By Mallory Sheff, Director, Financial Inclusion, DataKind
Connecting existing data can help social impact organizations make better decisions without collecting more information. Drawing on a Learning Circle supported by the Mastercard Center for Inclusive Growth, DataKind shares four lessons for making data more useful – and how these insights informed the development of FinPulse, a new product focused on community financial health.
For many social impact organizations, the data challenge isn’t simply collecting more information. It’s connecting the data they already have – across internal systems, frontline teams, and public sources – to generate insights that no single source can provide.
When these sources remain disconnected, organizations can struggle to see the bigger picture, understand the communities they serve, and make informed decisions. Closing that gap requires more than technical solutions. It starts with understanding what organizations need to learn and how connected data can help them get there.
This spring, with support from the Mastercard Center for Inclusive Growth, DataKind convened a Learning Circle on Data Integration for Social Impact. DataKind’s Learning Circles bring together peer organizations to work through a shared challenge. In this cohort, four organizations across India, Colombia, Mexico, and Canada, explored what it takes to connect fragmented datasets and turn them into more useful insights.
The organizations differed in geography, mission, and data maturity, but the underlying challenges were similar: training records were disconnected from performance data; program engagement data could not easily be joined to impact surveys; proprietary research sat apart from the public benchmarks that could help interpret it; and qualitative insights captured important dimensions of people’s experiences, but were not connected to quantitative data that could help organizations understand broader patterns.
The challenge wasn’t simply a lack of technical capacity. Better tools and technical resources can help, but organizations also need to answer more fundamental questions: What are we trying to understand? Which data matters? What can – and should – be connected? And what decisions could those connections help us make?
Four lessons from this work stood out.
1. Start with the decision, not the data
Data integration is often treated as a technical exercise: align schemas, resolve identifiers, move information between systems, and build pipelines. Those tasks matter, but starting there can lead organizations to invest considerable effort in connecting data without being clear about what the connection is supposed to accomplish.
We found that organizations made more progress when they could articulate a specific question they wanted their data to answer. That question created a filter for everything that followed: which datasets were relevant, which inconsistencies actually needed to be resolved, and which technical improvements could wait.
As one participant reflected, “Earlier, I saw data integration mainly as a technical task of joining datasets, but now I understand it as a strategic process focused on defining the right questions, ensuring data reliability, and enabling decision-making.”
A useful starting point is therefore not “Which datasets can we integrate?” but “What do we need to understand or decide that we cannot understand today?” The distinction may seem small, but it changes integration from an infrastructure project into a strategic capability.
2. Integration doesn’t have to mean centralization
For organizations with fragmented data, it is tempting to imagine the end state as a single, comprehensive system: a new data warehouse, a unified platform, or one database containing everything the organization collects. Sometimes, that investment is warranted, but often, it’s not the place to begin.
Integration exists on a spectrum. At one end, an organization might connect two datasets once to answer a specific question. Further along, teams might establish shared definitions or identifiers that allow information to be linked consistently. Only at the other end might an organization need automated pipelines and centralized infrastructure.
For many organizations, the most useful first step is relatively modest: choose two relevant data sources, establish a reliable way to connect them, and use the resulting information to answer the question you had set out to explore. A successful connection creates value quickly while also revealing what would be required to integrate data more systematically in the long run.
3. Many technical barriers are governance questions in disguise
Even a relatively simple integration can highlight issues that have little to do with technology. Who owns a particular dataset? Do two teams mean the same thing when they use the same term? Are identifiers consistent across programs? What information can appropriately be linked, and under what conditions?
Across the organizations we worked with, siloed systems, inconsistent identifiers, and mismatched definitions surfaced as obstacles. These can look like technical problems when teams encounter them during implementation, but better software alone rarely resolves them.
Instead, integration forces organizations to make explicit decisions about ownership, definitions, standards, privacy, and process. Those conversations can be difficult, but they are also valuable: they create the shared understanding required for data to become useful across teams rather than remaining an asset of individual programs or departments.
This is why governance should not be treated as something that comes after the technical design. Addressing these questions early can prevent them from resurfacing later as expensive technical blockers.
4. Publicly available data isn’t the same as usable data
Organizations don’t have to generate every piece of information they need themselves. Census data, household surveys, administrative records, financial and climate data, and other public sources can provide valuable context for understanding the communities organizations serve.
While in principle, the amount of information available has never been greater, in practice, using it remains surprisingly difficult. Finding the right source is only the beginning. Organizations must understand what the data actually measures, assess its quality and relevance, reconcile differences in geography and time, transform it into a usable format, and determine how it can meaningfully connect to their own information. If the connection needs to remain current, someone also needs to maintain it as public datasets change.
“Publicly available,” in other words, doesn’t necessarily mean “ready to use.”
This creates an important but often overlooked capacity gap: organizations can be surrounded by potentially valuable data while still lacking access to the context it could provide. The constraint isn’t necessarily data scarcity; it is the time, infrastructure, and specialized expertise required to turn available data into usable information.
From collecting more data to making existing data more useful
Taken together, these lessons suggest a broader shift in how the social sector approaches data. For many organizations, becoming more data-informed may not require another survey, dashboard, or database. It may start with making better use of information that already exists, both within the organization and beyond it.
The goal isn’t to connect everything simply because it can be connected. It’s to build the connections that help organizations answer important questions, understand their work in context, and make better decisions.
At DataKind, this insight increasingly informs how we approach product development. When organizations face recurring data challenges, there’s an opportunity to build reusable infrastructure that makes valuable information easier to access and use, rather than requiring each organization to solve the same problems independently.
DataKind will soon launch FinPulse, a new product focused on financial health. FinPulse brings together public data and indicators that can help organizations better understand the financial health of the communities they serve, and interpret their own knowledge in a broader context. By making this information easier to access and explore, FinPulse aims to reduce the effort required for organizations to navigate fragmented data sources independently.
Making data more useful requires more than collecting information. It requires the right connections, the context to interpret what that data reveals, and the ability to turn those insights into action.
If your organization is exploring how to make better use of its data, we’d love to hear from you at partners@datakind.org.
Hero image courtesy of iStock/ArtHead.
