If you’ve ever pulled up a map on your phone, checked a flood zone before buying a house, or wondered how a delivery app always seems to know exactly where you are, you’ve relied on geospatial data – probably without giving it a second thought.
Overture Maps CTO Amy Rose sat down with Grizz Griswold, the host of FINOS Open Source in Finance Podcast (FINOS is the FinTech Open Source Foundation under the Linux Foundation which aims to accelerate collaboration and innovation in financial services through the adoption of open source software). In this latest episode, Amy talked about just how much complexity lives behind the maps we use every day, and how Overture helps its members trust and validate open map data through open standards and high-quality metadata.
Open data doesn’t have software’s playbook yet
Open source software has decades of established practice behind it: version control, maintainer roles, code as a single source of truth. Open data doesn’t have that same maturity. Geospatial data adds even more complexity on top.
Map data is usually built from many different input sources, shaped by judgment calls, and inherently opinionated. That raises a hard question: how do you make a dataset’s lineage and provenance transparent enough that people can trust it, especially when most users can’t just “read the code” the way they could with software?
This is the problem GERS (Global Entity Reference System) and Overture’s validation pipelines exist to solve. GERS gives every map feature a persistent ID, so a building, road, or place can be tracked and referenced across sources and over time. The validation pipeline merges multi-source data into one predictable schema. Combine that with documented lineage and machine-readable formats like GeoParquet, and developers get a ground truth they can actually audit, not just a dataset they’re asked to take on faith.
Why metadata matters more in an agentic AI world
Metadata – data about data – usually gets treated as an afterthought. And when it does exist, it’s typically captured at the dataset level: one license, one lineage note, one accuracy statement, applied to the whole thing.
That’s a problem for agentic AI workflows, which often combine data at the record level rather than working with entire datasets. If the metadata only lives at the dataset level, it doesn’t travel with the individual records that get pulled out and recombined. And record-level metadata needs to answer real questions:
- How was this piece of data created? Was it surveyed on the ground, or extracted from imagery via machine learning?
- What’s the licensing? This matters a lot, since combining open licenses can create friction or outright incompatibility.
- What coordinate reference system was used to map it to the real world?
Our goal at Overture is to capture this metadata from the start and carry it through every processing step, every new attribute, every transformation, so that by the time a release goes out, the metadata is already built in. This isn’t a nice-to-have anymore. It’s a requirement.
Skip this step, and downstream systems can’t verify where their data came from, what licensing constraints apply, or whether the spatial reference is even correct. In an environment where records get dynamically isolated and recombined, stripping out that context turns structured data into a black box and leaves developers holding the risk of licensing conflicts and spatial misalignment they had no way to see coming.
The stakes are real
This isn’t just about clean pipelines and tidy licensing. In disaster response and preparedness, buildings get damaged or destroyed, and the data needs to be refreshed quickly and faithfully to reflect what’s actually happening on the ground. Fast, reliable, well-documented location data isn’t a convenience in those moments. It’s what lets people bring the right information together when it counts.
Where geospatial data is headed
Amy sees a clear trend: geospatial data is moving from a niche “add-on” to something mainstream. For years, spatial data needed special extensions – like GeoParquet layered on top of standard Parquet – just to be usable in common formats. That’s changing.
The bigger shift, in her view, is that location data is something every organization deals with, not just the ones that explicitly work in “maps.” As AI and automation reshape how data gets used, the goal is for spatial data to be built natively into AI systems, instead of bolted on as an afterthought, which is how it’s often treated today.
Core map data as shared infrastructure
Trusting open data, and understanding its context down to the individual record, is expensive and labor-intensive to do alone. That’s the whole reason Overture exists.
We think high-quality map data should work like electricity or the internet: a non-differentiating utility that everyone needs and nobody should have to rebuild from scratch. By treating core map layers like roads, buildings, places as shared infrastructure, Overture members lower their total cost of ownership and get a reliable, neutral foundation to build location intelligence on top of.
This data isn’t finished, and it isn’t perfect. It’s a foundation, not the whole building. That’s exactly why we need more people building on it and telling us where it breaks.
Come build with us
- Explore Overture’s vision for open spatial data and location grounding.
- Join Overture and help shape the knowledge layer — visit the Overture website to learn how your organization can become a member.
- Dig into the reference data and system through our developer documentation.
- For the latest schema releases and AI tooling, sign up for the monthly newsletter or follow Overture on LinkedIn, X, and Bluesky.
