This piece is part of the weekly series “Growing Forward: Insights for Building Better Food and Agriculture Systems,” presented by the Global Food Institute at the George Washington University and the nonprofit organization Food Tank. Each installment highlights forward-thinking strategies to address today’s food and agriculture related challenges with innovative solutions. To view more pieces in the series, click here.
Recent events in Gaza, Sudan, and South Sudan remind us that famine and starvation remain harsh realities, not just distant threats, even in our modern world. In 2025, more than 295 million people across 53 countries faced crisis levels of acute food and nutrition insecurity—the highest number since global tracking began in 2017*. And the threats are only growing: conflict, climate change, and economic instability are all projected to increase the frequency and severity of these crises.
For those of us in high-income countries, it can be difficult to fully grasp the toll these crises take. When food becomes scarce or inaccessible, families often sell their few remaining assets to survive. The poorest—those with little or nothing to sell—are hit the hardest. In the worst scenarios, malnutrition and mortality rates surge, particularly among children.
Despite the complexity of these crises—and the constraints that conflict imposes on humanitarian and development assistance—there is still much we can do to prevent the worst from happening. Acting early saves lives and livelihoods and is far more effective than responding once a crisis is well underway.
In recent years, there’s been a growing recognition of the need for more proactive responses. But early action depends on early warning—and that’s where data becomes critical. Food security crises have many moving parts and drivers. Tracking them requires timely, reliable information and the ability to detect real signals amidst the noise. When data is missing, mistrusted, or disconnected from decision-making, we lose valuable time—and opportunities to prevent suffering.
Efforts to build comprehensive early warning systems began in earnest in the mid-1980s with the creation of the Famine Early Warning Systems Network (FEWS NET). Since then, the standard approach for food security analysis has involved conducting field surveys, analyzing the results, projecting future trends, and circulating published reports to guide decisions about aid.
This approach has been invaluable, but not without challenges. Conducting household surveys in person is expensive and slow. By the time results come in, the situation may have changed—like driving while looking in the rearview mirror. Projections can also be inconsistent, inaccurate, and difficult to replicate–undermining their reliability for decision-making. And despite all these efforts to collect data and project scenarios, warnings are not always acted upon in time.
Since 2018, I’ve had the privilege of leading a team at the World Bank Group working to close critical gaps in the availability, quality, and frequency of food and nutrition security data–and to better connect this information to timely, evidence-based decision-making. One area we’ve focused on is building data-driven models to help identify early signs of food crises. The idea itself isn’t new—but making it work at scale has always seemed out of reach, largely due to the multiple, interacting drivers of these crises and the limited historical data available on them.
Still, the appeal of a modeled approach has been too enticing to ignore. With a good model, we can generate more frequent updates quickly and at much lower cost than traditional methods. We also gain transparency—so anyone running the model gets the same results—and we open the door to more reliable, longer-range projections. Of course, a model does not replace our traditional toolkit, but it significantly enhances it.
With support from United Nations agencies like the U.N. Food and Agriculture Organization, World Food Programme, UNICEF, and World Health Organization, international NGOs, and major tech companies including Amazon, Google, and Microsoft, we launched an ambitious agenda. We tested everything from basic econometric models to machine learning and deep neural networks.
I’m proud to say that we’ve now successfully deployed this first-of-its-kind model in Somalia and Yemen. It’s able to capture past crisis conditions with about 80 percent accuracy, and recent improvements—developed in collaboration with the Integrated Food Security Phase Classification—are improving accuracy even further. We’re preparing to expand deployment to five countries this year with the goal to reach 15 by 2026. You can follow this work in real time through the Global Food and Nutrition Security Dashboard.
What excites me most is not just that this effort enhances how we detect crisis risks—it’s changing how decisions are made. From the outset, we’ve emphasized transparency, technical rigor, and practical relevance by ensuring our data connects directly to decisions about funding and programming. This work is now helping shape the development of national preparedness plans with governments and their partners to systematize earlier, more impactful, and better coordinated responses to these crises.
We’re only beginning to scratch the surface of what’s technically possible in this space. Looking ahead, advances in higher-frequency indicators–combined with rapid innovations in AI–will provide earlier warnings and greater lead time to respond to emerging crises. The World Bank Group is also deepening its commitment to this agenda through the launch of the new Global Challenge Program on Food and Nutrition Security, which seeks to sharpen our focus on high-impact, high-return interventions that integrate social protection, food, and health measures. We are moving away from siloed approaches toward more integrated solutions—ensuring that the private sector also plays a central role in our engagements. But especially now—at a time when the value of international aid is being questioned—innovations like these feel especially critical. They have the potential to drive smarter, more cost-effective, and more accountable responses—marking a critical step forward in our broader effort to prevent, and ultimately end, food crises for good.
*The international community relies on a range of indicators to assess the severity and duration of food insecurity, based on the widely accepted framework of access, availability, utilization, and stability. While these measures form the foundation of food security analysis, they can be confusing—even for practitioners. The FAO offers a useful overview of the main indicators and their purposes, along with a short quiz to help test your understanding of the differences.
Photo courtesy of the U.S. Agency for International Development
