The use of non-traditional data is now a regular feature of applied research and policy analysis across a wide range of domains. Sources such as satellite imagery, mobile network data, retail transactions, social media content, wastewater signals, and sensor-generated data provide detailed, high-frequency signals that conventional surveys and administrative systems often do not capture. This data enables more timely and geographically detailed analysis of environmental conditions, economic activity, infrastructure systems, public health risks, and human behavior, especially when combined with surveys, administrative records, geospatial datasets, or other reference sources to improve coverage, refine estimates, or reveal patterns that would be difficult to observe through traditional data alone.

This update continues our ongoing monitoring of how non-traditional data is being reused for public-interest research and decision support (See prior updates: 1, 2, 3, and 4). Earlier editions documented examples across public health, economic measurement, mobility, environmental monitoring, urban systems, and governance. The cases collected here, published or released between January and April 2026, add further evidence on where these data sources are being applied and the kinds of questions they are helping researchers and practitioners address.

The examples in this update span multiple areas of public interest, illustrating how these data sources support measurement, monitoring, and decision-making in practice. Each entry summarizes the focus of the work, the role of non-traditional data, and why the example matters for public-interest analysis .

This update is organized into the following sections:

• Health, Healthcare Access, and Epidemiology

• Human Mobility and Population Dynamics

• Economic Measurement and Development

• Consumer Behavior, Markets, and Labor Dynamics

• Public Perception and Social Media

• Human Behavior and Smart Devices

• Urban Systems, Infrastructure, and Spatial Planning

• Agriculture and Food Production Systems

• Environmental Monitoring and Natural Systems

Each section highlights recent examples that demonstrate how non-traditional data is being applied to that problem domain, with an emphasis on practical use cases and measurable outcomes.

Health, Healthcare Access, and Epidemiology

OpenAI. “AI as a Healthcare Ally: How Americans are Navigating the System with ChatGPT.” January 2026. https://cdn.openai.com/pdf/2cb29276-68cd-4ec6-a5f4-c01c5e7a36e9/OpenAI-AI-as-a-Healthcare-Ally-Jan-2026.pdf

Focus: Examines how individuals use AI tools to navigate healthcare systems, access information, and manage care. 

Role of Non-Traditional Data: The analysis drew on large-scale, anonymized user interaction data from ChatGPT, including billions of healthcare-related queries. This data captured when and how people sought information, what topics they searched for, and how usage varied across locations. This provided a real-time view of information needs and barriers to accessing care.

Why This Matters: The findings highlight gaps in how people navigate healthcare systems, especially outside clinical settings and standard hours. The data reveals unmet needs related to insurance, billing, and access, particularly in underserved areas. These insights demonstrate how user-generated data can complement conventional health system data by capturing real-time demand for information and support, informing efforts to improve patient support and reduce friction in healthcare access.

Rey, Samantha. “Researchers to Use Retail Loyalty Card Data to Identify Early Signs of Cancer.” Imperial College London News. February 2, 2026. https://www.imperial.ac.uk/news/articles/2026/researchers-to-use-retail-loyalty-card-data-to-identify-early-signs-of-cancer/

Focus: Investigates whether patterns in retail purchasing behavior can reveal early warning signs of cancer before formal diagnosis.

Role of Non-Traditional Data: The study analyzed retail loyalty card data from Boots and Tesco, using detailed records of individuals’ purchasing behavior over time. Researchers focused on purchases of over-the-counter medications and compared patterns between cancer patients and healthy individuals. The dataset included approximately 3,000 consenting participants and allowed researchers to identify subtle changes in behavior prior to diagnosis.

Why This Matters: Earlier work showed that changes in purchasing patterns could signal ovarian cancer up to eight months before diagnosis. Expanding this approach across multiple cancer types could enable earlier detection and intervention, improving outcomes by prompting individuals to seek medical advice sooner.

DeJonge, Peter M. V., Ian Pray, Rachel Poretsky, Martin Shafer, Sandra L. McLellan, Alyse Kittner, Colin Korban et al. “Wastewater-Based Epidemiology to Enhance Public Health Preparedness and Response During Large-Scale Events: Experiences from the 2024 Republican and Democratic National Conventions – Milwaukee, WI and Chicago, IL.” Frontiers in Public Health, Volume 14, Article 1800385. March 19, 2026. https://doi.org/10.3389/fpubh.2026.1800385 

Focus: Examines how wastewater data can be used to monitor infectious disease risks during large-scale events.

Role of Non-Traditional Data: The study used wastewater samples as a population-level data source, capturing biological signals from entire communities. Researchers collected and analyzed samples before, during, and after two major events, testing for multiple pathogens. They combined these results with traditional public health data, such as hospital visits, to track changes in disease levels over time. 

Why This Matters: Wastewater monitoring provides continuous, low-cost insight into disease spread without requiring individual testing. This is especially valuable during large events where traditional surveillance may be incomplete or delayed, helping public health authorities detect risks earlier and respond more effectively.

Human Behavior and Population Dynamics

Andrich, Paolo, Shengjie Lai, Halim Jun, Qianwen Duan, Zhifeng Cheng, Seth R. Flaxman & Andrew J. Tatem. “Social Media Data for Population Mapping: A Bayesian Approach to Address Representativeness and Privacy Challenges.” arXiv preprint. January 30, 2026. https://arxiv.org/abs/2601.22104

Focus: Examines how social media data can be used to estimate population distribution in near real time, particularly for disaster response.

Role of Non-Traditional Data: The study used aggregated Facebook user location data to estimate where people were located across different regions and time periods. Researchers combined this data with additional signals such as nighttime light intensity and urbanization patterns to provide additional context. They developed statistical methods to adjust for missing data and uneven social media use, and calibrated the results against census data to improve accuracy.

Why This Matters: The analysis showed that social media data can provide timely population estimates in rapidly changing situations such as disasters. It also demonstrated that without adjustment, this data can underrepresent rural and low-density areas, but targeted corrections can improve reliability for real-world use.

Elejalde, Erick, Timur Naushirvanov, Kyriaki Kalimeri, Elisa Omodei, Márton Karsai, Loreto Bravo & Leo Ferres. “Use of Mobile Phone Data to Measure Behavioral Response to SMS Evacuation Alerts.” International Journal of Disaster Risk Reduction, Volume 131, 105919. December 2025. https://doi.org/10.1016/j.ijdrr.2025.105919

Focus: Analyzes how populations respond to emergency evacuation alerts during wildfires by measuring changes in mobility.

Role of Non-Traditional Data: The study used anonymized mobile phone network data, capturing activity from approximately 580,000 devices at 15-minute intervals. Researchers tracked changes in connections to cell towers as a proxy for population movement and compared these patterns before and after evacuation alerts. They combined this information with alert records and socioeconomic data to understand differences across communities.

Why This Matters: The analysis found that initial alerts triggered strong movement, while repeated alerts led to weaker responses. It also showed that higher-income areas responded more quickly and that movement occurred even in areas without direct alerts, indicating spillover effects such as congestion or precautionary evacuation.

Huang, Ming-Wey, Chia-Ying Lin, Ming-Chun Ke, Wei-Sen Li & Tzu-Yin Chang. “Analysis of Human Flow during a Natural Disaster Utilizing Trajectory-Free Mobile Network Data: A Case Study of Earthquake.” Scientific Reports, Volume 16, Article 5275. 2026.  https://doi.org/10.1038/s41598-026-36255-1

Focus: Examines how population movement changes during an earthquake using aggregated mobile network data.

Role of Non-Traditional Data: The study used aggregated mobile network data showing how many devices were connected to cell towers over time. Researchers grouped these data points into small geographic areas and tracked how counts changed before and after an earthquake. This allowed them to estimate population shifts without tracking individual movements.

Why This Matters: The analysis showed that population movement shifted quickly after the earthquake, with clear directional patterns emerging within the first hour. These insights can support faster and more informed disaster response without relying on detailed individual tracking.

Economic Measurement and Development

Aiken, Emily, Joshua E. Blumenstock, Sveta Milusheva & M. Merritt Smith. “Predicting Well-Being with Mobile Phone Data: Evidence from Four Countries.” arXiv preprint. February 2, 2026. https://arxiv.org/pdf/2602.02805

Focus: Examines how mobile phone data can be used to estimate well-being, including poverty and vulnerability, across multiple countries.

Role of Non-Traditional Data: The study uses mobile phone metadata, including records of calls, text messages, airtime top-ups, mobile data usage, and mobile money transactions, as indicators of economic behavior. Researchers linked this data to household survey responses from four countries, enabling direct comparison between digital activity patterns and measured well-being. They constructed features capturing communication patterns, mobility, and usage behavior, and applied machine learning models to predict outcomes such as asset-based wealth, consumption, and food security. 

Why This Matters: The analysis showed that mobile phone data can predict poverty-related outcomes with moderate accuracy and that performance improves significantly with more representative data. This creates a path toward faster and lower-cost measurement of economic well-being in places where traditional surveys are infrequent or incomplete.

Maduako, Iyke, Dharana Rijal & Alberto Sanchez Rodelgo. “Satellite Data for Nowcasting: Estimating Cambodia’s GDP in Real Time Using Satellite Data in a Machine Learning Framework.” IMF Selected Issues Paper No. 2026/001, January 2026. https://www.imf.org/-/media/files/publications/selected-issues-papers/2026/english/sipea2026001.pdf 

Focus: Examines how satellite-based indicators can be used to estimate economic activity in near real time, addressing delays and gaps in official GDP statistics.

Role of Non-Traditional Data: The study used satellite-derived measures such as nighttime light intensity, air pollution levels, rainfall patterns, and vegetation conditions to track changes in economic activity. Researchers combined these signals with traditional economic data in a predictive model that links environmental patterns to output. Because satellites capture data frequently and across all regions, they provide a continuous view of economic activity at a finer geographic level than official statistics.

Why This Matters: Incorporating satellite data improved the accuracy of GDP estimates by more than 20% and made it possible to observe economic differences within the country that are not visible in national statistics. This supports more timely and targeted economic decision-making, particularly in data-constrained settings. 

Sherman, Luke, Jonathan Proctor, Hannah Druckenmiller, Heriberto Tapia & Solomon Hsiang. “Global High-Resolution Estimates of the UN Human Development Index Using Satellite Imagery and Machine Learning.” Nature Communications, Volume 17, Article 1315. January 2026. https://doi.org/10.1038/s41467-026-68805-6

Focus: Develops a method to estimate the Human Development Index (HDI) at a much finer geographic level using satellite imagery and machine learning.

Role of Non-Traditional Data: The study used large-scale satellite imagery, including daytime images and nighttime lights, as indicators of socioeconomic conditions. Researchers trained models using existing regional HDI data and applied them to estimate development levels for smaller geographic areas, including tens of thousands of municipalities and grid cells. This approach combined satellite data with official statistics to produce more detailed estimates without requiring new surveys.

Why This Matters: The results reveal large differences in development within countries that are hidden in national or regional averages. The analysis shows that more than half of the global population would be placed in a different development category when using high-resolution data, highlighting the importance of more precise measurement for policy targeting and resource allocation.

Consumer Behavior, Markets, and Labor Dynamics

Desiderio, Antonio, Alessia Galdeman, Franziska Bäuerlein & Sune Lehmann. “Mapping Regional Disparities in Discounted Grocery Products.” npj Science of Food, Volume 10, Article 112. February 20, 2026. https://doi.org/10.1038/s41538-026-00764-0

Focus: Examines how the spatial distribution of discounted near-expiry grocery products reflects patterns in food waste, retail behavior, and supply-demand dynamics.

Role of Non-Traditional Data: The study analyzed high-frequency retail data obtained through a public API, capturing daily information on discounted near-expiry products across 542 stores over a 153-day period. Researchers combined this data with geospatial datasets from OpenStreetMap to understand store location and proximity, and with product-level data from OpenFoodFacts data on nutrition and environmental impact. They used these combined datasets to identify patterns in how and where discounting occurs.

Why This Matters: The analysis showed clear regional differences in discounting, with rural stores more frequently discounting perishable items like meat and dairy, while urban stores focused more on convenience products. These patterns point to local mismatches between supply and demand that contribute to food waste. Understanding these differences can help retailers and policymakers design more targeted strategies to reduce waste and improve inventory management .

Hiltz, Barbara S., Bryan G. Victor & Brian E. Perron. “From Job Postings to Curriculum Decisions: Using AI to Generate Workforce Intelligence for MSW Program Planning.” University of Michigan & Wayne State University. March 6, 2026. https://arxiv.org/pdf/2603.06839

Focus: Examines how large-scale job posting data can be used to align education curricula with real-time employer demand in social work.

Role of Non-Traditional Data: The study analyzed more than 41,000 online job postings collected from platforms such as Indeed, LinkedIn, and Glassdoor. Researchers used AI methods to categorize roles and extract information about required skills, competencies, and areas of specialization. They then examined patterns in employer demand across different types of positions.

Why This Matters: The analysis found that around 70% of job postings focused on clinical roles and that a consistent set of core skills appeared across specialties. These insights provide a clearer picture of workforce demand and can help education providers adjust curricula to better prepare graduates for the roles that are actually available.

Public Perception and Social Media

Botas Etcheverría, Bruno, Jenny Alexandra Cifuentes Quintero, Yury Andrea Jiménez Agudelo & Maria Espinosa Ruiz. “Public Perception on Immigration and Racial Discrimination in Spain: A Social Media Analysis Using X Data.” Journal of Computational Social Science, Volume 9, pp. 25-1–25-39. January 27, 2026. https://link.springer.com/article/10.1007/s42001-025-00459-8

Focus: Examines public perceptions of immigration and racial discrimination in Spain using social media data.

Role of Non-Traditional Data: The study analyzed large volumes of user-generated content from X to identify common themes and overall sentiment. Researchers applied automated text analysis methods to group discussions by topic and classify whether posts expressed positive or negative views.

Why This Matters: The analysis showed that negative sentiment dominated across several areas of the discussion and increased around specific events. These patterns provide a real-time view of how public attitudes evolve and how online platforms shape public debate.

Human Behavior and Smart Devices 

Fiori, Michele, Gabriele Civitarese, Flora D. Salim & Claudio Bettini. “DomusFM: A Foundation Model for Smart-Home Sensor Data.” arXiv preprint. February 2, 2026. https://arxiv.org/pdf/2602.01910

Focus: Examines how smart-home sensor data can be used to understand and predict human activity in everyday environments.

Role of Non-Traditional Data: The study used data from smart-home sensors such as motion detectors, door sensors, and environmental monitors. These devices generated continuous streams of simple signals, such as whether a door opened or movement was detected. Researchers combined multiple publicly available smart-home datasets and analyzed these patterns over time to identify common behaviors and routines.

Why This Matters: The analysis showed that these passive data streams can capture detailed patterns of daily life without directly observing individuals. This supports applications such as health monitoring, assisted living, and home automation, while maintaining privacy by relying on indirect signals rather than cameras or personal reporting.

Urban Systems, Infrastructure, and Spatial Planning

Bektemyssova, Gulnara and Galymzhan Shaikemelev. “An Intelligent Approach to Public Transport Route Optimization Based on Mobile Network Data and Geographic Information Systems.” INASS Express, Vol. 2, Article 5. April 17, 2026. https://doi.org/10.22266/inassexpress.2026.005 

Focus: Examines how mobile network data can be used to model urban transport demand and optimize public transport routes in rapidly growing cities.

Role of Non-Traditional Data: The study used anonymized mobile network data, aggregated into small geographic areas and time intervals, to estimate where people were located and how they moved across the city. Researchers treated the number of mobile users in each area as a proxy for demand and combined this with data on existing transport routes and infrastructure from OpenStreetMaps. They then used machine learning models to evaluate how well current routes matched demand and identify opportunities for improvement. 

Why This Matters: The analysis reveals mismatches between where people travel and where services are provided, enabling targeted adjustments such as adding routes in high-demand areas or reallocating resources. This supports more efficient and responsive public transport systems.

Czaplicki, Nicole, Colin J. Shevlin, Hector R. Ferronato, Aidan D. Smith, Dwarakh V. Nayam, Lei Peng, Scott W. Springer & Doren Walker. “A Blended Data Approach to Measuring Monthly Housing Starts: Satellite Imagery, Survey Data and More!” NBER Working Paper No. 35113. April 2026. http://www.nber.org/papers/w35113

Focus: Develops a method to produce monthly estimates of housing starts by combining satellite imagery with traditional survey data.

Role of Non-Traditional Data: The study used high-resolution satellite imagery collected on a monthly basis to directly observe new residential construction activity. Researchers applied image analysis techniques to identify different stages of construction and detect likely housing starts. They combined these observations with administrative and survey data, including building permits and construction surveys, and incorporated geospatial data such as building footprints and road networks to improve accuracy.

Why This Matters: Satellite data enabled direct observation of construction activity at scale, reducing reliance on interviews and improving coverage of housing starts, including projects that may be missed or delayed in traditional reporting. This improves the ability of policymakers and analysts to track housing supply in near real time and respond to changes in the market.

DataVLab. “Case Study: Annotating Urban Expansion with AI and Satellite Maps.” April 22, 2026. https://datavlab.ai/post/case-study-annotating-urban-expansion-with-ai-and-satellite-maps

Focus: Examines how satellite imagery and AI can be used to track urban expansion and classify different types of city growth over time.

Role of Non-Traditional Data: The study used high-frequency satellite imagery collected over a ten-year period,  combined with geospatial data from OpenStreetMap and municipal zoning datasets, to track changes in urban areas. Researchers manually labeled more than 400 satellite images to identify different types of development, including residential, industrial, and informal settlements, and used these labeled examples to train AI models that automatically detect and classify new growth. The system then produced maps showing how cities expanded and what types of development occurred.

Why This Matters: The results highlight where and how cities are growing, including patterns that diverge from planned development. This provides concrete evidence to support zoning updates, infrastructure planning, and efforts to manage informal expansion.

Agriculture and Food Production Systems

Monday Robotics Editorial. “Precision Farming in 2026: How AI and Satellite Data Are Transforming Crop Management.” Monday Robotics. March 27, 2026. https://mondayrobotics.com/blog/precision-farming-ai-satellite-2026

Focus: Explores how farms use AI and satellite data to optimize crop management, reduce input use, and improve yields through precision agriculture techniques.

Role of Non-Traditional Data: The article describes how farmers rely on high-frequency satellite imagery, combined with drone data and in-field sensors, to monitor conditions across their fields. These data streams capture changes in soil moisture, plant health, and temperature, allowing AI systems to detect early signs of disease, pest activity, or water stress. Farmers used these insights to guide targeted actions, adjusting water, fertilizer, and pesticide use in specific areas as opposed to applying them uniformly

Why This Matters: Farms using these approaches reported input cost reductions of 15-25% and crop loss reductions of 15-30%, while improving yield predictions by up to 30%. This makes agricultural production more efficient and better able to respond to changing conditions.

Environmental Monitoring and Natural Systems

Lusk, Daniel et al. “Crowdsourced biodiversity monitoring fills gaps in global plant trait mapping.” Nature Communications, Volume 17, 1203. January 30, 2026. https://doi.org/10.1038/s41467-026-68996-y

Focus: Develops global, high-resolution maps of plant characteristics to improve understanding of how vegetation responds to different environmental conditions.

Role of Non-Traditional Data: The study used large-scale citizen science observations alongside vegetation plot data, plant trait databases, and satellite-based environmental data. Citizen-reported species observations provided broad geographic coverage, which researchers linked to plant characteristics using statistical models. This allowed them to estimate traits in areas where direct field measurements were limited or unavailable.

Why This Matters: The approach fills major geographic gaps in plant data, especially in regions that are difficult or costly to survey. It demonstrates how crowdsourced observations can improve coverage and consistency, thereby strengthening the data used in large-scale ecological models that inform climate, biodiversity, and land management decisions.

Periopsis. “AI for Cleaner Cities: Automated Illegal Dumping Detection at Scale.” February 5, 2026. https://www.periopsis.com/blog/ai-for-cleaner-cities-automated-illegal-dumping-detection-at-scale/

Focus: Examines how AI and satellite imagery can be used to detect illegal waste dumping and support environmental monitoring.

Role of Non-Traditional Data: The study used more than 30,000 satellite images collected from different regions, combined with annotated examples of dumping sites, to train a machine learning model that identifies waste in imagery. The system processes satellite data through an automated pipeline to flag locations where dumping was likely occurring.

Why This Matters: The results show that large areas can be monitored continuously without relying on manual inspections. This allows authorities to identify dumping sites earlier, prioritize enforcement actions, and respond more quickly to environmental risks.

Smith, Oliver. “Global Crackdown on Illegal Logging Intensifies in 2026.” Informed Clearly. January 22, 2026. https://informedclearly.com/en/environment/36031/global-crackdown-illegal-logging-intensifies-2026

Focus: Examines how governments and international organizations are increasing enforcement efforts against illegal logging using new monitoring technologies and data sources.

Role of Non-Traditional Data: The article describes how authorities use satellite imagery, remote sensing data, and AI-based monitoring systems to detect deforestation and illegal logging activity in near real time. These systems analyze changes in forest cover across large geographic areas and flag suspicious activity, enabling enforcement agencies to identify logging hotspots, track patterns of land-use change, and prioritize inspections. The data is often combined with geospatial mapping tools and environmental datasets to support coordinated responses across jurisdictions.

Why This Matters: Earlier detection of illegal logging reduces the extent of environmental damage and improves the ability of authorities to intervene effectively. This can strengthen enforcement in regions where illegal activity has historically been difficult to monitor.

Knoblauch, Steffen, Ram Kumar Muthusamy, Hao Li, Iddy Chazua, Benedcto Adamu, Innocent Maholi & Alexander Zipf. “Assessing Building Heat Resilience Using UAV and Street-View Imagery with Coupled Global Context Vision Transformer.” arXiv. January 16, 2026. https://arxiv.org/pdf/2601.11357

Focus: Examines how aerial and street-level imagery can be used to identify building features that influence heat exposure in cities.

Role of Non-Traditional Data: The study used drone imagery and street-view images to capture building features such as rooftops, walls, and surrounding vegetation. Researchers combined this with building maps and temperature data to identify which characteristics were associated with higher or lower heat exposure. They trained a model to extract attributes such as roofing material, wall material, vegetation presence, building density, and surface brightness from imagery and linked these features to observed temperature patterns.

Why This Matters: The analysis highlights specific building and environmental features that reduce heat exposure, providing practical guidance for urban design and retrofitting. This is especially relevant for cities facing increasing heat risks and aiming to protect vulnerable populations.

Firoze, Adnan, Akshaj Uppala, Lindsay Darling, Raymond A. Yeh, Bedrich Benes, Brady Hardiman, Songlin Fei & Daniel Aliaga. “Where Are the City Trees? Monitoring Urban Trees across the U.S. Using Generative AI.” Communications of the ACM. March 30, 2026. https://cacm.acm.org/research/where-are-the-city-trees-monitoring-urban-trees-across-the-u-s-using-generative-ai/

Focus: Develops a scalable method to identify and monitor individual urban trees across U.S. cities using satellite imagery and AI.

Role of Non-Traditional Data: The study used high-frequency satellite imagery to observe vegetation patterns over time and combined it with map data, such as building outlines and road networks from OpenStreetMap, to provide urban context. Researchers applied AI methods to detect tree cover and estimate individual tree locations, producing a large dataset covering hundreds of cities. This produced a national-scale dataset of approximately 278 million urban trees and supported repeated analysis using updated imagery.

Why This Matters: The method provides a consistent way to measure urban tree coverage at scale, including in areas not captured in official records. This makes it possible to identify gaps in tree distribution across neighborhoods and supports more targeted urban planning and environmental interventions.

Reflections

1. Non-traditional data is most useful where existing measurement is slow, incomplete, or too coarse 

Many of the examples in this update address persistent information gaps: estimating GDP between official releases, mapping human development below the national level, monitoring urban trees on private land, tracking illegal dumping across large areas, and detecting population movement after disasters. In these cases, non-traditional data adds timeliness, geographic detail, and visibility into places or behaviors that conventional systems often miss. 

2. Certain data sources are reliable tools for recurring public-interest problems

There are clear patterns around which data types fit which questions. Satellite imagery is repeatedly used to monitor land, infrastructure, crops, forests, housing, and urban growth. Mobile network data is used to understand mobility, evacuation, transport demand, and economic well-being. Retail and platform data help reveal behavior that occurs outside formal systems, from job demand to healthcare navigation and medication purchases. Wastewater data continues to mature as a tool for population-level disease surveillance. These recurring applications suggest the gradual emergence of a more standardized playbook for NTD reuse, while also highlighting the growing need for shared methodological, metadata, and quality standards.

3. The strongest applications combine non-traditional data with trusted reference points

Across the examples, researchers often combined non-traditional data with surveys, administrative records, census data, clinical data, permits, maps, or ground observations. This improved accuracy and interpretation, but also helped address the limits of data sources that can be incomplete, unevenly distributed, or influenced by platform use, device ownership, retail access, or digital participation. The most credible applications used non-traditional data as part of a broader evidence base, turning partial signals into more reliable findings.

4. AI is making more data usable, but validation and interpretability remain central challenges

AI methods appeared across perhaps the majority of examples, especially where raw data was visual, textual, large, or difficult to interpret manually. Researchers used these methods to identify trees, buildings, dumping sites, crop stress, urban expansion, public sentiment, mobility patterns, and behavioral signals from sensors and digital platforms. But the key question is whether the signals extracted through the use of AI are accurate enough, fair enough, and interpretable enough to support real-world decisions. The strongest cases validated outputs against trusted benchmarks, connected findings to operational decisions, or combined automated analysis with domain expertise and human oversight.

5. Responsible reuse depends on institutions, trust, and preparedness 

Many of these datasets are privately held, passively generated, or collected for purposes unrelated to research or policy. Their public value depends on access arrangements, documentation, safeguards, and long-term stewardship. Privacy-preserving aggregation already plays an important role in applications involving mobile network data, wastewater surveillance, platform data, and smart-device ecosystems, but major governance questions remain around access, representation, accountability, and responsible use. These themes closely align with the conclusions of “Nontraditional Data in Pandemic Preparedness and Response: Identifying and Addressing First- and Last-Mile Challenges” (2026), which highlights both “first-mile” barriers related to data access, standardization, and discovery, and “last-mile” barriers related to translating analysis into operational decision-making. The examples in this update similarly suggest that durable public-interest use of non-traditional data depends not only on analytical capacity, but also on institutions that can identify useful data, broker access responsibly, maintain trust, and connect evidence to decisions before urgent needs arise.

Back to the Blog