evidoria

← Back to browse

Good practice Imported

Novissi: AI-Targeted Digital Cash Transfers and Togo's ASTRE Social Registry

Togo · Lomé · See the Togo profile · See the Lomé profile

Evidence: Quasi-experimental Top 1% 100/100 · Ask Evidence Copilot about this practice

Togo built a fully digital, mobile-money cash-transfer system using satellite imagery and phone data to target its poorest citizens during COVID-19, then turned it into a permanent, independently evaluated social-protection platform now scaling to 1.24 million people.

920,000+ people
People reached by Novissi Phase 1 cash transfers (2020–2022)
34 USD million
Total disbursed in Novissi Phase 1 (2020–2022)
154,238 people
Additional beneficiaries reached via AI/satellite targeting (Dec 2020–Apr 2021)
4–21 %
Reduction in exclusion errors vs. geographic targeting
42 %
Improvement in targeting precision
9–35 %
Higher exclusion errors vs. a full social registry
700,000 households
Households targeted under ASTRE's new cash-transfer round (from Dec 2025)
1.24 million people
People targeted for ASTRE by 2029 (by 2029)
Novissi: AI-Targeted Digital Cash Transfers and Togo's ASTRE Social Registry

Details

Maturity
Scaling
Promoter
Togolese Ministry of Digital Economy and Digital Transformation (with UC Berkeley CEGA and J-PAL)
Period
2020–2026 (ongoing)
Keywords
social protection, cash transfers, artificial intelligence, mobile money, financial inclusion

Context

Togo launched Novissi in April 2020 as one of the world's first fully digital, mobile-money-based social protection programmes, created in response to COVID-19 income shocks among informal workers with no formal safety net. The programme used self-registration by mobile phone and mobile-money delivery to reach over 920,000 people between 2020 and 2022.

Objectives

From September 2020, the Togolese government worked with researchers from UC Berkeley's CEGA and J-PAL to extend the programme into rural areas that lacked prior survey data, aiming to identify and reach the poorest households more precisely than simple geographic targeting could.

Activities

Machine-learning models trained on satellite imagery (roof material, plot size) and mobile-phone metadata (call patterns, mobile-money balances) were used to rank poverty across Togo's 100 poorest cantons, reaching 154,238 additional beneficiaries between December 2020 and April 2021 with $20 a month for three months. Building on this experience, Togo's government subsequently launched ASTRE, a national social-protection programme backed by the World Bank that combines a biometric ID system, an AI-updated dynamic social registry and an interoperable digital payment platform.

Results

A peer-reviewed evaluation (Aiken, Bellue, Blumenstock, Karlan & Udry, published in the Journal of Development Economics and covered by Nature in 2022) found that phone-based targeting cut exclusion errors by 4-21% compared with simple geographic targeting and improved targeting precision by 42%. The same study was explicit about the approach's limits: compared with a full social registry, which Togo did not have, the ML method still produced 9-35% higher exclusion errors, and it could not reach anyone without a mobile phone. A new cash-transfer round under ASTRE launched in December 2025, delivering CFAF 25,000 to a first wave within a programme targeting 700,000 households, with plans to reach 1.24 million people by 2029.

Conclusions

World Bank monitoring in 2025 noted a slower-than-planned rollout of ASTRE's Programme Coordination Unit and Independent Verification Agent, though a new government appointed in October 2025 was expected to accelerate delivery — illustrating that institutionalising an emergency digital-cash innovation into permanent infrastructure remains a work in progress.

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

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