Proof of Good | SDG 1: No Poverty

Leveraging Mobile and Satellite Data (Case Study: Togo)

In the quest to achieve the United Nations Sustainable Development Goals (SDGs), innovative applications of data science are becoming pivotal in addressing entrenched socio-economic challenges. One particularly illustrative example is Togo’s use of mobile phone and satellite data to effectively target emergency cash transfers during the COVID-19 pandemic, exemplifying SDG 1: No Poverty. This initiative demonstrates how advanced data analytics, machine learning, and cross-sector collaboration can translate raw data into meaningful social impact, rapidly identifying vulnerable populations in need.

Context and Challenge

In April 2020, amid the global COVID-19 pandemic, the Togolese government launched the Novissi program to provide emergency financial support to residents in the Greater Lomé region. The program aimed to mitigate the economic shock of lockdowns for individuals unable to work due to public health restrictions. Monthly cash transfers ranged from $10 to $20, targeted to residents who were registered to vote in specific regions and engaged in particular occupations.

However, Togo faced a critical challenge: the lack of recent, reliable poverty data. Traditional household surveys were outdated, costly, and logistically impractical due to social distancing measures. The government needed a solution that could accurately identify the poorest residents quickly and at scale, ensuring that aid reached those most in need without delays.

Data-Driven Solution

To address this gap, a team of researchers led by UC Berkeley, in collaboration with the Togolese government, applied an innovative combination of satellite imagery and mobile phone data, analyzed using machine learning algorithms. Key contributors included UC Berkeley researchers, who designed and implemented the data models, the Government of Togo who oversaw program implementation and coordinated the distribution of funds via the Novissi mobile platform, Tech entities such as GiveDirectly, who facilitated cash transfers and provided operational expertise in digital payment systems, and the World Bank who provided financial support to conduct a large phone survey.


Source: UC Berkeley, Vice Chancellor for Research Office

 The methodology unfolded in two complementary stages:

    1. Identifying Vulnerable Regions via Satellite Data
      Satellite imagery was analyzed to determine the 100 poorest “cantons,” each comprising approximately three to four villages. Machine learning models extracted variables correlated with poverty, including roof material density and forest coverage. This approach allowed researchers to pinpoint areas with the highest concentration of impoverished residents, even in the absence of up-to-date census data.
    2. Targeting Individuals via Mobile Phone Data
      Within these selected cantons, mobile phone metadata was used to identify the poorest individuals. Machine learning algorithms assessed behavioral indicators such as the volume of international calls, frequency of mobile internet usage, and other usage patterns. These signals, when combined, provided a predictive model of poverty status, allowing for remote, rapid targeting of financial assistance.
Source: Nature

Implementation and Results

The combined satellite-phone approach enabled the government to distribute aid with unprecedented speed and precision. Approximately 140,000 residents received emergency cash transfers as a direct result of this targeting methodology. Compared to conventional methods, such as social registries, which were either unavailable or incomplete, the phone-based approach minimized misclassification, reducing both exclusion of eligible recipients and inclusion of ineligible individuals.

Critically, while traditional social registries might be more accurate, they were not feasible under the constraints imposed by the pandemic. The satellite and phone data approach demonstrated a unique advantage: it could be deployed entirely remotely within just a few weeks, addressing both the urgency and scale required during a crisis.

Outcomes and Impact

    • Scale of Aid: 140,000 individuals reached, translating to substantial immediate economic relief.
    • Efficiency: Aid delivery was faster and more precise compared to traditional methods, circumventing delays associated with field surveys and paper-based registries.
    • Innovation in Poverty Targeting: The initiative showcased the potential of integrating multiple data streams (satellite imagery and mobile metadata) into machine learning frameworks for social good.
    • Policy Implications: The success of the Togo case provides a replicable model for other countries facing crises without up-to-date poverty data, whether during pandemics, natural disasters, or conflicts.

Conclusion

The Togo Novissi program exemplifies how modern data science can directly support SDG 1 by effectively identifying and assisting those living in poverty. By leveraging satellite imagery and mobile phone data through sophisticated machine learning algorithms, the program reached 140,000 of the country’s most vulnerable residents with timely financial support. This initiative demonstrates that when governments, researchers, and NGOs collaborate strategically, data can become a powerful instrument for social good, enabling rapid, large-scale interventions that traditional methods alone cannot achieve. As countries worldwide seek innovative approaches to meet the SDGs, the Togo case provides both a blueprint and inspiration for using data to address poverty with precision, speed, and ethical rigor.

Leveraging Mobile and Satellite Data (Case Study: Togo)

In the quest to achieve the United Nations Sustainable Development Goals (SDGs), innovative applications of data science are becoming pivotal in addressing entrenched socio-economic challenges. One particularly illustrative example is Togo’s use of mobile phone and satellite data to effectively target emergency cash transfers during the COVID-19 pandemic, exemplifying SDG 1: No Poverty. This initiative demonstrates how advanced data analytics, machine learning, and cross-sector collaboration can translate raw data into meaningful social impact, rapidly identifying vulnerable populations in need.

Context and Challenge

In April 2020, amid the global COVID-19 pandemic, the Togolese government launched the Novissi program to provide emergency financial support to residents in the Greater Lomé region. The program aimed to mitigate the economic shock of lockdowns for individuals unable to work due to public health restrictions. Monthly cash transfers ranged from $10 to $20, targeted to residents who were registered to vote in specific regions and engaged in particular occupations.

However, Togo faced a critical challenge: the lack of recent, reliable poverty data. Traditional household surveys were outdated, costly, and logistically impractical due to social distancing measures. The government needed a solution that could accurately identify the poorest residents quickly and at scale, ensuring that aid reached those most in need without delays.

Source: UC Berkeley, Vice Chancellor for Research Office

Data-Driven Solution

To address this gap, a team of researchers led by UC Berkeley, in collaboration with the Togolese government, applied an innovative combination of satellite imagery and mobile phone data, analyzed using machine learning algorithms. Key contributors included UC Berkeley researchers, who designed and implemented the data models, the Government of Togo who oversaw program implementation and coordinated the distribution of funds via the Novissi mobile platform, Tech entities such as GiveDirectly, who facilitated cash transfers and provided operational expertise in digital payment systems, and the World Bank who provided financial support to conduct a large phone survey.

The methodology unfolded in two complementary stages:

    1. Identifying Vulnerable Regions via Satellite Data
      Satellite imagery was analyzed to determine the 100 poorest “cantons,” each comprising approximately three to four villages. Machine learning models extracted variables correlated with poverty, including roof material density and forest coverage. This approach allowed researchers to pinpoint areas with the highest concentration of impoverished residents, even in the absence of up-to-date census data.
    2. Targeting Individuals via Mobile Phone Data
      Within these selected cantons, mobile phone metadata was used to identify the poorest individuals. Machine learning algorithms assessed behavioral indicators such as the volume of international calls, frequency of mobile internet usage, and other usage patterns. These signals, when combined, provided a predictive model of poverty status, allowing for remote, rapid targeting of financial assistance.


Source: Nature

Implementation and Results

The combined satellite-phone approach enabled the government to distribute aid with unprecedented speed and precision. Approximately 140,000 residents received emergency cash transfers as a direct result of this targeting methodology. Compared to conventional methods, such as social registries, which were either unavailable or incomplete, the phone-based approach minimized misclassification, reducing both exclusion of eligible recipients and inclusion of ineligible individuals.

Critically, while traditional social registries might be more accurate, they were not feasible under the constraints imposed by the pandemic. The satellite and phone data approach demonstrated a unique advantage: it could be deployed entirely remotely within just a few weeks, addressing both the urgency and scale required during a crisis.

Outcomes and Impact

  • Scale of Aid: 140,000 individuals reached, translating to substantial immediate economic relief.
  • Efficiency: Aid delivery was faster and more precise compared to traditional methods, circumventing delays associated with field surveys and paper-based registries.
  • Innovation in Poverty Targeting: The initiative showcased the potential of integrating multiple data streams (satellite imagery and mobile metadata) into machine learning frameworks for social good.
  • Policy Implications: The success of the Togo case provides a replicable model for other countries facing crises without up-to-date poverty data, whether during pandemics, natural disasters, or conflicts.

Conclusion

The Togo Novissi program exemplifies how modern data science can directly support SDG 1 by effectively identifying and assisting those living in poverty. By leveraging satellite imagery and mobile phone data through sophisticated machine learning algorithms, the program reached 140,000 of the country’s most vulnerable residents with timely financial support. This initiative demonstrates that when governments, researchers, and NGOs collaborate strategically, data can become a powerful instrument for social good, enabling rapid, large-scale interventions that traditional methods alone cannot achieve. As countries worldwide seek innovative approaches to meet the SDGs, the Togo case provides both a blueprint and inspiration for using data to address poverty with precision, speed, and ethical rigor.

Leveraging Mobile and Satellite Data (Case Study: Togo)

In the quest to achieve the United Nations Sustainable Development Goals (SDGs), innovative applications of data science are becoming pivotal in addressing entrenched socio-economic challenges. One particularly illustrative example is Togo’s use of mobile phone and satellite data to effectively target emergency cash transfers during the COVID-19 pandemic, exemplifying SDG 1: No Poverty. This initiative demonstrates how advanced data analytics, machine learning, and cross-sector collaboration can translate raw data into meaningful social impact, rapidly identifying vulnerable populations in need.

Context and Challenge

In April 2020, amid the global COVID-19 pandemic, the Togolese government launched the Novissi program to provide emergency financial support to residents in the Greater Lomé region. The program aimed to mitigate the economic shock of lockdowns for individuals unable to work due to public health restrictions. Monthly cash transfers ranged from $10 to $20, targeted to residents who were registered to vote in specific regions and engaged in particular occupations.

However, Togo faced a critical challenge: the lack of recent, reliable poverty data. Traditional household surveys were outdated, costly, and logistically impractical due to social distancing measures. The government needed a solution that could accurately identify the poorest residents quickly and at scale, ensuring that aid reached those most in need without delays.

Data-Driven Solution

To address this gap, a team of researchers led by UC Berkeley, in collaboration with the Togolese government, applied an innovative combination of satellite imagery and mobile phone data, analyzed using machine learning algorithms. Key contributors included UC Berkeley researchers, who designed and implemented the data models, the Government of Togo who oversaw program implementation and coordinated the distribution of funds via the Novissi mobile platform, Tech entities such as GiveDirectly, who facilitated cash transfers and provided operational expertise in digital payment systems, and the World Bank who provided financial support to conduct a large phone survey.


Source: UC Berkeley, Vice Chancellor for Research Office

 The methodology unfolded in two complementary stages:

    1. Identifying Vulnerable Regions via Satellite Data
      Satellite imagery was analyzed to determine the 100 poorest “cantons,” each comprising approximately three to four villages. Machine learning models extracted variables correlated with poverty, including roof material density and forest coverage. This approach allowed researchers to pinpoint areas with the highest concentration of impoverished residents, even in the absence of up-to-date census data.
    2. Targeting Individuals via Mobile Phone Data
      Within these selected cantons, mobile phone metadata was used to identify the poorest individuals. Machine learning algorithms assessed behavioral indicators such as the volume of international calls, frequency of mobile internet usage, and other usage patterns. These signals, when combined, provided a predictive model of poverty status, allowing for remote, rapid targeting of financial assistance.
Source: Nature

Implementation and Results

The combined satellite-phone approach enabled the government to distribute aid with unprecedented speed and precision. Approximately 140,000 residents received emergency cash transfers as a direct result of this targeting methodology. Compared to conventional methods, such as social registries, which were either unavailable or incomplete, the phone-based approach minimized misclassification, reducing both exclusion of eligible recipients and inclusion of ineligible individuals.

Critically, while traditional social registries might be more accurate, they were not feasible under the constraints imposed by the pandemic. The satellite and phone data approach demonstrated a unique advantage: it could be deployed entirely remotely within just a few weeks, addressing both the urgency and scale required during a crisis.

Outcomes and Impact

    • Scale of Aid: 140,000 individuals reached, translating to substantial immediate economic relief.
    • Efficiency: Aid delivery was faster and more precise compared to traditional methods, circumventing delays associated with field surveys and paper-based registries.
    • Innovation in Poverty Targeting: The initiative showcased the potential of integrating multiple data streams (satellite imagery and mobile metadata) into machine learning frameworks for social good.
    • Policy Implications: The success of the Togo case provides a replicable model for other countries facing crises without up-to-date poverty data, whether during pandemics, natural disasters, or conflicts.

Conclusion

The Togo Novissi program exemplifies how modern data science can directly support SDG 1 by effectively identifying and assisting those living in poverty. By leveraging satellite imagery and mobile phone data through sophisticated machine learning algorithms, the program reached 140,000 of the country’s most vulnerable residents with timely financial support. This initiative demonstrates that when governments, researchers, and NGOs collaborate strategically, data can become a powerful instrument for social good, enabling rapid, large-scale interventions that traditional methods alone cannot achieve. As countries worldwide seek innovative approaches to meet the SDGs, the Togo case provides both a blueprint and inspiration for using data to address poverty with precision, speed, and ethical rigor.

Leveraging Mobile and Satellite Data (Case Study: Togo)

In the quest to achieve the United Nations Sustainable Development Goals (SDGs), innovative applications of data science are becoming pivotal in addressing entrenched socio-economic challenges. One particularly illustrative example is Togo’s use of mobile phone and satellite data to effectively target emergency cash transfers during the COVID-19 pandemic, exemplifying SDG 1: No Poverty. This initiative demonstrates how advanced data analytics, machine learning, and cross-sector collaboration can translate raw data into meaningful social impact, rapidly identifying vulnerable populations in need.

Context and Challenge

In April 2020, amid the global COVID-19 pandemic, the Togolese government launched the Novissi program to provide emergency financial support to residents in the Greater Lomé region. The program aimed to mitigate the economic shock of lockdowns for individuals unable to work due to public health restrictions. Monthly cash transfers ranged from $10 to $20, targeted to residents who were registered to vote in specific regions and engaged in particular occupations.

However, Togo faced a critical challenge: the lack of recent, reliable poverty data. Traditional household surveys were outdated, costly, and logistically impractical due to social distancing measures. The government needed a solution that could accurately identify the poorest residents quickly and at scale, ensuring that aid reached those most in need without delays.

Source: UC Berkeley, Vice Chancellor for Research Office

Data-Driven Solution

To address this gap, a team of researchers led by UC Berkeley, in collaboration with the Togolese government, applied an innovative combination of satellite imagery and mobile phone data, analyzed using machine learning algorithms. Key contributors included UC Berkeley researchers, who designed and implemented the data models, the Government of Togo who oversaw program implementation and coordinated the distribution of funds via the Novissi mobile platform, Tech entities such as GiveDirectly, who facilitated cash transfers and provided operational expertise in digital payment systems, and the World Bank who provided financial support to conduct a large phone survey.

The methodology unfolded in two complementary stages:

    1. Identifying Vulnerable Regions via Satellite Data
      Satellite imagery was analyzed to determine the 100 poorest “cantons,” each comprising approximately three to four villages. Machine learning models extracted variables correlated with poverty, including roof material density and forest coverage. This approach allowed researchers to pinpoint areas with the highest concentration of impoverished residents, even in the absence of up-to-date census data.
    2. Targeting Individuals via Mobile Phone Data
      Within these selected cantons, mobile phone metadata was used to identify the poorest individuals. Machine learning algorithms assessed behavioral indicators such as the volume of international calls, frequency of mobile internet usage, and other usage patterns. These signals, when combined, provided a predictive model of poverty status, allowing for remote, rapid targeting of financial assistance.


Source: Nature

Implementation and Results

The combined satellite-phone approach enabled the government to distribute aid with unprecedented speed and precision. Approximately 140,000 residents received emergency cash transfers as a direct result of this targeting methodology. Compared to conventional methods, such as social registries, which were either unavailable or incomplete, the phone-based approach minimized misclassification, reducing both exclusion of eligible recipients and inclusion of ineligible individuals.

Critically, while traditional social registries might be more accurate, they were not feasible under the constraints imposed by the pandemic. The satellite and phone data approach demonstrated a unique advantage: it could be deployed entirely remotely within just a few weeks, addressing both the urgency and scale required during a crisis.

Outcomes and Impact

  • Scale of Aid: 140,000 individuals reached, translating to substantial immediate economic relief.
  • Efficiency: Aid delivery was faster and more precise compared to traditional methods, circumventing delays associated with field surveys and paper-based registries.
  • Innovation in Poverty Targeting: The initiative showcased the potential of integrating multiple data streams (satellite imagery and mobile metadata) into machine learning frameworks for social good.
  • Policy Implications: The success of the Togo case provides a replicable model for other countries facing crises without up-to-date poverty data, whether during pandemics, natural disasters, or conflicts.

Conclusion

The Togo Novissi program exemplifies how modern data science can directly support SDG 1 by effectively identifying and assisting those living in poverty. By leveraging satellite imagery and mobile phone data through sophisticated machine learning algorithms, the program reached 140,000 of the country’s most vulnerable residents with timely financial support. This initiative demonstrates that when governments, researchers, and NGOs collaborate strategically, data can become a powerful instrument for social good, enabling rapid, large-scale interventions that traditional methods alone cannot achieve. As countries worldwide seek innovative approaches to meet the SDGs, the Togo case provides both a blueprint and inspiration for using data to address poverty with precision, speed, and ethical rigor.