Fix the Broken Map: Rebuilding America's Homeless Resources Data

## Overview Across America, a deadly information gap exists between people experiencing homelessness and the resources meant to help them. Shelter hotlines list disconnected phone numbers. Online directories point to closed facilities. Capacity information is hours or days out of date. The result: people in crisis spend precious hours calling numbers that don't work, walking to addresses that don't exist, and missing opportunities for shelter, meals, and services. This isn't a problem of missing data. Multiple systems collect homeless resources information: HUD Continuum of Care databases, 211 information systems, municipal homeless services portals, nonprofit directories, and street outreach records. The problem is that these sources contradict each other, contain errors, and become outdated within months of publication. When datasets disagree, frontline workers and people seeking help have no way to know which information is correct. During extreme weather events, this information crisis becomes life-threatening. Winter storms and summer heat waves kill homeless individuals who couldn't find shelter because the information they had was wrong. This challenge asks: can we build systems that ensure accurate, real-time homeless resources data reaches people when they need it most? ## Background The fragmentation of homeless resources data creates cascading failures throughout the system. A shelter changes its phone number, but only updates it in two of the five databases where it's listed. A facility expands capacity, but the information takes three weeks to propagate. An organization closes its doors, yet remains listed in directories for months. Current approaches to maintaining this data rely heavily on manual updates, periodic surveys, and phone verification campaigns. These methods are labor-intensive, expensive, and immediately out of date. By the time someone calls to verify a shelter's information, circumstances may have already changed. The result is a constant state of data decay where accuracy rates of 50-60% are considered acceptable. The consequences hit hardest during crises. When a polar vortex hits, emergency shelters open overflow capacity, but this information may not reach coordination systems for hours or days. When heat waves strike, cooling centers activate, but outdated contact information means vulnerable populations can't find them. Outreach workers waste time traveling to facilities that have moved or closed, time they could spend connecting people to actual resources. Meanwhile, advances in data integration, conflict resolution algorithms, automated validation techniques, and AI-powered information maintenance offer new possibilities. Systems that can cross-reference multiple data sources, identify discrepancies, validate information through various signals, and flag probable errors could dramatically improve data accuracy and freshness. ## What We're Looking For We're seeking solutions that address the core challenge of maintaining accurate, up-to-date homeless resources data across fragmented systems. Your approach should demonstrate: Data Integration and Conflict Resolution: Methods to combine multiple data sources (government databases, nonprofit directories, 211 systems, outreach records) and resolve conflicts when sources disagree. How do you determine which source is most reliable? How do you handle partial information or missing fields? Automated Validation: Techniques to verify information accuracy without relying solely on manual phone calls. This might include web scraping for facility websites, social media monitoring, pattern analysis of call outcomes, or other signals that indicate whether listed information is current and correct. Continuous Maintenance: Systems that keep data fresh over time, flagging information that's likely outdated based on age, change patterns, or other indicators. How do you identify when information needs reverification? How do you prioritize validation efforts? Real-World Application: Your solution should show how improved data reaches the people who need it: individuals seeking shelter, outreach workers, emergency coordinators, or service providers. Consider the full chain from data improvement to actual impact. Your submission should include a working prototype or detailed technical approach, demonstration using real or realistic test data (with appropriate privacy protections), documentation of your methodology, and clear metrics showing improvement over current approaches. ## Expected Outcomes Successful solutions will demonstrate measurable improvements in data accuracy, freshness, and usability. We expect to see approaches that can achieve 90%+ accuracy rates for critical contact information (phone numbers, addresses, hours) and identify outdated information within days rather than months. The best solutions will be practical and scalable. They should work with existing data sources and systems rather than requiring wholesale replacement. They should reduce the manual effort required to maintain accurate data while improving outcomes. They should be deployable in cities and regions of varying sizes and technical sophistication. Ultimately, we're looking for solutions that save lives by ensuring that when someone experiencing homelessness seeks help, the information they receive is accurate the first time. Your work could become the foundation for better homeless services coordination across the country, helping cities turn fragmented data into reliable, life-saving information.