evidoria

← Back to browse

Good practice Imported

ERA, RESNET & Smart Scanning — Madagascar Customs' AI-Powered Risk Analysis and Cargo Screening

Madagascar · Antananarivo · See the Madagascar profile · See the Antananarivo profile

Evidence: Descriptive / self-reported Top 75% 47/100 · Ask Evidence Copilot about this practice

Madagascar's customs authority, an IMF pilot country for AI in customs, credits its AI risk-scoring tool ERA with a 68% year-on-year revenue jump in January 2025 — a real, officially cited figure, but a single self-reported month with no independent audit or cost data published.

68 %
Year-on-year customs revenue increase attributed to ERA risk-scoring (self-reported, single month, unaudited) (January 2025 vs. January 2024)
ERA, RESNET & Smart Scanning — Madagascar Customs' AI-Powered Risk Analysis and Cargo Screening

Details

Maturity
Scaling
Promoter
Direction Générale des Douanes de Madagascar (DGD), with IMF technical assistance
Period
AI tools in active use by January 2025; IMF technical-assistance mission April 2025; regional scale-up planned through 2029
Keywords
customs, dynamic risk analysis, computer vision, revenue collection, e-government

Context

Madagascar's Direction Générale des Douanes (DGD) operates three AI systems: ERA (Enhanced Risk Assessment) for dynamic risk-scoring of shipments, RESNET for automated X-ray and scanner image analysis, and Smart Scanning for container/cargo screening. Officials describe Madagascar as 'the first public institution to adopt artificial intelligence' in the country's public sector. The IMF designated Madagascar its African pilot country for AI-in-customs integration, and IMF specialists Victor Budeau and François Chastel launched an on-site technical-assistance mission on 24 April 2025.

Results

Director General Lainkana Zafivanona Ernest publicly reported a 68% year-on-year customs revenue increase (January 2025 vs. January 2024) at a technology conference in Antananarivo on 6 March 2025, attended by 2,500+ students and academics from Durham University, LSE, and Universitas Airlangga. This figure is a genuine, officially disclosed result but derives from a single self-reported month-on-month comparison, without independent audit, cost data, fraud-detection rates, or false-positive analysis.

Conclusions

DGD aims to extend the tools to additional control sectors and share its experience regionally by 2029. No published appeals process, independent oversight body, or data-protection framework governing the risk-scoring model currently exists.

Implementation
Indicative cost
Medium (€50k–€500k) — No cost data published for ERA, RESNET, or Smart Scanning; IMF technical assistance mission (April 2025) supported implementation.
Time to results
Medium (1–3 years) — Tools active by January 2025; IMF technical-assistance mission April 2025; DGD plans extension to additional control sectors and regional knowledge-sharing through 2029.
Staffing & skills
Direction Générale des Douanes de Madagascar (DGD), IMF technical assistance specialists Victor Budeau and François Chastel

Conditions for success

  • IMF pilot-country designation providing technical assistance and international visibility
  • Leadership sponsorship from the DGD Director General
  • Live deployment integrated into nationwide customs clearance operations rather than laboratory trials

Common failure modes

  • The 68% revenue-increase figure is self-reported, covers a single month, and lacks independent audit, cost data, fraud-detection rates, or false-positive analysis
  • No published appeals process, independent oversight body, or data-protection framework governing the risk-scoring model

Where it fits

Governance type
national government
Scale
national customs authority
Income level
low income (Madagascar)

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

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

Attachments

Similar practices you may find useful