A polygon is a closed shape on a map, stored as an ordered list of coordinates, used by advertising systems to decide whether a device counts as being somewhere. It is the alternative to a circle. Radius targeting takes one point and a distance; a polygon traces the outline of the thing itself: a supermarket footprint, a stadium concourse, a postcode cell, a forecourt without the road running past it. The shape exists because places are not round, and because systems that treat them as round bid on people in the car park next door and count them as visitors.

What the shape is made of

A polygon is built from linear rings. The GeoJSON format, published by the Internet Engineering Task Force as RFC 7946 in August 2016, defines a linear ring as a closed line of at least four positions in which the first and last are identical. The first ring bounds the surface. Further rings cut holes inside it, which is how an atrium, a courtyard or an excluded loading bay gets represented. RFC 7946 fixed two things its 2008 predecessor left open: coordinates must use World Geodetic System 1984 in decimal degrees, with the option to declare an alternative reference system removed, and rings must follow the right-hand rule, exterior rings running counterclockwise and holes clockwise. Because the older specification was silent on winding order, parsers are told not to reject shapes that get it wrong.

Older encodings remain in circulation. Esri shapefiles and well-known text are accepted as geofence import formats by mapping platforms, and location vendors take spreadsheet uploads too. Foursquare supports custom polygon uploads in both its Proximity Designer and Audience Designer tools, accepting CSV, TSV and Excel files, capping point uploads at 5 MB or roughly 100,000 rows, and rendering only the first 1,000 rows on the map while targeting against the full set.

How a device is matched to a shape

Nothing in the bid request carries the polygon. OpenRTB's Geo object transmits a point and its provenance: latitude, longitude, a type code indicating whether the fix came from device location services, an IP lookup or the user, estimated accuracy in metres, and lastfix, the number of seconds since the fix was taken. The specification warns that coordinates should match the accuracy the type claims, and that the centroid of a postal code should not be passed as though it were a device position.

Boundary evaluation therefore happens on the buy side, against shapes the platform already holds. The underlying operation is a point-in-polygon test, classically solved by casting a ray from the coordinate and counting edge crossings: an odd number means inside. Running that geometry across tens of thousands of shapes inside a bid timeout is not viable, so systems precompute. Bounding boxes discard obvious misses, and grid indexes carry the rest. H3, the hexagonal grid Uber open sourced in 2018, encodes any cell as a 64-bit integer across sixteen resolutions, so a polygon becomes a covering set of cells and membership becomes a lookup. The cost is precision at the edges, since a hexagon approximates the outline it fills.

Where polygons sit in the buying stack

Support is uneven, and the largest platform offers none. Display & Video 360 exposes two location list types, regional and proximity, each holding up to 10,000 entries: postcodes in one, street addresses or coordinate pairs in the other. Proximity lists are circles, and Google sets a minimum radius of 1 kilometre that must enclose a population of at least 1,000, a privacy floor rather than a technical one. The radius cannot be changed after selection, and reporting rolls up to the list rather than the individual site.

Custom shapes are usually bought from a specialist. According to documentation from the buying platform Choozle, radius geofences executed in The Trade Desk carry no incremental cost per thousand impressions, while polygon targeting routed through Foursquare adds roughly 1.50 dollars on top of media: a price on the difference between a circle and an outline.

On the measurement side the shape is the product. Adsquare, the Berlin location intelligence firm, maps audience signals onto two-dimensional polygons of store locations, a method its DACH general manager David Luchtenburg described when the company extended store visit tracking to Austria and Switzerland on January 30, 2025. StackAdapt and Adsquare had already put custom radius and polygon setup inside a buying interface on September 10, 2024, across EMEA, Asia Pacific, North America and Latin America.

Planning systems have started shipping polygons as inventory currency. On September 8, 2026, microm and Azerion pushed Germany's PLZ-8 grid into the Hawk demand-side platform, carving the country into roughly 82,000 cells averaging about 500 households each and turning survey-defined target groups into map shapes buyable across digital out-of-home, connected television and audio without cookies or device identifiers.

Origin and evolution

The vocabulary was fixed by measurement bodies rather than by buying platforms. The Media Rating Council issued its Location-Based Advertising Measurement Guidelines on March 29, 2017, developed with the Interactive Advertising Bureau and the Mobile Marketing Association after a public comment draft the previous November. Those guidelines split the geo-fence into two named methods: point-radius, a circle of stated radius around a coordinate, and boundary measurement, in which a complex polygon represents the confines of the place itself. Foursquare Visits became the first location product accredited under those guidelines in August 2020, an assurance the Media Rating Council grants by independent audit rather than by measuring anything itself.

A 2018 PPC Land guide to location marketing described geofencing as targeting a custom area on a map when demand-side platforms and data providers enforced minimum radii of around 50 metres and the technique worked only in application environments. What changed after 2020 was not the geometry but the law around it.

Why the outline matters

Boundary choice is a budget decision disguised as a drawing exercise. A generous radius around a shopping centre entrance captures passers-by, while a tight polygon around a single storefront can miss people who parked and walked in from another side. Both errors flow into the same reported number, and neither is visible in a dashboard showing only attributed visits.

There is a regulatory asymmetry too. Statutes define precise geolocation by radius: the SECURE Data Act unveiled in April 2026 uses 1,750 feet, and Vermont's Act 138 uses 1,850 feet. A polygon traced around a single unit encloses an area orders of magnitude smaller than either circle, so the compliance test is written in a shape the discipline abandoned for its most precise work.

Limitations and disputes

The MRC guidelines concede the structural flaw: geo-fencing generally cannot represent multi-storey buildings. A polygon is two-dimensional, so a shopping centre, an office tower or a supermarket above a car park collapses into one outline covering several unrelated businesses. Altitude is recommended as a validation input, and rarely supplied reliably.

Precision also outruns the signal. The same guidelines place GPS accuracy at roughly 10 to 100 metres, degrading indoors, and Wi-Fi at 30 to 100 metres. An outline drawn to the wall of a shop is finer than the measurement being tested against it, which means false positives at the boundary are arithmetically guaranteed rather than occasional. Network-derived positioning compounds it: identifiers such as the BSSID of a fixed router now appear by name in enforcement orders as sources of precise location.

Legal exposure has grown fastest. Washington's My Health My Data Act, effective March 31, 2024, bans geofencing within 2,000 feet of an in-person care facility and carries a private right of action. The Federal Trade Commission closed its four-year case against Kochava with a stipulated order signed on May 4, 2026, barring sales of sensitive location data without affirmative express consent. Accuracy claims now carry an evidentiary duty: Cox Media Group and two other vendors accepted 20 years of substantiation obligations covering representations about geographic targeting. The United Kingdom's Information Commissioner's Office advised in 2026 that region or city level targeting is acceptable without consent while finer signals are not.

Not the same as

A point-radius geofence is a circle: one coordinate and one distance. Cheaper, faster to evaluate, supported everywhere, and what most platforms mean by geofencing. A polygon is the boundary method the MRC defined against it.

A geo region is an administrative unit already in a platform's taxonomy, such as a city, state or postcode. Google Business Profile service areas now work this way, since radius targeting has been retired in favour of named places.

An isochrone encloses everywhere reachable within a travel time. The output is a polygon, but the input is a road network, not a building footprint.

An H3 or geohash cell is a tile from a fixed global grid. Grids approximate arbitrary shapes and are used to index them, not to define them.

Polygon is also the Ethereum scaling network used in advertising's Web3 phase, including the NFL's commemorative Super Bowl LVI ticket tokens, and a unit of 3D geometry, where polygon reduction trims model complexity in interactive product creative. Neither has anything to do with location targeting.

Recent developments

Boundary-based measurement kept moving into the largest buying platforms through 2026. The Trade Desk plugged Adsquare into Audience Unlimited on June 17, 2026, and Amazon DSP listed Adsquare segments across six markets five weeks later. Azira and Adobe opened cross-border footfall attribution inside Adobe Advertising DSP on August 25, 2026, live first in Asia Pacific. On September 10, 2026, Drako began bundling foot traffic attribution with its audiences at no charge, against a market rate it put at about 50 cents per thousand impressions.

Timeline

  • August 2016: RFC 7946 standardises GeoJSON, mandating WGS 84 coordinates and right-hand winding for polygon rings
  • November 2016: The MRC issues its draft Location-Based Advertising Measurement Guidelines for public comment
  • March 29, 2017: The final guidelines define point-radius and boundary geo-fencing
  • 2018: Uber open sources H3, a hexagonal grid index for approximating arbitrary shapes
  • August 2020: Foursquare Visits receives the first MRC accreditation for location data
  • September 10, 2024: StackAdapt and Adsquare place custom radius and polygon setup inside a demand-side platform
  • March 31, 2024: Washington's My Health My Data Act takes effect, banning geofencing within 2,000 feet of care facilities
  • January 30, 2025: Adsquare extends polygon-based store visit tracking to Austria and Switzerland
  • April 2026: The SECURE Data Act defines precise geolocation at a radius of 1,750 feet
  • May 4, 2026: The FTC signs a stipulated order closing its Kochava location data case
  • June 17, 2026: The Trade Desk integrates Adsquare into Audience Unlimited
  • August 25, 2026: Azira and Adobe launch cross-border footfall attribution in Asia Pacific
  • September 8, 2026: microm and Azerion route Germany's 82,000 PLZ-8 cells into the Hawk DSP as polygons

Summary

Who. Location intelligence vendors including Adsquare, Foursquare, Azira and GroundTruth draw and license the shapes; demand-side platforms evaluate device coordinates against them; the Media Rating Council defines and audits the methods; the Federal Trade Commission and state legislatures set the limits on the data involved.

What. A polygon is a closed boundary expressed as ordered coordinates, used to define a targeting area or a measurement place. It is the boundary alternative to point-radius geofencing, encoded in formats such as GeoJSON and evaluated by point-in-polygon tests, usually accelerated by grid indexes.

When. The geometry predates advertising, the interchange format was standardised in August 2016, the measurement vocabulary was fixed in March 2017, and commercial deployment accelerated from 2024 as location measurement moved inside major buying platforms.

Where. In location list and geofence configuration on the buy side, in the visit matching layer of location intelligence vendors, and in planning grids such as Germany's PLZ-8 cells now buyable as map shapes.

Why. Circles buy and count the wrong people at the edges of a place, and outlines reduce that error at a cost in price, processing and regulatory exposure. Precision is also the property regulators target, which makes the shape a compliance question as much as a targeting one.