Key Takeaways
- Polygon data represents the true shape and boundaries of a real-world place using connected coordinate pairs, not a single latitude and longitude like point data.
- Polygons are created through manual digitization, photogrammetry, LiDAR, or machine learning extraction, each with different accuracy and cost tradeoffs.
- Polygon data is delivered in formats like Shapefile, GeoJSON, WKT, and increasingly GeoParquet, which became a native Apache Parquet type in 2026.
- Beyond mapping and visit attribution, polygon data powers site selection, insurance risk modeling, urban planning, telecom network design, and real estate analytics.
- Spatial hierarchy and provenance transparency (knowing whether a polygon is machine-generated or hand-drawn) separate reliable polygon datasets from approximate ones.
Every physical place has a shape, and that shape carries information a single coordinate point never can. Polygon data captures the true boundaries of buildings, parcels, and properties, giving analysts, insurers, retailers, and city planners a far more precise foundation than a dot on a map. This guide covers what polygon data actually is, how it gets made, the formats it comes in, and the use cases where precision makes or breaks the analysis.
What Is Polygon Data?
Polygon data is a type of geospatial data that represents the true shape and boundaries of a real-world feature, such as a building, parcel, or park, using a connected series of x and y coordinate pairs rather than a single latitude and longitude.
In vector geography, every feature is represented as a point, a line, or a polygon. Points mark a single location, like a store’s front door. Lines represent linear features, like roads or pipelines. Polygons enclose an area, capturing everything a point cannot: shape, size, and how one place sits relative to the places around it. Esri defines a polygon as a GIS object that stores its geographic representation as a series of coordinate pairs enclosing an area, as one of its properties in a database row. That last part matters: a polygon in geography is never just a shape. It is a shape paired with data.
Polygon = Geometry + Attributes
The boundary coordinates are only half the story. The attributes of a polygon, things like address, category, spatial hierarchy, and parking availability, are what turn a shape on a map into something you can actually analyze. A polygon with no attributes tells you where a building sits. A polygon with attributes tells you what it is, what it contains, and how it relates to everything around it.
Polygon Data vs. Point Data vs. Centroid Radius
Not every use case needs a full polygon. But when precision matters, the difference between a point, a centroid radius, and a true polygon is the difference between a good analysis and a wrong one.
Data Type | Precision | Best For | Common Failure Mode |
Point | Single coordinate, no shape or size | Quick mapping, competitive counts, trade area centers | Cannot distinguish between adjacent or stacked locations |
Centroid radius | Approximate circle around a center point | Rough proximity searches, low-density areas | Over-attributes visits to small stores, under-attributes to large ones; breaks down in dense areas |
Polygon (building footprint) | True shape and boundary of the structure | Visit attribution, risk modeling, site selection, urban planning | Requires accurate, current source data to stay reliable |
There is no “close enough” with polygons. A centroid radius approach tends to produce geocodes that overrepresent small sub-stores and underrepresent large ones, since it is modeling a circle, not a building. A true polygon does not have that problem because it reflects the physical footprint itself.
How Polygon Data Is Created
Polygon data does not exist by default. It has to be built, and how it is built determines how much you can trust it.
Manual digitization
An analyst traces building outlines directly from aerial or satellite imagery. It is slow and expensive at scale, but it remains the most reliable method for complex or unusual structures where automated methods struggle, like irregularly shaped campuses or buildings partially obscured by tree cover.
Photogrammetry and remote sensing
Overlapping aerial or satellite images are stitched together to reconstruct a 3D representation of the ground, from which building outlines are extracted. This scales far better than manual work and is the backbone of most large, wide-coverage footprint datasets.
LiDAR
Airborne LiDAR sensors fire laser pulses at the ground and measure the return time to build a dense 3D point cloud. Because it captures elevation directly, LiDAR is especially good at separating buildings from trees and terrain, and it increasingly feeds directly into machine learning pipelines rather than being processed on its own.
Machine learning extraction
Deep learning models, typically convolutional neural networks trained on labeled imagery, segment building outlines automatically and then regularize them into clean, rectilinear polygons. This is now the dominant method for generating footprints at global scale. A 2026 peer-reviewed benchmark on very-high-resolution satellite imagery reported 97.6% precision and a 94.5% F1-score for automated building footprint extraction, evidence that ML-based methods have closed much of the gap with manual digitization.
How SafeGraph does it
SafeGraph Geometry draws from thousands of sources, including satellite imagery, aerial photography, and municipal records, to generate a first-pass polygon for every POI using machine learning and computer vision. Where automated extraction falls short, particularly for complex structures like malls, campuses, or buildings partially hidden by tree cover, polygons are hand-drawn and verified by human analysts instead.
Every record carries a metadata field flagging which method produced it: machine-generated or hand-drawn. That provenance flag is a differentiator most vendors do not expose. Rather than asking you to trust a black box, it lets you decide, POI by POI, how much scrutiny a given polygon needs before it goes into a high-stakes model like insurance risk or site selection. The same metadata also carries spatial hierarchy: whether a polygon is a standalone footprint or shares its boundary with a parent place, like a Subway sharing its shape with the Walmart that contains it. (See Geometry: The Foundation of SafeGraph Places for the full technical breakdown of how spatial hierarchy is calculated.)
Polygon Stats to Know in 2026
- The global GIS market is valued at roughly $14.56 billion in 2025 and is projected to reach $31.8 billion by 2031, at a 13.94% CAGR (Mordor Intelligence).
- Microsoft’s GlobalMLBuildingFootprints dataset held 1.4 billion building records as of early 2026, with the combined Microsoft-Google Open Buildings dataset spanning over 2.5 billion footprints worldwide.
- Apache Parquet added native GEOMETRY and GEOGRAPHY types in February 2026, and more than 20 tools across seven languages now support GeoParquet.
- Leading ML extraction pipelines are now reporting F1-scores above 94% on high-resolution satellite imagery benchmarks.
Polygon Data Formats and Delivery
Polygon data is only as useful as your ability to load and query it. The most common formats are:
- Shapefile: The long-standing GIS industry default. Widely supported, but limited by field name length and a 2GB file size cap that makes it clumsy for large-scale delivery.
- GeoJSON: A human-readable, web-friendly format built on JSON. Excellent for web mapping and lightweight applications, less efficient for very large polygon databases.
- WKT and WKB (Well-Known Text and Well-Known Binary): Text and binary representations of geometry that plug directly into spatial databases and SQL queries. SafeGraph delivers Geometry data in WKT for exactly this reason: it drops straight into a warehouse without a conversion step.
- GeoParquet: A columnar format built for cloud-scale analytics. GeoParquet only got more central to the ecosystem in 2026, when Apache Parquet added native geometry and geography types, cementing it as the standard for moving large polygon datasets between cloud systems.
Delivery matters as much as format. Flat-file downloads still work for one-time projects, but warehouse-native delivery, querying Geometry data directly in Snowflake or receiving scheduled S3 bulk drops, is now the standard for teams running polygon data in production pipelines. See the full setup on the Integrations page.
Want the full technical breakdown of SafeGraph’s Places and Geometry schema?
Download the Places Data Technical Guide.
7 Use Cases for Polygon Data
1. Mapping and Visualization
This one might seem obvious, but it shouldn’t be overlooked. All geospatial data eventually gets mapped in some form, but mapping with polygons instead of points is a deliberate choice: it signals that you need to understand how properties actually relate to one another, not just where they sit. A single point can tell you a building exists. A polygon tells you its shape, its footprint, and how close it sits to the building next door.
Mapping building footprints also reveals relationships that point data simply cannot show. With spatial hierarchy metadata, polygon data can be mapped to show which places sit inside another place, a food court inside a mall, a clinic inside a hospital campus, and so on. Accurate geometry additionally enables visualization of co-tenancy and adjacency, which feeds directly into site selection and risk assessment work later in this list. Mapping polygons also captures accessibility details a point never could: where an entrance is actually located, or where the nearest parking lot sits relative to the building.
2. Visit Attribution
Mobility data only becomes meaningful once it is matched to a real boundary. Polygon-based visit attribution shows which places people actually enter and how long they stay, while centroid radius models tend to both over- and under-attribute visits, especially for locations stacked close together or inside the same building. Spatial hierarchy resolves this by correctly separating a mall from the dozens of individual stores inside it.
3. Insurance Risk Assessment
Precise co-tenancy and adjacency data is fundamental to property risk modeling. A nail salon that opens next door to a fireworks retailer carries a meaningfully different risk profile than one that opens across the street or beside a grocery store, and only polygon data captures that distinction reliably. In a competitive, data-driven insurance market, underassessing risk increases exposure, while overassessing drives customer churn; polygon precision is what lets insurers avoid both.
4. Site Selection and Co-Tenancy Analysis
Retailers and commercial real estate teams use polygon data to evaluate potential locations before signing a lease. Accurate footprints reveal what else occupies a shopping center, how much frontage and parking a site actually has, and how nearby tenants might drive or cannibalize foot traffic. Co-tenancy analysis, understanding which brands perform well next to which other brands, depends entirely on knowing the real boundaries and adjacency of every location in a trade area.
5. Urban Planning and Infrastructure
City planners use building footprints to model development constraints, plan utility and transportation infrastructure, and estimate how a neighborhood’s built environment is likely to grow. Comparing footprint area against available land helps planners balance development with the public space communities still need, from parks to sidewalks to accessibility infrastructure like curb ramps.
6. Telecom and RF Planning
Telecom providers rely on building footprints to plan network infrastructure as they expand into new coverage areas. Positioning a cell tower requires weighing available land against signal range, and accounting for the footprints of surrounding buildings that could obstruct or interfere with transmission. Precise polygon data helps carriers achieve maximum coverage with the fewest towers, which is increasingly central to how telecom teams justify network capital spend.
7. Real Estate and Property Analytics
Polygon data lets real estate analysts calculate a property’s true footprint and, from there, derive its approximate square footage, a real-world example of a polygon’s area doing genuine analytical work rather than just sitting on a map. Combined with parcel and zoning data, this supports valuation models, portfolio analysis, and due diligence at a scale manual measurement can’t match. (SafeGraph’s guide on estimating POI square footage from polygons walks through the methodology.)
What Makes Polygon Data “Good”? A Buyer’s Checklist
Not all polygon datasets are built the same way, and the differences show up fast once you’re relying on the data in production. Before buying or building a polygon database, check for:
Precision at the building level: The polygon should reflect the actual footprint, not a generalized shape or a buffered approximation.
Spatial hierarchy: Parent-child relationships between polygons (a mall containing individual stores, an airport containing a Starbucks) are what let you tell a standalone building apart from a shared structure.
Freshness and verified status: The physical world changes constantly: stores close, buildings get renovated, new construction appears. A polygon dataset is only as reliable as its update cadence.
Coverage transparency: Know exactly what you’re getting. SafeGraph Geometry currently provides building footprints and spatial hierarchy metadata for POIs across the US, Canada, and the UK. Confirm current coverage against your target markets before you build a project around it.
Provenance transparency: Whether a polygon was machine-generated or hand-drawn changes how much confidence you should place in it for edge cases. Datasets that expose this metadata field, rather than hiding it, are being straight with you about accuracy.
Where to Get Polygon Data
Open sources
OpenStreetMap, Microsoft’s GlobalMLBuildingFootprints, and Google’s Open Buildings dataset (now combined into a single dataset spanning over 2.5 billion footprints) are free and genuinely useful starting points.
Their limitations are consistent: attribute depth is thin, update cadence varies by region, and spatial hierarchy (the parent-child relationships that make polygon data analytically useful) is largely absent.
They’re a solid base layer, not a production-ready dataset on their own.
Commercial datasets
Vendors that specialize in curating, verifying, and maintaining polygon databases are worth the cost when accuracy, freshness, and metadata depth actually matter to your outcomes, which is most B2B use cases beyond basic mapping.
SafeGraph Geometry delivers building footprints in WKT format with spatial hierarchy, parking lot attributes, and ML-versus-hand-drawn provenance flags built in.
Get a free sample of SafeGraph Geometry data to see the schema and coverage before committing to anything.
Closing Thoughts
Points tell you where something is. Polygons tell you what it actually looks like, how it relates to everything around it, and whether you can trust the answer.
As more of the physical world gets mapped by machine learning at global scale, the datasets that win will be the ones that pair that scale with real verification, not just automated guesses.
Whichever source you choose, judge it the same way you’d judge any data you’re about to build decisions on: by its precision, its freshness, and how honest it is about where the data came from.
More About Polygon Data
FAQ’s
1. What is polygon data in geospatial analysis?
Polygon data represents areas using boundaries defined by connected coordinate pairs, capturing the true shape and size of a feature like a building or land parcel, rather than a single point location.
2. Why is polygon data more accurate than point data?
It reflects a location’s actual shape and footprint instead of a single coordinate, which makes it possible to correctly model adjacency, overlap, and spatial relationships between places, especially in dense areas.
3. How is polygon data created?
Through manual digitization, photogrammetry and remote sensing, LiDAR, or machine learning extraction from satellite and aerial imagery. Most large-scale datasets today combine automated extraction with human verification for accuracy.
4. What format does polygon data come in?
Common formats include Shapefile, GeoJSON, WKT/WKB, and GeoParquet. GeoParquet has become the standard for cloud-scale delivery since Apache Parquet added native geometry support in 2026.
5. How is polygon data used in visit attribution?
It determines whether a person actually entered a specific building rather than just passing nearby, which centroid radius models routinely get wrong in dense or stacked locations.
6. Why is polygon data important for insurance risk assessment?
It reveals co-tenancy and adjacency, which businesses share a structure or sit next to which others, information that is critical for accurately pricing property risk.
7. What's the difference between a polygon and a geofence?
A polygon typically represents a real, physical boundary like a building footprint. A geofence is a virtual boundary drawn for a specific purpose (like triggering a mobile notification) that may or may not align with a physical structure.
8. Can polygon data be generated automatically?
Yes. Machine learning models can extract building footprints directly from satellite imagery with over 94% accuracy on recent benchmarks, though hand-drawn verification is still used for complex or high-stakes structures.
