The GDR Data Model — Methodology

How GDR builds precise, scalable audiences from geography — not from people.

One model. Six data layers. Fused into a single, privacy-native targeting fabric — every layer free of personal identifiers.

GDR builds audience targeting from the ground up at the household and area level, never the individual.

Through IDfree — GDR's privacy-native data product and philosophy — official statistics, environmental intelligence, and consumer research are combined into a single score for every geographical unit in a market.

The result is precise, scalable targeting for display, video, social, DOOH and CTV, without cookies, device IDs, or any personal data.

Because the model reads places rather than people, it is durable against cookie deprecation and signal loss, works in any market on Earth, and stays clean from a compliance standpoint everywhere it runs.

Privacy-Native by Design

Privacy isn't a constraint bolted on after the fact — it is the foundation the model is built on.

No individuals. All targeting resolves to households and geographical units, never to named or identified people.

No children. The model targets households, not individuals, and never children.

No tracking, no PII. No cookies, device IDs, or private data are used at any point in the model.

Probabilistic, not deterministic. Audiences are modeled from authoritative and environmental signals — not assembled from ID-based profiles.

The model provides targeting for households, not individuals, and not children. No tracking or private data is used in the data model.

The Six Data Layers

Each layer contributes an independent view of a place. Combined, they describe who lives in an area, how it is built, and what its residents value — without ever describing a person.

Official Statistics

Government data on income, population, household composition, and housing from national statistical offices. The factual backbone of the model — authoritative, regularly updated, and available at small geographical units across all markets.

PlacePrint

The unique digital fingerprint of a neighborhood, derived from everything that exists there — every building, road, shop, and transit stop — without using any personal data. Proven by Google Research to predict neighborhood income and demographic composition with accuracy competitive with traditional survey-based methods.

Satellite Intelligence

Machine learning applied to satellite imagery — capturing urban density, vegetation, land use, and building patterns from above. Updates continuously rather than on census cycles, and works in any market on Earth regardless of local data availability.

Street Intelligence

Building-level analysis from street-level imagery — facade material, architectural era, building condition, and maintenance quality. Where satellite reads areas from above, Street Intelligence reads buildings from the front, processing each in under half a second.

GDR Consumer Classification

Geodemographic segment profiles built from decades of behavioral and lifestyle research. Translates raw statistics into meaningful consumer typologies — who lives there, how they spend, and what they value.

Survey

Attitudinal and behavioral data from consumer research — media consumption, brand affinity, purchase intent, and lifestyle values. Connects the geographic and physical profile of an area to what residents actually think and buy.

How the Layers Fuse

The six layers are fused at the geographical unit level — never at the individual level. For each unit, the signals are normalised and combined into a single composite score. An audience is then built by defining a target profile and indexing every geographical unit against it, so units are ranked by how strongly they match.

Step 1

Six layers scored for every geographical unit

→

Step 2

Layers normalised & fused into one composite score

→

Step 3

Units indexed against the target audience profile

→

Step 4

Ranked units activated across DSPs & channels

One score per place. The output is a single index value per geographical unit — ready to activate for display, video, social, DOOH and CTV, with no personal data attached.

Full-Population Reach

Everyone is on the map — no one is excluded.

Because every geographical unit in a market is scored, the entire population is segmented. For any given audience, every single unit carries a value — from zero to an open-ended index, with no upper limit. Nobody falls off the addressable map because they lack a cookie, an ID, or a third-party data match.

INDEX 0 — no affinityHIGHER AFFINITY →

This is what makes the model unique. Reach is set by choosing how far down the index to activate — not by how many identifiers happen to exist. GDR can deliver broad scale or tight precision from the same model, and tune the trade-off campaign by campaign.

100% of the population scoredReach dialed by threshold, not ID supplyScalable reach at a fair priceNo tracking, no personal data

Accuracy & Validation

✓Independently validated approach. Google Research has shown that neighborhood-fingerprint methods like PlacePrint predict income and demographic composition with accuracy competitive with traditional survey-based methods.

✓Always current. Satellite and street layers update continuously rather than on census cycles, so the model reflects how areas actually change over time.

✓Universal coverage. Because environmental layers work anywhere on Earth, the model extends to any market regardless of local data availability.

✓Multi-layer corroboration. No single source drives a score; independent layers reinforce one another, reducing the error any one dataset would carry alone.

Questions & Answers

The questions media and data teams ask most often when they first meet the model.

Does the model use any personal data, cookies, or tracking?+

No. No cookies, device IDs, or personal identifiers are used at any point. Every signal resolves to households and geographical units, never to a named or identified person. The model reads places, not people.

How is this different from cookie- or ID-based targeting?+

Traditional targeting assembles profiles from individual identifiers that are increasingly fragile as cookies deprecate and signal is lost. GDR instead models audiences probabilistically from authoritative statistics and environmental intelligence, scored at the area level. There is no ID to lose, so the approach is durable through cookieless and consent-limited environments.

Are children ever targeted or identified?+

No. The model targets households, not individuals, and never children. Even a family-oriented audience such as “households with school-age children” resolves to the home as a whole — no child, and no individual person, is ever targeted, profiled, or identified.

How accurate can targeting be without individual-level data?+

Accuracy comes from corroboration across six independent layers — no single source drives a score, so the layers reinforce one another and reduce the error any one dataset would carry alone. The underlying approach is independently validated: Google Research has shown neighborhood-fingerprint methods like PlacePrint predict income and demographic composition with accuracy competitive with traditional survey-based methods.

How much reach can the model deliver — and why is it cost-effective?+

Because every geographical unit is scored, the entire population is segmented — every unit holds a value for any audience, from zero to an open-ended index. Reach is therefore set by where you draw the index threshold, not by how many identifiers happen to be available for purchase. That means broad scale or tight precision from the same model, and reach delivered at a fair price without buying or matching personal data.

How granular is the targeting?+

Down to small geographical units — neighborhood and grid-cell level — rather than broad regions. Every unit in a market receives its own composite score, so audiences can be built and ranked at fine resolution. To protect privacy, GDR never segments geographical areas smaller than 110 m × 110 m (120 yd × 120 yd), and never with fewer than 15 households.

Which channels and platforms can I activate on?+

The output is a single index per geographical unit, ready to activate across display, video, social, DOOH and CTV through standard DSP and platform workflows — with no personal data attached to the audience.

Which markets can you cover?+

Any market on Earth. Because the environmental layers — satellite and street intelligence — work anywhere regardless of local data availability, the model extends globally and is not limited to markets with rich third-party data.

How fresh is the data?+

Satellite and street layers update continuously rather than on census cycles, so the model reflects how areas actually change over time. Official statistics refresh on their national release schedules, and survey and classification inputs are maintained on an ongoing basis.

Is the approach privacy-compliant?+

The model is privacy-native by design — it never ingests or produces personal identifiers, so there is no personal data to consent, store, or expose. This makes it well suited to privacy-first regulation across markets. (GDR can share market-specific compliance detail on request.)

Fair Pricing

Pricing aligned with value

Because GDR owns the IP and the underlying algorithms, we are not bound to expensive third-party licensing or intermediary markups. That means we can offer our data at a very fair price — one where the cost of data and the effect it delivers stay closely aligned.

✓IP owned in-house. No external licensing fees passed on to clients.

✓Transparent model. Pricing reflects actual data and processing costs, not arbitrary premiums.

✓Aligned with your outcomes. Cost scales with the value and reach you activate, not with data scarcity.