Geomarketing organises marketing decisions around places rather than people. Customer records, sales figures, survey panels and census data are attached to defined map units, the units are ranked against one another, and the ranking drives where a shop opens, where a leaflet drops, how a sales territory is cut and which areas receive advertising. The unit of analysis is the small area rather than the individual. A neighbourhood of 500 households can be described when the households inside it cannot, and that asymmetry is the reason the discipline exists.

The German business reference Gabler Wirtschaftslexikon defines it as a marketing instrument built on geographic information systems (GIS), locating internal company data in space and relating it to external market data such as sociodemographic structure to create a basis for business decisions. The word travels unevenly: standard usage in German, French, Spanish and Portuguese-speaking markets, while a Wageningen University thesis observed that the term is not used in English. English-speaking practitioners use location intelligence, geodemographics or business geography for the same work.

How the analysis is built

Four steps run in sequence. The first is geocoding: turning addresses or postcodes in a company's own records into points a mapping system can place.

The second is the spatial unit, which determines everything downstream. Official geographies are free but uneven: a five-digit German postcode can cover a village or a slice of central Munich. Commercial vendors therefore build their own. microm, based in Neuss and part of the Creditreform group, maintains a level called PLZ-8 below the postal code Deutsche Post administers. Azerion puts the count at roughly 82,000 cells averaging about 500 households each; a 2012 academic description cited approximately 85,000, alongside 1.5 million street sections averaging 27 households. The design goal is homogeneity rather than equal area, so an index computed in a dense Berlin block means roughly what it means across several rural streets.

The third step fills the cells, because survey research cannot be matched to individuals at scale. The German joint industry study best for planning surveyed 30,390 people in its 2025 edition to represent 71.00 million German-speaking residents aged 14 and over. None of those respondents is the person behind any given ad request. Locating the target group in space, then buying the space, is the standard bridge.

The fourth step is spatial modelling, resting on work published long before computers ran it. William J. Reilly set out the law of retail gravitation in 1931, calculating a breaking point between two centres from their populations and the distance between them. David Huff replaced that hard boundary with a probability in 1963, substituting retail floorspace for population and travel time for distance.

Activation is where the analysis meets ad tech. Geography arrives in the Geo object of an OpenRTB bid request, OpenRTB being the real-time bidding specification set by the Interactive Advertising Bureau's Tech Lab. The object carries lat, lon, type, accuracy, lastfix, ipservice, country, region, metro, city, zip and utcoffset. The IAB Tech Lab specification instructs that coordinates should only be passed when they match the accuracy implied by the type field, and that a postal code centroid should not be sent as a coordinate pair. Exchanges coarsen further. Google's Authorized Buyers documentation describes supplying only a geolocation shared by a sufficiently large number of users, limiting lat and lon to 0.01 degrees, and returning a city centre point with accuracy set to the radius of an equal-area circle.

On the buying interfaces the controls are blunter. Google Ads offers presence targeting and a broader presence-or-interest default reaching people who have shown interest in a place without being in it; its documentation reports internal 2022 test data showing 5% more conversions in travel, real estate and education campaigns after the switch. Google Ad Manager added postal code targeting for Spain and Italy on June 16, 2025, joining a framework already covering Germany, France and the United Kingdom. Amazon released a geographic optimization API for DSP advertisers sorting markets into performance bands at postal code granularity, with smart locations objects applying bid adjustments at that level.

Origin and evolution

The computing layer arrived first. The Canada Geographic Information System, led by Roger Tomlinson from 1963, is generally treated as the first true GIS; commercial packages reached personal computers in the 1980s.

Segmentation arrived alongside it. Claritas introduced PRIZM, the Potential Rating Index for ZIP Markets, in 1974, linking census data to the then-new American ZIP codes. In Britain, Richard Webber built a neighbourhood classification at the government-funded Centre for Environmental Studies in the 1970s to identify deprivation in Liverpool. He joined CACI in 1979, where it was rebranded ACORN, and later developed Mosaic with CCN, now Experian. The first British Mosaic worked at postcode level, roughly 20 properties per unit.

The word itself belongs to continental Europe in the early 1990s, when desktop mapping made cartographic processing cheap enough for commercial use; microm has operated in the field since around 1992. Through the 2000s the digital branch grew separately as IP-based geotargeting in search and display. The 2010s added mobile software development kits reporting satellite coordinates, producing both proximity targeting and the footfall panels measuring it. The current phase is a reversal: as cookies and device identifiers degraded, aggregated geography became one of the few signals resolving in every channel, and planning-side geography moved into buying platforms.

Why it matters for the marketing community

The practical argument is cross-channel consistency. An outdoor screen has no cookie and no user, a connected television device carries a household identifier at best, and a mobile application may or may not surrender an advertising identifier. Geography resolves, to some approximation, everywhere.

That logic is now being productised. microm and Azerion integrated the PLZ-8 grid into the Hawk demand-side platformin an announcement dated September 8, 2026, letting target groups from the German market media study enter as polygons and run across outdoor, television, audio, mobile and desktop without cookies or user identifiers. The same platform had already taken Global Data Resources data across 13 European markets in January 2026. Germany carries roughly 111,000 digital outdoor screens according to an August 2025 study, placing more than one screen in the average cell.

Measurement followed the same path. Footfall attribution moved inside buying platforms in September 2024, when StackAdapt and Adsquare integrated it into a single interface. Geography also serves as a randomisation unit: Meridian GeoX runs market-level causal experiments to calibrate a mix model, descended from the geo experiments methodology Google researchers published in 2011.

Limitations and disputes

The most serious objection is that the error bar often exceeds the cell. A study published on November 5, 2025 benchmarked nearly a billion address records from six providers and found IP-to-postal accuracy of 13%, with providers agreeing on the same household linkage 6.4% of the time and state-level accuracy of 5% to 17%. FreeWheel used that research in February 2026 to argue that IP-based targeting can miss 87% of households, disclosing that addresses account for 17% of targeting on its own platform. Adstra and InterMedia Advertising reported on July 15, 2026 that 23% of residential addresses reached the intended geographic target. All three sell identity products. Against a 500-household polygon, those margins are not a rounding detail.

The second objection is statistical. Area averages do not describe the households inside them, a trap William Robinson named the ecological fallacy in 1950. A cell indexed highly for a segment still contains households outside it; the index is a propensity, not a fact.

The third is regulatory. Aggregated geography sits outside the consent machinery governing identifier-based targeting, but law now defines precise location by radius. The stipulated order closing the Federal Trade Commission's Kochava case, signed on May 4, 2026, treats data finer than an 1,850 foot radius as precise location. New Jersey's Assembly Bill A5328, signed on June 30, 2026, bans sales of geolocation finer than 1,750 feet with no consent exception and penalties reaching $50,000 per record. Where aggregated ends and precise begins is unsettled, and a small enough polygon can single out a household.

Dependency is the fourth. Proprietary grids are licensed by single owners, and the Azerion announcement described no portability path, no accuracy benchmark and no method for resolving a bid request to a cell at auction time, and its claims of reduced wastage carried no figures.

Not the same as

Geotargeting is the delivery control: the campaign setting including or excluding named places, radii or postal codes. It is an activation route out of the analysis, not the analysis.

Geofencing draws a boundary and treats entry into it as the trigger. It answers where an impression is served, not which areas deserve budget.

Location intelligence is the English-market label for a largely overlapping activity, weighted toward mobile movement data rather than census and survey projection.

Geo experiments use geography as a randomisation unit to establish causality. The market split is a measurement design, not a targeting method.

Recent developments

September 2026 produced four moves in nine days. Stagwell renamed its People Platform location business Numetrix on September 1 with no disclosed product change, and Azira opened natural-language querying over 13 trillion location signals the same day. The microm integration followed on September 8, and Meridian GeoX left beta on September 9 claiming budget savings above 31% for large advertisers.

Screen-level planning is the newer frontier. A8 Media built a system to select individual outdoor screens rather than whole city buys, and Drako began giving away foot traffic reports normally priced near $0.50 per thousand impressions, its chief executive calling the metric a partial signal. Pricing pressure is arriving alongside demand.

Timeline

  • 1931: William J. Reilly publishes The Law of Retail Gravitation
  • 1963: David Huff publishes a probabilistic trade area model in Land Economics; the Canada Geographic Information System is developed under Roger Tomlinson
  • 1974: Claritas introduces PRIZM, linking United States census data to ZIP codes
  • 1979: Richard Webber joins CACI; his neighbourhood classification is rebranded ACORN
  • 1980s: Mosaic is developed with CCN at British postcode level, around 20 properties per unit
  • Early 1990s: the term geomarketing enters commercial use in continental Europe as desktop mapping spreads
  • 1997: CCN becomes Experian
  • 2011: Google researchers publish the geo experiments methodology for market-level testing
  • November 5, 2025: the Coalition for Innovative Media Measurement and Go Addressable publish research putting IP-to-postal accuracy at 13%
  • February 28, 2026: FreeWheel publishes its analysis of address-based targeting limits
  • May 4, 2026: the Federal Trade Commission's Kochava order defines precise location at an 1,850 foot radius
  • June 30, 2026: New Jersey signs Assembly Bill A5328, banning sales of geolocation finer than 1,750 feet
  • July 15, 2026: Adstra and InterMedia report 23% geographic accuracy for residential addresses
  • September 8, 2026: microm and Azerion announce the PLZ-8 integration into Hawk DSP
  • September 9, 2026: Google's Meridian GeoX reaches general availability

Summary

Who: Geomarketing suppliers including microm in Germany, Experian and CACI in Britain, and location intelligence firms such as Adsquare, Azira, Foursquare and Numetrix. Buyers are retailers, multi-location service businesses, media planners and the agency teams translating catchment analysis into media schedules. Platforms including Google Ads, Google Ad Manager, Amazon DSP and Hawk carry the activation layer, and IAB Tech Lab maintains the bid request fields that transport geography.

What: The use of spatial data to make marketing decisions, combining geocoding of internal records, a chosen spatial unit, projected survey or census attributes, and gravity or catchment modelling, with activation through geographic targeting controls and measurement through footfall attribution and market-level experiments.

When: Methodologically dated to Reilly in 1931 and Huff in 1963, computationally to the Canada Geographic Information System from 1963, commercially to PRIZM in 1974 and ACORN in 1979. The term entered continental European business use in the early 1990s. The current revival tracks the degradation of cookies and device identifiers through the 2020s.

Where: Strongest as a named discipline in Germany, France, Spain, Portugal and Latin America; practised under other labels in English-speaking markets. Activation now spans digital out-of-home, connected television, audio, mobile, desktop and retail media.

Why: Aggregated geography resolves in channels where individual identifiers do not, gives planners one key for cross-channel reach arithmetic, and sits outside the consent machinery attached to identifier-based targeting. The counterweight is accuracy: independent and vendor research puts address-derived geographic precision between 13% and 23%, which is coarser than the neighbourhood grids being sold, and area-level indices describe populations rather than the households inside them.