Assessing hunger in the age of big data: a decade of remote monitoring

Hunger poses a perennial challenge that requires timely and cost-effective data for effective humanitarian action. Over the past decade, the World Food Programme (WFP) has pioneered the adoption of a remote survey methodology that provides real-time data, marking a transformational step in producing the evidence needed for the fight against global hunger. Here, we explore the operational experiences within the humanitarian sector, the lessons learned, and the path forward.

The starting point

Food security is dynamic. Traditionally, the measurement of food security relied heavily on face-to-face surveys and assessments, which were time-consuming, resource-intensive and often delayed critical evidence needed to respond in crisis situations.

In 2013, WFP initiated a remote data-collection pilot in refugee camps in the Democratic Republic of Congo (DRC), marking the inception of a real-time, data-driven approach in humanitarian action. The initiative, known as mVAM (mobile Vulnerability Analysis and Mapping), leveraged mobile technologies like phone surveys, SMS and chatbots to collect high-frequency food security data, particularly from hard-to-access areas, setting the stage for a scalable, cost-effective, and efficient data-collection system. The advent of remote monitoring technologies catalysed a paradigm shift, enabling real-time data collection, analysis and visualisation even from places that were traditionally not accessible. This approach has revolutionised the measurement of food insecurity.

Operational experiences

Since then, remote monitoring has become a key tool for evidence generation in inaccessible areas and during rapid-onset crises. For instance, during the Ebola crisis in West Africa in 2014, traditional assessment methods were severely limited due to quarantines and movement restrictions. WFP’s remote systems became instrumental in generating the data needed for humanitarian operations across Guinea, Sierra Leone and Liberia. Similarly, the use of remote monitoring during the Mozambique humanitarian crisis post-Cyclone Idai facilitated immediate data collection, helping in understanding the extent of damage and the immediate needs of the affected population.

By 2019, advancements in big data analytics allowed WFP to process thousands of surveys daily, through automated data pipelines and visualisation, forming the world’s first real-time food security monitoring system. This system further expanded during the Covid-19 pandemic, scaling from 11 countries to 35, and becoming a primary source of food security information when traditional face-to-face assessments became nearly impossible. Moreover, the use of AI methodologies has enabled the nowcasting of key food security indicators in up to 50 countries where data collection has not been feasible.

In addition to WFP, many other organisations have deployed similar approached to enable rapid data collection and evidence generation. For example, UNICEF has utilised remote monitoring to assess the social and economic welfare of households post-pandemic. Similarly, the Food and Agriculture Organization’s (FAO’s) Global Information and Early Warning System (GIEWS) employs remote monitoring to deliver timely updates on global food and agriculture situations.

Risks and challenges

Assessing thousands of households daily, and assisting millions every year, data collection presents several challenges for an operation as sizable as WFP’s, including how to address the privacy and ethical concerns associated with remote data collection in crisis situations. WFP, with its longstanding commitment to data protection, strives to ensure that the privacy of vulnerable populations is safeguarded in every aspect of its operations, whether through traditional on-the-ground efforts or cutting-edge remote monitoring systems. To that end, WFP and our partners strictly adhere to established data protection standards and apply industry-leading encryption and data management tools to mitigate any data-exposure risks.

The digital divide is another key challenge for remote surveys. Households with limited access to phones, sim cards and network, or in certain cases, charging devices or power are excluded or underrepresented in remote surveys. Therefore, the population reached through mobile phone surveys may result in a sample that is artificially wealthier and more urban that the actual population in the country. Additionally, contacting specific population groups through phone surveys might be challenging due to socioeconomic norms – for example, women might be less willing to respond to phone surveys from unknown numbers.

Lastly, resource constraints form another critical challenge. Securing sufficient funding for technology procurement, personnel training, and maintenance of remote monitoring systems is essential. Additionally, the demand for technical expertise to manage and analyse remotely collected data can pose a barrier for some humanitarian organisations, limiting the scope of their engagement. 

Lessons learned and best practices

In remote data collection, prioritising data accuracy and reliability is vital and requires a customised approach. Adjusting methodologies, such as refining questionnaires and training enumerators, is key. An iterative system design, focusing on the right sampling methods, helps meet information needs and reduces bias. Critical to this process are detailed call statistics, which guide sample size and frame decisions, and verification processes that validate data collection methods. A thorough examination of mode and tool biases is essential, leading to a comprehensive data quality assurance plan covering the entire data collection lifecycle, from questionnaire design to call centre operations.

Before implementing remote data collection, assessing network coverage and phone/SIM card ownership is crucial, as is evaluating the feasibility of reaching target populations, such as specific groups like refugees or displaced people. This ensures the effectiveness of remote survey methodologies. Addressing sampling bias in remote surveys is important, necessitating validation with other data sources and socio-demographic analysis, and applying appropriate weighting methods for representativeness.

Remote surveys can also be used to assess food insecurity drivers. Conducting mode experiments and validation studies before adding new questions or indicators helps identify biases and guides statistical adjustments. The survey mode can influence respondent answers, particularly for sensitive questions, leading to biases.

Investing in technology and capacity-building is essential for sustainable systems, enhancing digital skills among humanitarian workers and communities and ensuring access to digital technologies, especially in underserved areas.

Finally, securing sustainable funding is crucial for these initiatives’ longevity. Exploring various funding models and focusing on cost-efficiency is necessary in an era of limited humanitarian budgets, ensuring resource maximisation.

The impact: more than just numbers

In hard-to-reach areas across the world, WFP’s real-time monitoring system has consistently provided the first food security data, enabling quicker and more targeted interventions. The data and evidence generated by the real-time monitoring systems have been keys factor in combating food insecurity globally, delivering timely information, enabling evidence-based decision-making, and facilitating early warnings, thereby making a significant impact on the lives of millions of individuals.

For example, in Haiti, deteriorations in food security were detected months before traditional assessments could have caught them, showcasing the efficacy of real-time monitoring. The immediate detection of food security issues was also seen in conflict zones like Afghanistan and Ukraine, where real-time monitoring facilitated quicker and more targeted interventions.

Outside WFP, the ability of organisations like the UN Children’s Fund (UNICEF), the Famine Early Warning Systems Network (FEWS NET), the UN Refugee Agency (UNHCR), the FAO, and others to provide early warnings and timely updates on household vulnerability through remote monitoring has significantly contributed to better informed decision-making and response. Such operational experiences underscore the transformative power of real-time monitoring in not just understanding the scale and scope of food insecurity but also in ensuring that aid is delivered to those in need promptly and effectively, thereby making a tangible impact on the lives of millions facing hunger.

The road ahead

As we reflect on the past decade, it’s clear that digital technologies have revolutionised humanitarian action. The journey ahead requires us to continue evolving alongside technological advancements while nurturing a culture of ongoing learning and collaborative innovation. This approach is key to boosting the effectiveness and impact of humanitarian programmes, particularly in our collective efforts to combat global hunger.

WFP, and similar humanitarian organisations, are now turning towards advanced machine-learning methodologies to predict food security trends well in advance. This innovative approach is currently being piloted in countries like Cameroon, Syria, Yemen, Haiti and Nigeria. It represents a paradigm shift, enabling us to respond proactively and mitigate crises before they intensify. Although these methodologies are in their nascent stages, the potential they hold is immense. By enabling more effective responses, we can save more lives and ensure that our resources are utilised more efficiently and impactfully. This is not just a step forward; it’s a leap towards a future where our fight against hunger is proactive, data-driven, and more effective than ever before.

Kyriacos M. Koupparis, PhD is Head of Hunger Monitoring Unit, United Nations World Food Programme. Kusum Hachhethu is Food Security Analysis Team Lead, United Nations World Food Programme.