaps

Three-year product, channel, and geography forecasts were not public.
No source disclosed board-approved capital-expenditure, hiring, or financing assumptions.

Hidden risks

Forecasts may be highly sensitive to product-mix shifts from subject tutoring toward AI and non-subject content.
Current growth assumptions could depend on regulatory interpretations that are not visible in public sources.

Follow-up questions

Provide the post-2021 board plan, sensitivity cases, and management assumptions by product line and entity.

I.C Capital Structure

not publicly verifiable confidence: low

Historical investors are public, but shares outstanding, current ownership, employee equity, debt, and off-balance-sheet liabilities are not.

EC-005 EC-006 EC-007 EC-017 EC-018

Evidence gaps

No reviewed public source disclosed the current cap table, option pool, debt, or guarantees.
No public source reviewed disclosed current shareholder percentages or liquidation preferences.

Hidden risks

Late-stage financing terms, preferences, and any post-2020 secondaries could materially change common-equity value.
Entity conversion and regulatory restructuring after 2021 may have changed ownership architecture.

Follow-up questions

Provide the current cap table, debt schedule, shareholder register, and every financing instrument executed since 2020.

Capital structure and ownership snapshot
stakeholder or class	public position	diligence caveat
Founder / CEO leadership	CEO Li Yong was publicly named in 2020 reporting.	Founder equity, voting rights, and current officer status are private.
Late-stage investors	Tencent, Hillhouse, DST Global, and other funds are publicly disclosed historical investors.	Current ownership percentages, preferences, side letters, and any post-2020 transfers are private.
Employee equity / options	Not publicly disclosed in reviewed sources.	Material to retention, dilution, and transaction outcomes.

I.D Other financial information

verified confidence: high

The public financing path is clear through 2020: valuation increased from above US$3B in 2018 to US$7.8B in April 2020 and US$15.5B in October 2020, after which current valuation became opaque.

EC-003 EC-004 EC-005 EC-006 EC-007 EC-008 EC-018

Evidence gaps

Tax positions, revenue-recognition policy, and post-2020 financing details were not public.
No reviewed source reconciled valuation, entity changes, and the current product mix after 2021.

Hidden risks

The valuation story is stale and may no longer reflect the economics of a post-crackdown education business.
Current debt, tax, and accounting policies were not publicly disclosed.

Follow-up questions

Reconcile all historical funding rounds to current capitalization, tax structure, and present-day revenue composition.

Public funding-round history
date	round	amount	valuation	lead or key investors
2018-12	Growth round	US$300M	>US$3B	Tencent-led; Warburg Pincus, Matrix China, IDG Capital
2020-04	Series G1	US$1B	US$7.8B	Hillhouse Capital; Tencent, IDG Capital, Boyu Capital
2020-10	Series G2 / Series G total	US$1.2B (US$2.2B total Series G)	US$15.5B	DST Global; CITICPE, GIC, Temasek, TBP, DCP, Ocean Link, Greenwoods, Danhe Capital
2021-present	Post-crackdown financing	not publicly verified	not publicly verified	Private diligence required

Yuanfudao funding timeline Timeline of the key public funding milestones.

Public valuation trajectory and current gap Line chart of the publicly disclosed valuation anchors.

Chapter 02

02Products

Yuanfudao publicly presents a broad post-crackdown product stack centered on AI-enabled learning, but pricing, learning outcomes, content economics, and product-level profitability are mostly private.

EC-001 EC-002 EC-004 EC-007 EC-009 EC-014 EC-015 EC-016 EC-018

II.A Description of each product

partially verified confidence: high

The current product portfolio includes Yuanfudao quality courses, Zebra AI Study, Xiaoyuan AI, Yuan Programming, and Haitun AI, while historical sources add live tutoring, a Q&A arm, and homework/problem-checking tools.

EC-001 EC-002 EC-004 EC-007 EC-009 EC-014 EC-015 EC-016 EC-018

Evidence gaps

Product-level pricing, refunds, learning outcomes, gross margins, and retention were not publicly disclosed.
No public roadmap or defect/incident history was reviewed for the current AI product stack.

Hidden risks

The current balance between regulated subject tutoring, non-subject products, and AI utilities is not publicly quantified.
Consumer app popularity does not prove sustainable monetization or educational outcomes.

Follow-up questions

Provide current product matrix with revenue, gross margin, user cohorts, learning outcomes, and regulatory classification by product.

Product and SKU matrix
product	audience	public evidence	risk or gap
Yuanfudao SuYang / course products	Students and parents	Homepage lists Yuanfudao quality-course offerings and digital content products.	Current pricing, regulatory classification, and product-level revenue are not public.
Zebra AI Study	Younger learners / early education	Homepage and app-store metadata show Zebra AI as an active branded product with a large review base.	Outcome quality, retention, and monetization are not public.
Xiaoyuan AI / Xiaoyuan problem-solving tools	Students and parents checking homework and exercises	Homepage and app-store metadata show Xiaoyuan AI as a major utility app with multi-million review volume.	Revenue conversion, data/privacy controls, and content moderation remain opaque.
Yuan Programming / Haitun AI and adjacent digital content	Students seeking broader non-subject or digital learning	Current homepage lists programming and additional AI-learning brands.	Product economics and regulatory treatment are not public.

Pricing and monetization evidence gaps
surface	public status	historical signal	diligence request
Course pricing	Not publicly detailed in reviewed sources	2018 coverage said most revenue came from live courses.	Price lists, refunds, scholarships, discounts, and revenue by course type.
App monetization	Apps are visible in the App Store, but paid conversion is not public	Very large review bases imply significant distribution reach.	Freemium conversion, subscriptions, ARPU, and in-app purchase terms.
Post-2021 product mix	Current site shows non-subject and AI products	Regulatory shock implies monetization changed materially after 2021.	Revenue mix before and after the crackdown, by legal entity and product.

Current product and dependency architecture High-level public view of Yuanfudao's product stack and AI layer.

Chapter 03

03Customer Information

Public evidence supports very large consumer reach and app adoption, but top-customer revenue, churn, cohort quality, and supplier spend are not publicly available.

EC-003 EC-005 EC-006 EC-007 EC-014 EC-015 EC-016 EC-018

III.A Top customers by application

partially verified confidence: high

Yuanfudao is a consumer education business, so the best public customer proxies are user counts, paid-user counts, and app-review volume rather than named enterprise accounts.

EC-003 EC-005 EC-006 EC-007 EC-014 EC-015 EC-016 EC-018

Evidence gaps

No public top-15 customer or household schedule was available.
No public source broke users down by paid status, city tier, or learning outcome cohort.

Hidden risks

Large user numbers mix different products, periods, and user definitions, making concentration and monetization hard to infer.
App ratings are a trust and reach signal, not a revenue or retention metric.

Follow-up questions

Provide current and historical paid-user cohorts, segment retention, refund rates, and product-level revenue concentration.

Public user and customer-scale evidence
signal	public value	source period	verification note
Total users	>200M users in 2018; >400M users in 2020	2018-2020	User definitions vary across sources and are not the same as current paying customers.
Paid / full-price users	>1M paid users in 2018; >1M long-term full-price users in 2020	2018-2020	No current public equivalent and no churn or refund detail.
Current app trust / adoption proxy	113,550 Yuanfudao ratings; 3,140,579 Xiaoyuan AI ratings; 344,123 Zebra AI ratings	2026 fetch	Ratings are not the same as active users or paying customers.

Public scale anchors for users and app trust Bar chart of publicly visible scale anchors.

III.B Strategic relationships

partially verified confidence: medium

Public relationships include large investors, academic and technology partners, media collaborations, and Olympic or TV-linked brand partnerships.

EC-002 EC-007 EC-009

Evidence gaps

No public material-partner contract schedule was reviewed.
No public source disclosed revenue contribution or dependency by partner.

Hidden risks

Partnership announcements do not disclose exclusivity, revenue contribution, data rights, or termination terms.
Research partnerships can create hidden IP, compliance, and publication-rights questions.

Follow-up questions

Provide all material partnership agreements, research MoUs, revenue contribution, and data/IP terms.

Strategic relationships and partnerships
partner or relationship	nature	public evidence	gap
Tencent and late-stage investors	Capital and strategic backing	Multiple financing rounds publicly identified Tencent and other major funds.	Investor rights, governance, and current ownership percentages are private.
Tsinghua, Peking University, Chinese Academy of Sciences, Microsoft	AI / research collaboration	TechCrunch described AI-lab collaboration with elite schools and Microsoft.	Data rights, publication rights, IP ownership, and contract terms are not public.
Winter Olympics / Strongest Brain / Oxford University Press	Brand and education partnerships	Current homepage highlights these public collaborations and endorsements.	Revenue contribution, exclusivity, and ongoing status are not public.

III.C Revenue by customer

not publicly verifiable confidence: low

No reviewed public source disclosed revenue by customer, household, cohort, channel, or product.

EC-018

Evidence gaps

No public product-level or cohort-level revenue concentration schedules were available.
No public source disclosed any customer accounting for 5% or more of revenue.

Hidden risks

Consumer education businesses can mask heavy concentration in a few products, geographies, or promotional periods.
Refund rates and deferred revenue could materially change perceived demand quality.

Follow-up questions

Provide revenue by product, geography, cohort, and channel, including refund and deferred-revenue schedules.

III.D Significant relationships severed within the last two years

not publicly verifiable confidence: low

No public schedule of severed customer, partner, or supplier relationships was reviewed.

EC-011 EC-018

Evidence gaps

No public lost-customer, lost-partner, or supplier-termination schedule was available.
No public customer-reference or churn documentation was reviewed.

Hidden risks

The 2021 regulatory reset likely forced material commercial changes even if counterparties were not publicly named.
Partner or supplier exits can be hidden inside broader product-line withdrawals.

Follow-up questions

Provide the last two years of lost customer, partner, and supplier reports with reasons and remediation.

III.E Top suppliers

partially verified confidence: medium

The public record points to app stores, academic/technology partners, and teaching-center operations, but not to supplier spend or contract terms.

EC-002 EC-007 EC-014 EC-015 EC-016 EC-018

Evidence gaps

No public top-supplier schedule, cloud commitments, content-license register, or app-store commercial agreement was reviewed.
No public purchase amounts or minimum commitments were available.

Hidden risks

A child-focused app portfolio may depend on opaque data, moderation, content, cloud, or device suppliers.
Supplier concentration is likely but not publicly quantified.

Follow-up questions

Provide supplier concentration, app-store terms, AI/cloud contracts, content licenses, and teacher or contractor dependency data.

Supplier, platform, and concentration dependency matrix
dependency	role	public evidence	risk
Apple App Store and mobile distribution	Distribution and discovery for flagship apps	Active Yuanfudao, Xiaoyuan AI, and Zebra AI listings with large review bases.	Platform-policy or ranking changes could affect discovery and trust.
Teaching centers and workforce footprint	Delivery and support infrastructure	TechCrunch cited 30,000 employees in teaching centers across China in 2020.	Large fixed-cost footprint and workforce transition risk after the crackdown.
Research and AI collaborators	AI capability and technical inputs	AI-lab collaborations and public technical achievements were highlighted by company and press sources.	Data, IP, and AI-governance obligations may be material but not public.

Chapter 04

04Competition

Yuanfudao operates in a crowded Chinese edtech market where regulation changed the basis of competition from growth-at-all-costs tutoring scale toward compliant AI-enabled learning tools and adjacent consumer products.

EC-004 EC-005 EC-007 EC-019

IV.A Competitive landscape by market segment

partially verified confidence: medium

Public sources place Yuanfudao in direct competition with Zuoyebang, VIPKid, Yiqizuoye, Byju's, and broader AI-learning tools, with regulation sharply increasing the importance of compliant product positioning.

EC-004 EC-005 EC-007 EC-019

Evidence gaps

No public win/loss analysis, net-promoter benchmarking, or verified market-share data was reviewed.
No public source benchmarked learning outcomes or teacher quality across the competitive set.

Hidden risks

Competition after 2021 may be less about pure user scale and more about regulatory classification, trust, and affordable AI features.
Competitor fundraising and product pivots can reset market structure quickly.

Follow-up questions

Provide competitor win/loss analysis, pricing comparisons, user acquisition costs, and learning-outcome benchmarks.

Competitor comparison matrix
competitor	segment	why it matters	source basis
Zuoyebang	K-12 tutoring and homework tools	Repeatedly named as a primary rival and also targeted by the same 2021 regulatory and advertising actions.	TechNode / KR Asia sector reporting.
VIPKid	Online tutoring / live classes	Public coverage places Yuanfudao in the same broader China online-learning race.	TechNode and TechCrunch competitive context.
Yiqizuoye / 17zuoye	Online education and learning tools	Competes for digital-learning distribution and parent spend.	TechNode / TechCrunch context.
Byju's	Global edtech valuation benchmark	TechCrunch framed Yuanfudao as surpassing Byju's on valuation in 2020.	TechCrunch valuation article.

Basis-of-competition scoring
axis	yuanfudao position	pressure	diligence need
Consumer scale and app reach	Strong historical user scale and current app-review volume	Low-cost alternatives and shifting demand can erode consumer advantage.	Current active users, paid conversion, cohort retention, and CAC.
AI and personalization	AI lab, public research claims, and current AI branding are strong	AI quality, child-safety, and regulation can narrow product differentiation.	Learning outcomes, model governance, and privacy controls.
Regulatory adaptability	Current product mix suggests adaptation, but evidence is incomplete	Regulation can instantly change what is sellable or scalable.	Product classification memos, entity maps, and current compliance reviews.

Competitive positioning map after the tutoring crackdown Qualitative positioning across AI breadth and regulatory exposure.

This is a positioning aid, not a market-share chart.

Chapter 05

05Marketing, Sales, and Distribution

Historical distribution relied on apps, live courses, and high-visibility partnerships, but the marketing model changed materially once regulators targeted tutoring advertising and profit-driven expansion.

EC-001 EC-003 EC-008 EC-011 EC-012 EC-014 EC-015 EC-016

V.A Strategy and implementation

partially verified confidence: medium

The public record shows an app-led, mass-market education strategy amplified by live courses, brand partnerships, and AI-based learning utilities, but 2021 regulation changed the permissible GTM playbook.

EC-001 EC-003 EC-011 EC-012 EC-014 EC-015 EC-016

Evidence gaps

No current GTM budget, CAC, payback, or pipeline data was public.
The balance between paid marketing, organic app discovery, and partnership-led distribution is not public.

Hidden risks

Regulators may continue to challenge aggressive pricing, advertising, or content classification.
Current marketing efficiency is not knowable from public app ratings alone.

Follow-up questions

Provide current GTM plan, CAC/payback, channel attribution, and all compliance controls for marketing claims and pricing.

Distribution channels and GTM motions
channel	public evidence	current question
Flagship learning apps	Yuanfudao, Xiaoyuan AI, and Zebra AI are active in the China App Store with significant ratings.	What share of gross bookings and customer acquisition now comes through apps?
Live courses / tutoring	Historical reporting said most revenue came from live courses and that Yuanfudao offered live tutoring.	What portion remains in compliant form after the crackdown?
Brand and media partnerships	Homepage highlights Olympics and TV-show partnerships.	Do these partnerships materially drive user acquisition or only brand awareness?
Non-subject and AI products	Current site centers non-subject and AI products such as Zebra AI Study and Haitun AI.	Are these products replacing restricted tutoring economics?

Public marketing-signal summary
signal	observed value	diligence caveat
Educational app filing	Yuanfudao and related apps appeared in the first filing list reported by Chinanews.	Filing improves compliance optics but does not prove sustainable economics.
Mass brand campaigns and partnerships	Homepage highlights Winter Olympics and Strongest Brain partnerships and high-attendance promotional events.	Brand reach does not disclose CAC or long-term retention.
App-store social proof	Flagship apps show hundreds of thousands to millions of ratings in the China App Store.	Ratings are dynamic and not the same as monetized active users.

Public GTM and distribution anchors Bar chart of public GTM anchors tied to apps, paid users, and partnerships.

V.B Major Customers

not publicly verifiable confidence: low

The company appears to serve mass-market student and parent cohorts rather than named enterprise accounts, but public sources do not disclose major paying cohorts or renewal behavior.

EC-018

Evidence gaps

No major-customer cohort data, pipeline, or renewal calendar was public.
No public household-level spend or premium-segment retention data was reviewed.

Hidden risks

A consumer model can still hide dependence on a small set of premium products or seasonal campaigns.
Policy shocks can change cohort behavior very quickly.

Follow-up questions

Provide cohort-level revenue, repeat purchase, retention, and renewal metrics for all flagship products.

V.C Principal avenues for generating new business

partially verified confidence: medium

Public evidence suggests new business historically came from mobile learning apps, live courses, and branded educational partnerships, with today's growth more dependent on AI and non-subject products.

EC-001 EC-014 EC-015 EC-016 EC-018

Evidence gaps

No source disclosed current new-business mix by channel, product, or geography.
No public channel-conversion or referral data was reviewed.

Hidden risks

The revenue contribution of each acquisition path is not public.
Marketing channels after the crackdown may have shifted in ways not obvious from app metadata alone.

Follow-up questions

Provide new-customer mix, conversion funnel, and channel attribution by product line.

V.D Sales force productivity model

not publicly verifiable confidence: low

No reviewed public source disclosed a sales productivity model, sales compensation, or quota structure.

EC-018

Evidence gaps

No public compensation plans, quotas, sales-cycle metrics, or hiring productivity data were available.
No public source broke down go-to-market staffing between teaching, support, and commercial roles.

Hidden risks

The business may rely more on consumer growth mechanics than on a classic B2B salesforce, but staffing and compensation data are private.
Any large teaching or support workforce can still create high fixed-cost obligations.

Follow-up questions

Provide sales/teaching organization productivity metrics, compensation plans, and role-by-role economics.

V.E Ability to implement marketing plan with current and projected budgets

not publicly verifiable confidence: low

Current and projected marketing budgets are not public, and the sector's regulatory environment makes historical growth spending a poor proxy for future budget efficiency.

EC-011 EC-018

Evidence gaps

No board-approved marketing budget or headcount plan was public.
No public source disclosed how budget allocation changed after 2021.

Hidden risks

Marketing plans may now depend more on compliance and trust than on raw paid acquisition.
Budget adequacy is impossible to judge without current product-line economics.

Follow-up questions

Provide current and projected marketing budgets, channel ROI, and compliance review workflow.

Chapter 06

06Research and Development

Yuanfudao publicly describes an AI-first R&D posture with a dedicated research institute, leadership quotes, and algorithmic learning tools, but R&D spend, model governance, and roadmap execution quality are private.

EC-002 EC-007 EC-010 EC-013

VI.A Description of R&D organization

partially verified confidence: medium

Public sources identify an AI Research Institute, academic collaborations, and named technical leaders, but the current R&D org chart and spend levels are not public.

EC-002 EC-007 EC-010 EC-012 EC-018

Evidence gaps

No public R&D org chart, budget, capitalized software policy, or model-risk management framework was available.
No public source disclosed the current staffing of the AI Research Institute or teaching-center engineering support.

Hidden risks

Research intensity can create high fixed costs and data-governance risk without making outcomes durable.
The current split between research, content, teaching, and product engineering is not public.

Follow-up questions

Provide the current R&D org chart, headcount by team, budget, model governance framework, and research-partner agreements.

Key R&D personnel and public technical leadership
name or group	role	public evidence	gap
Li Xin	Co-founder	Quoted in PeopleApp on AI and education change.	Current scope, equity, and reporting lines are not public.
Guo Changzhen	CTO	Quoted in PeopleApp on personalized teaching and AI support.	Current team size, roadmap ownership, and controls are not public.
Li Yunjin	Visual algorithm group lead	PeopleApp described him discussing a virtual teacher feature.	Launch timing and commercialization are not public.
AI Research Institute and collaborators	Research function	Company and TechCrunch highlight AI research work and collaborations.	Current staffing and IP governance remain private.

Yuanfudao R&D and AI leadership map High-level public R&D organization based on named leaders and functions.

This is a conservative public-org sketch, not a confirmed internal reporting chart.

VI.B New Product Pipeline

partially verified confidence: medium

Public sources show continued AI-led product development, including virtual-teacher ambitions and a current portfolio of AI learning apps, but roadmap timing, cost to complete, and commercialization are not public.

EC-001 EC-010 EC-012 EC-018

Evidence gaps

No public roadmap, release cadence, or post-launch KPI reporting was available.
No source disclosed R&D cost, AI infrastructure spend, or product-specific commercialization assumptions.

Hidden risks

Pipeline complexity can increase content, compliance, and model-evaluation risk.
Current hardware or device ambitions were not verified beyond broad public AI-transformation commentary.

Follow-up questions

Provide the product roadmap, major releases since 2021, AI-model vendors, and product-level commercialization targets.

Public product and research pipeline
initiative	status	evidence	risk
Virtual teacher	Publicly described as upcoming in 2019	PeopleApp said a virtual teacher was coming and would adapt learning based on student responses.	Execution, educational efficacy, and child-safety controls.
AI-enabled personalized study tools	Current and historical public focus area	Homepage plus AI claims on learning tools and research rankings.	Model governance, privacy, and measurement of actual learning outcomes.
Curriculum and service expansion	2020 funding use-of-proceeds	2020 financing coverage said new capital would fund curriculum innovation and service expansion.	Post-crackdown monetization may differ from original plans.

Chapter 07

07Management and Personnel

Public management visibility is partial: founder, co-founder, CTO, and one vice-president are identifiable, and 2020 coverage described a 30,000-employee footprint, but compensation, attrition, and current org design are private.

EC-005 EC-007 EC-010 EC-012 EC-017

VII.A Organization Chart

partially verified confidence: low

Only a high-level public organization view is available, centered on founder leadership, AI/R&D leadership, and a vice-president spokesperson role.

EC-005 EC-010 EC-012 EC-017

Evidence gaps

No public executive org chart, board roster, or legal-entity leadership map was reviewed.
No public succession or delegated-authority materials were available.

Hidden risks

The current org may have changed materially after the 2021 sector reset.
Large teaching-center and app-operations teams can create hidden coordination risk.

Follow-up questions

Provide the executive org chart, board list, legal-entity structure, and succession plans.

High-level management org chart Publicly visible management and spokesperson roles.

VII.B Historical and projected headcount by function and location

partially verified confidence: medium

The best public headcount anchor is TechCrunch's 2020 reference to 30,000 employees in teaching centers across China; current and projected headcount are not public.

EC-007 EC-010 EC-012 EC-018

Evidence gaps

No current headcount by function, location, or legal entity was public.
No public hiring plan, teaching-center count, or severance schedule was reviewed.

Hidden risks

Headcount may have fallen materially after the 2021 crackdown, making the 2020 anchor stale.
Public app activity does not reveal the current cost base or support staffing.

Follow-up questions

Provide current and projected headcount by function/location and all major workforce-change documents since 2021.

Headcount and hiring signals
signal	public value	period	risk or request
Teaching-center workforce	30,000 employees	2020	Potentially stale after the 2021 crackdown; needs current verification.
Regional office footprint	Chengdu, Xi'an, and Zhengzhou publicly named	2019	Need full location map, legal entities, and current employment counts.
Current headcount	not publicly verifiable	current	Requires current org chart, payroll, and severance schedule.

Public headcount anchors over time Bar chart of the limited public headcount anchors available.

VII.C Senior management biographies

partially verified confidence: medium

Senior-management visibility is limited but not absent: public sources identify founder/CEO Li Yong, co-founder Li Xin, CTO Guo Changzhen, and vice-president Cheng Qun.

EC-005 EC-010 EC-012

Evidence gaps

No comprehensive public executive biography set was reviewed.
Board membership, legal-entity officer status, and compensation are not public.

Hidden risks

A thin public management bench can hide key-person and succession risk.
Public biographies do not reveal tenure economics, equity, or outside obligations.

Follow-up questions

Provide full executive biographies, officer lists, board composition, and current responsibilities across entities.

Senior management public roster
name	role	public evidence	diligence gap
Li Yong	CEO	TechNode said CEO Li Yong announced the 2020 financing in an internal letter made public.	Current role, equity, and board authority are not public.
Li Xin	Co-founder	PeopleApp quoted Li Xin as a Yuanfudao co-founder discussing AI in education.	Current responsibilities and equity are not public.
Guo Changzhen	CTO	PeopleApp quoted Guo Changzhen on personalized teaching and AI.	Current reporting line, org size, and incentive package are not public.
Cheng Qun	Vice-president	China Daily quoted Cheng Qun in 2026 on AI-driven educational transformation.	Division ownership and formal seniority are not public.

VII.D Compensation arrangements

not publicly verifiable confidence: low

Executive and workforce compensation arrangements are not public.

EC-018

Evidence gaps

No employment agreements, compensation schedules, or benefit plans were public.
No public change-of-control or retention terms were reviewed.

Hidden risks

Historic scale and regulatory change can create material severance, teacher-compensation, and retention obligations.
App-store success is not a substitute for compensation sustainability.

Follow-up questions

Provide executive agreements, teacher and technical compensation plans, benefit plans, and severance obligations.

VII.E Incentive stock plans

not publicly verifiable confidence: low

No public equity incentive or option-plan detail was reviewed.

EC-018

Evidence gaps

No option plan, 409A-equivalent valuation, or employee-liquidity information was public.
No public source disclosed the current option pool or award terms.

Hidden risks

Late-stage private companies often rely heavily on opaque option refreshes and secondary-liquidity promises.
Regulatory restructuring could affect vesting and employee equity value.

Follow-up questions

Provide all equity-incentive plans, grant ledgers, valuation support, and employee liquidity programs.

VII.F Significant employee relations problems, past or present

not publicly verifiable confidence: low

No public employee-relations case schedule was reviewed.

EC-011 EC-018

Evidence gaps

No public HR investigation logs, labor claims, or employee-relations reports were reviewed.
No public compliance audits or works-council equivalents were available.

Hidden risks

Large teacher and support operations can create labor, classification, or complaint risk that is invisible in general public sources.
Downsizing risk is elevated in a sector that saw abrupt policy-driven contraction.

Follow-up questions

Provide all labor claims, employee-relations logs, and compliance reviews since 2020.

VII.G Personnel Turnover

inconclusive confidence: low

Sector-wide downsizing was publicly reported in 2021, but Yuanfudao-specific turnover data and retention program details were not public.

EC-012 EC-018

Evidence gaps

No public two-year attrition data or retention-program detail was available.
No public source quantified post-crackdown restructuring for Yuanfudao specifically.

Hidden risks

Turnover may have been substantial after 2021, but the public record is not quantitative.
Retention of top AI and technical talent may be difficult under regulatory and valuation uncertainty.

Follow-up questions

Provide attrition by function/location, retention plans, and all major restructuring or severance programs since 2021.

Departures, turnover, and personnel evidence gaps
topic	public signal	status	diligence need
Sector downsizing	KR Asia reported Yuanfudao and peers were downsizing amid regulatory uncertainty.	inconclusive	Quantify role-by-role layoffs, severance, and retained talent by function.
Yuanfudao-specific attrition	No public attrition percentages or regretted-loss metrics reviewed.	not_publicly_verifiable	Provide two-year attrition by function/location and talent-retention plans.
Employee incentive retention	No public option-refresh or liquidity data reviewed.	not_publicly_verifiable	Provide stock-plan, refresh, and retention-grant data.

Chapter 08

08Legal and Related Matters

The strongest public legal signals concern education regulation, app filing, and advertising-compliance issues; litigation, IP ownership, contracts, and insurance remain largely private.

EC-008 EC-009 EC-011 EC-012 EC-018

VIII.A Pending lawsuits against the Company

not publicly verifiable confidence: low

No pending lawsuit against Yuanfudao was verified in the reviewed accessible public sources.

EC-018

Evidence gaps

No docket-grade litigation search across all relevant entities and jurisdictions was completed.
No counsel letter, reserve schedule, or threatened-claims log was available.

Hidden risks

A lack of accessible public lawsuit results is not equivalent to a counsel-grade negative search.
Child-data, advertising, teacher, and refund issues can create claims that do not surface prominently in English-language sources.

Follow-up questions

Provide all pending, threatened, and settled claims with counsel assessment and reserves.

Pending lawsuits against Yuanfudao
matter type	public status	status	request
Pending lawsuits against Yuanfudao or principal entities	No verified pending civil case was identified in the reviewed accessible public sources.	not_publicly_verifiable	Counsel letter, full docket search, threatened-claims log, and reserve schedule.
Administrative matters	Public sources instead highlight regulatory actions such as app filing and false-advertising penalties.	partially_verified	Distinguish regulatory actions from private litigation in counsel materials.

VIII.B Pending lawsuits initiated by Company

not publicly verifiable confidence: low

No public schedule of company-initiated lawsuits or enforcement actions was reviewed.

EC-018

Evidence gaps

No company-initiated litigation schedule or demand-letter register was public.
No public source summarized historical IP enforcement by the company.

Hidden risks

App, content, and AI businesses often bring IP or unfair-competition disputes that are not visible from high-level public review.
Any historical enforcement around content, exams, or trademarks may sit in local registries not searched here.

Follow-up questions

Provide all company-initiated litigation, demand letters, and IP-enforcement matters.

Pending lawsuits initiated by Yuanfudao
matter type	public status	status	request
Company-initiated litigation or IP enforcement	No public schedule was verified in reviewed accessible sources.	not_publicly_verifiable	Provide all demand letters, infringement actions, and company-initiated cases.
Brand or content enforcement history	Material brand and content rights appear important, but enforcement history was not public.	not_publicly_verifiable	Provide trademark, content, and anti-piracy enforcement history.

VIII.C Environmental and employee safety issues and liabilities

not publicly verifiable confidence: low

As a software and digital-content business, Yuanfudao has limited obvious environmental exposure, but office, teaching-center, and workforce liabilities are not public.

EC-001 EC-012 EC-018

Evidence gaps

No public environment, safety, office, or workplace-incident records were reviewed.
No public insurance or lease schedules were available.

Hidden risks

Teaching centers and offices can still create workplace safety, lease, and labor-compliance exposures.
Any device-related or hardware pilot efforts would add operational risk not visible in the reviewed sources.

Follow-up questions

Provide office and teaching-center footprint data, safety reports, and labor-compliance reviews.

VIII.D Material patents, copyrights, licenses, and trademarks

partially verified confidence: medium

The public record supports a meaningful portfolio of app brands, AI-research outputs, and software assets, but registry-level IP schedules and license chains are not public.

EC-002 EC-007 EC-010 EC-018

Evidence gaps

No comprehensive patent, copyright, trademark, or open-source register was publicly available in reviewed sources.
No public source disclosed full license terms for content, research outputs, or AI components.

Hidden risks

AI partnerships and educational content can create complicated rights, data-license, and model-governance questions.
Brand portfolios and content rights are material for a consumer learning-app company but are not fully visible from public sources.

Follow-up questions

Provide all trademark, copyright, patent, content-license, open-source, and research-collaboration schedules.

Material IP, brands, and license dependencies
asset or dependency	public evidence	status	diligence note
Yuanfudao, Xiaoyuan, and Zebra app / brand family	Current homepage and app-store metadata support a meaningful branded app portfolio.	verified	Trademark ownership and current filing status still require registry review.
AI Research Institute outputs and public technical achievements	Homepage and press coverage cite MS MARCO / Stanford results and open-sourced technologies.	verified	Open-source terms, data rights, and model governance remain private.
Research / education collaboration rights	TechCrunch referenced collaboration with Tsinghua, Peking University, Chinese Academy of Sciences, and Microsoft.	partially_verified	Need agreement terms, IP ownership, publication rights, and data-sharing rules.

VIII.E Insurance coverage and material exposures

not publicly verifiable confidence: low

Insurance coverage, limits, and exclusions were not public.

EC-018

Evidence gaps

No policy schedule, claims history, or broker summary was public.
No public source described cyber, D&O, E&O, media, or employment-practices coverage.

Hidden risks

The company's exposure profile likely spans advertising, privacy, child safety, educational outcomes, and labor matters.
The adequacy of insurance against AI or content-related claims cannot be inferred from public sources.

Follow-up questions

Provide current insurance schedules, claims history, broker letters, and all relevant exclusions.

VIII.F Material contracts

not publicly verifiable confidence: low

Public sources imply important contracts with investors, app stores, research partners, media partners, and educational-content stakeholders, but none were publicly available in reviewable form.

EC-001 EC-012 EC-017 EC-018

Evidence gaps

No current material-contract schedule or major agreement text was public.
No public source described app-store, partner, or content-license commercial terms.

Hidden risks

Post-2021 compliance may depend on contract clauses around curriculum classification, platform distribution, and content ownership.
AI and research partnerships could contain important IP and data-use obligations.

Follow-up questions

Provide all material contracts, including app-store, research, content, media, cloud, and financing agreements.

VIII.G Regulatory agency problems

partially verified confidence: high

The public record clearly shows meaningful regulatory exposure: app filing obligations, the 2021 Double Reduction policy, and false-advertising / deceptive-pricing penalties, while broader agency correspondence remains private.

EC-009 EC-011 EC-012 EC-018

Evidence gaps

No current regulator correspondence, remediation tracker, or compliance audit file was public.
No public source quantified the full operational or financial impact of enforcement actions on Yuanfudao.

Hidden risks

Regulatory interpretation risk remains high because product categories, pricing, and tutoring boundaries can shift.
The absence of a public schedule of current agency correspondence should not be treated as evidence of clean regulatory status.

Follow-up questions

Provide all regulatory correspondence, penalties, remediation plans, product classification memos, and compliance committee materials.

Regulatory and agency-action summary
event	date or period	public signal	diligence implication
Education app filing list	2019-12	Chinanews reported Yuanfudao and related apps were among the first 152 education apps to pass filing review.	Shows baseline compliance visibility but not broader regulatory cleanliness.
Double Reduction policy	2021-07	The State Council / CPC policy stopped new for-profit K-12 approvals and barred public-market and foreign-capital involvement in subject tutoring.	Transforms entity design, revenue model, and growth assumptions.
False-advertising / deceptive-pricing penalties	2021	KR Asia said Yuanfudao and Zuoyebang were each fined RMB 2.5M.	Indicates compliance and marketing-control risk.
Current AI-led transformation messaging	2026	China Daily quoted Cheng Qun on AI enabling personalized high-quality education at scale.	Suggests the company is still active, but also reframes the business around AI and policy-compliant positioning.

Legal and regulatory timeline Timeline of the strongest public compliance and regulatory events.

Yuanfudao risk heatmap Risk heatmap across the main diligence priorities.

Evidence

Evidence claims
ID	Claim	Status	Sources
EC-001	The current Yuanfudao homepage describes a live portfolio of AI and digital-learning products including Yuanfudao courses, Zebra AI Study, Xiaoyuan AI, Yuan Programming, and Haitun AI, and frames the company as using technology to deliver high-quality personalized education.	verified high	SRC-001
EC-002	Yuanfudao publicly claims significant AI capability, including top rankings on MS MARCO and Stanford Q&A benchmarks and an AI Research Institute that has open-sourced more than five internally developed technologies.	verified medium	SRC-001
EC-003	36Kr reported that Yuanfudao completed a Tencent-led US$300M round in December 2018 at a valuation above US$3B and disclosed 160M registered users, more than 1M paid users, RMB 1.5B recognized revenue, 70%-80% renewal, and roughly 42% marketing conversion.	verified medium	SRC-002
EC-004	TechCrunch reported that Yuanfudao raised US$300M in 2018 at a valuation above US$3B, served more than 200M users, derived most revenue from live courses, and operated a suite spanning live courses, a problem database, and homework-help tools plus an AI research institute.	verified high	SRC-003
EC-005	TechNode reported that Yuanfudao raised a US$1B Series G in April 2020 at a US$7.8B valuation, had more than 1M long-term users who paid full price, Zebra AI had 500,000 students, the firm had more than 400M users, and CEO Li Yong announced the financing.	verified high	SRC-004
EC-006	TechNode reported in October 2020 that Yuanfudao completed a US$1.2B G2, bringing total Series G financing to US$2.2B and the valuation to US$15.5B, while Yuanfudao and Zebra AI had 3.7M combined users.	verified high	SRC-005
EC-007	TechCrunch reported in October 2020 that Yuanfudao raised US$2.2B, reached a US$15.5B valuation, had doubled total users to 400M, employed 30,000 people in teaching centers across China, and collaborated with Tsinghua University, Peking University, the Chinese Academy of Sciences, and Microsoft through an AI research institute.	verified high	SRC-006
EC-008	A TechNode summary citing LatePost reported a US$1B G2 led by DST Global and a US$15.5B valuation in October 2020.	partially verified medium	SRC-007
EC-009	Chinanews reported in December 2019 that Yuanfudao and five affiliated apps passed the first batch of education mobile-app filing review tied to the Ministry of Education rules.	verified medium	SRC-008
EC-010	PeopleApp quoted Yuanfudao co-founder Li Xin, CTO Guo Changzhen, and visual-algorithm lead Li Yunjin on AI-enabled grading, a virtual teacher, an AI Research Institute, and offices in Chengdu, Xi'an, and Zhengzhou.	verified medium	SRC-009
EC-011	The official Double Reduction policy stopped new approvals for K-12 subject-tutoring institutions, required existing institutions to register as non-profits, and barred public-market and foreign-capital involvement in subject tutoring.	verified high	SRC-010
EC-012	KR Asia wrote that all Chinese edtech companies had to operate on a not-for-profit basis, that Yuanfudao and Zuoyebang had been fined RMB 2.5M for false advertising and deceptive pricing, and that firms including Yuanfudao halted IPO plans and downsized amid regulatory uncertainty.	partially verified medium	SRC-011
EC-013	China Daily quoted Yuanfudao Group vice-president Cheng Qun in March 2026 saying AI is enabling high-quality, large-scale personalized education and that individualized teaching is becoming reality.	verified medium	SRC-012
EC-014	Apple App Store metadata showed the Yuanfudao app carried a 4.55091 rating with 113,550 ratings under seller Beijing Yuanfudao Online Education Organisation.	verified medium	SRC-013
EC-015	Apple App Store metadata showed Xiaoyuan AI carried a 4.70633 rating with 3,140,579 ratings under seller Beijing Yuanli Science and Technology.	verified medium	SRC-014
EC-016	Apple App Store metadata showed Zebra AI Study carried a 4.66608 rating with 344,123 ratings under seller Beijing Yuanli Science and Technology.	verified medium	SRC-015
EC-017	The combination of an active current website, live current apps, and a named 2026 executive quote supports Yuanfudao as an active private operator, but definitive no-IPO / no-acquisition / no-shutdown confirmation requires corporate records.	partially verified medium	SRC-001 SRC-012 SRC-013 SRC-014 SRC-015
EC-018	Several diligence-critical items were not publicly verifiable from the reviewed source set, including audited financials, current cap table, debt, customer concentration, current headcount, litigation schedule, insurance, and material contracts.	not publicly verifiable high	SRC-001 SRC-010 SRC-011
EC-019	Independent 2018-2020 reporting placed Yuanfudao in competitive context with Zuoyebang, VIPKid, Yiqizuoye / 17zuoye, and Byju's.	verified medium	SRC-003 SRC-004 SRC-006

Sources
ID	Publisher	Title	Accessed
SRC-001	Yuanfudao	Yuanfudao homepage	2026-06-10
SRC-002	36Kr	36Kr: Yuanfudao completes a new US$300M round led by Tencent at a valuation above US$3B	2026-06-10
SRC-003	TechCrunch	TechCrunch: Tencent-backed homework app jumps to $3B valuation after raising $300M	2026-06-10
SRC-004	TechNode	TechNode: Yuanfudao is now one of China's most valuable ed-tech startups	2026-06-10
SRC-005	TechNode	TechNode: Online tutor Yuanfudao lands $1.2 billion on doubled valuation	2026-06-10
SRC-006	TechCrunch	TechCrunch: Chinese live tutoring app Yuanfudao is now worth $15.5 billion	2026-06-10
SRC-007	TechNode	TechNode Retailheads: ed tech firms nab hefty funding	2026-06-10
SRC-008	Chinanews	Chinanews: Ministry of Education publishes first education-app filing list	2026-06-10
SRC-009	People's Daily app	PeopleApp: Teaching and learning become smarter	2026-06-10
SRC-010	The State Council of the People's Republic of China	State Council / CPC General Office: Double Reduction policy	2026-06-10
SRC-011	KR Asia	KR Asia: China mandates all edtech companies to operate on not-for-profit basis	2026-06-10
SRC-012	China Daily	China Daily: Business leaders call for embrace of smart tech	2026-06-10
SRC-013	Apple App Store	Apple App Store / iTunes Search API results for Yuanfudao app (China storefront)	2026-06-10
SRC-014	Apple App Store	Apple App Store / iTunes Search API results for Xiaoyuan AI app (China storefront)	2026-06-10
SRC-015	Apple App Store	Apple App Store / iTunes Search API results for Zebra AI Study app (China storefront)	2026-06-10

Disclaimer

This report is a public-evidence diligence snapshot, not investment advice. Important financial, legal, technical, and contractual facts remain non-public and should be verified directly with management and primary documents before any investment decision.