evidoria

← Back to browse

Good practice

Vietnam General Department of Taxation — Big Data and AI for e-commerce taxpayer identification and compliance

Vietnam · Hanoi · See the Vietnam profile

Vietnam's GDT uses Big Data and ML to identify e-commerce tax evaders: Hanoi's AI matched national registries to uncover 627,000 businesses and collect VND38.8 trillion (~$1.52 B) in taxes — a 2.8x rise. Nationally, 20M invoice lines from 5 platforms are reviewed for anomalies.

627,000 businesses
E-commerce businesses newly identified (Hanoi)
38.8 trillion VND (~1.52 billion USD) tax revenue
E-commerce tax revenue collected (Hanoi)
2.8 x (fold increase, confounded by concurrent regulatory change)
Year-on-year increase in e-commerce tax revenue (Hanoi)
20 million lines
Invoice data lines reviewed nationally
2.7 million
Vendor stalls identified nationally
30,000 queries
Taxpayer chatbot queries processed (from Nov 2024)

Details

Maturity
Scaling
Promoter
General Department of Taxation (GDT), Ministry of Finance, Vietnam
Period
2023-present
Keywords
tax compliance, e-commerce, big data analytics, machine learning, invoice fraud detection

Context

Vietnam's General Department of Taxation, under the Ministry of Finance, runs a national programme applying Big Data analytics and machine learning across core tax-administration functions — automated refund classification, invoice-anomaly detection, risk alerting, e-invoice oversight and VAT reconciliation. The Department of Information Technology oversees technical delivery, with the Hanoi and Ho Chi Minh City tax departments leading localised roll-outs.

Objectives

The programme aims to bring previously untaxed e-commerce sellers into the formal tax net and to detect anomalies across a rapidly growing volume of electronic invoices, alongside piloting an AI virtual assistant to help taxpayers navigate tax rules.

Activities

The Hanoi Tax Department synchronised tax records with national population and business-registration databases using AI to cross-match e-commerce sellers against formal registration status. At national level, GDT reviewed data submitted by five major e-commerce platforms to flag unregistered vendors and anomalous transactions. In November 2024, Hanoi and GDT jointly launched an AI-powered Virtual Assistant for Taxpayers, developed over eight months with the Hanoi People's Committee, and piloted an extension to debt-management support for tax officials in January 2025.

Results

The Hanoi AI data-matching effort identified 627,000 previously unregistered e-commerce businesses, associated with VND38.8 trillion (approximately USD1.52 billion) in tax revenue from the e-commerce sector — a 2.8-fold increase on the prior year. Nationally, GDT reviewed over 20 million data lines from five e-commerce platforms, identifying 2.7 million vendor stalls with aggregate reported revenues of VND654 trillion (approximately USD24.9 billion). The taxpayer chatbot integrated more than 100 laws and had processed 30,000 queries from 4,500 users, generating over 15,000 bilingual responses.

Conclusions

GDT's own account attributes strong gains to the AI/big-data deployment, but the increase coincided with new e-commerce tax regulation (Decree 91/2022), so the 2.8x revenue rise cannot be attributed to the technology alone. The underlying ML algorithms are not publicly documented and no independent algorithmic audit has been published, so figures should be read as the agency's own reporting rather than externally verified outcomes.

Implementation
Indicative cost
Medium (€50k–€500k) — No public budget figures are disclosed; cost estimated as medium based on the scope of a national multi-function Big Data/ML deployment across a large tax administration plus an eight-month chatbot build.
Time to results
Medium (1–3 years) — Programme integrates multiple ML use cases already in operation; the taxpayer chatbot specifically took about eight months to develop (launched November 2024), with a further pilot expansion in January 2025.
Staffing & skills
Department of Information Technology (GDT), led by Director Pham Quang Toan, oversees technical implementation, Hanoi and Ho Chi Minh City tax departments lead localised deployment, Hanoi People's Committee co-developed the AI taxpayer virtual assistant over eight months

Conditions for success

  • Cross-matching AI outputs against national population and business-registration databases rather than relying on tax records alone
  • Concurrent regulatory change (Decree 91/2022) reinforcing the technology's enforcement effect
  • Reviewing platform-submitted transaction data directly from major e-commerce platforms

Common failure modes

  • Sole attribution of the 2.8x revenue increase to AI is not possible because it coincides with new e-commerce tax regulation
  • ML algorithms are not publicly documented and no independent algorithmic audit has been published

Where it fits

Governance type
national government agency
Scale
national, with city-level pilots (Hanoi, Ho Chi Minh City)
Income level
lower-middle-income

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful