01 / overview

A fingerprint for how a place behaves — built from signals no one else has

Population Dynamics Insights (PDI) turns Google's own real-world signals — actual search interest, actual Maps activity, actual environmental conditions — into a single fingerprint for every corner of the country. Two places with matching fingerprints behave alike, even in a market you've never set foot in. This isn't a panel, a survey, or a modeled proxy standing in for reality — it's Google's first-party signal, at a scale no other data provider can touch.

See what's inside vs. Places Insights

in one line

What it unlocks: find every location in the country that behaves like your best one — in a single query, in seconds.
Where it comes from: Google's own search and Maps signal, fused by a foundation model most competitors simply can't replicate.

About PDI data

population_dynamics.v1_s2_z12314K+ rows

geo_idSTRING

featuresFLOAT64[330]

administrative_area_level_1_idSTRING

administrative_area_level_2_idSTRING

···· one embedding, no named columns

0signals fused per fingerprint

0places fingerprinted nationwide

0states, incl. DC & PR

0real-world signal types, fused into one

02 / what's inside

Three real-world signal types, fused into one number

No competitor combines what people are actually searching for, how a place actually gets used, and the environment people actually experience there — into a single, trainable signal. Google does, and documents exactly how: PDI's 330 dimensions resolve into three named blocks, so you can see what's driving a result, not just trust a black box.

0–127 128–255 256–329

Aggregated Search Trends — what people nearby are actually searching for: real interest, real intent, real concerns.

Maps & Busyness — the real energy of a place: what's built there, and how much human activity actually happens around it.

Weather & Air Quality — the real environmental conditions that shape how people behave, day to day.

Because the blocks are documented, you can isolate exactly what's driving a match — Google calls this feature ablation. Ask "how much of this is search interest versus weather?" and get a real answer, not a guess.

About PDI data

built for power

one fingerprint per location, ready for instant, nationwide similarity matching
readable at the block level — isolate what's really driving a result

kept private

no raw search queries or device-level mobility traces
no individual-level data of any kind

03 / data & schema

Lives in BigQuery — join it to your data in minutes

Every one of the 314,000+ fingerprints sits one join away from your own tables, in the same BigQuery project as Places Insights. There's no pipeline to build and no vendor to integrate — the hardest part of "know your market's twin" is already solved, and it's ready the moment you point a query at it.

key facts

geo_id is an S2 level-12 cell token — join your own customer or store data on it directly, no crosswalk table needed
state and county resolve to Google Place IDs — clean, unambiguous joins with no name-matching guesswork
coverage is national and concentrated where people actually live — dense where your customers are, exactly where it matters most

Core schema reference PDI overview docs

Real-world signalssearch, Maps, weather — captured first-hand by Google

Foundation modellearns one embedding per S2 cell

Population Dynamics Insights (BigQuery)pre-GA, S2 level 12

Joins your BigQuery dataon geo_id — standard SQL, no ETL

BigQuery MLVertex AIYour GCP data

example query

Watch it find a market's twin, live

This is the query that makes PDI different from anything else available: one line of SQL compares a location's real fingerprint against every other place in the country and ranks them by how alike they truly are. No panel data to license, no data science team to spin up, no months-long model to build — the pattern-matching engine is already built, at a scale nobody else can offer. Click run on either card to see it in action.

similarity analysis

rank every cell by likeness to a winning store

SELECT
  t.geo_id,
  t.administrative_area_level_2_name AS county,
  (1 - ML.DISTANCE(t.features, target.features, 'COSINE')) AS similarity
FROM population_dynamics.v1_s2_z12 AS t,
  (SELECT features FROM population_dynamics.v1_s2_z12
   WHERE geo_id = '8841a0d') AS target
WHERE t.administrative_area_level_1_name != 'Tennessee'
ORDER BY similarity DESC
LIMIT 5

Query complete · 6.2s elapsed · 5 rows

county	metro	similarity
Jefferson County, KY	Louisville	ranked #1
Jefferson County, AL	Birmingham	ranked #2
Wake County, NC	Raleigh	ranked #3
Hamilton County, OH	Cincinnati	ranked #4

block ablation

same query, search-trend dims only

WITH sliced AS (
  SELECT geo_id, administrative_area_level_2_name AS county,
    features[0:127] AS search_block
  FROM population_dynamics.v1_s2_z12
)
SELECT s.county,
  (1 - ML.DISTANCE(s.search_block, t.search_block, 'COSINE')) AS similarity
FROM sliced AS s, (SELECT search_block FROM sliced
  WHERE geo_id = '8841a0d') AS t
ORDER BY similarity DESC LIMIT 5

Query complete · 5.8s elapsed · 5 rows

county	slice	similarity
Kenton County, KY	search-only	ranked #1
Durham County, NC	search-only	ranked #2
Milwaukee County, WI	search-only	ranked #3
Madison County, AL	search-only	ranked #4

Use the PDI embeddings

04 / analysis modes

What PDI makes possible

A fingerprint for every location changes what's possible with the data you already have. Thin data stops being a dead end — a handful of known locations is enough to reason about hundreds you've never touched. Select any pattern below to see how.

A related pattern, clustering, groups areas into named archetypes — the answer to "what's the vibe of this area," like a nightlife corridor, a commuter suburb, or a retiree community.

Use the PDI embeddings

1

Find your next winning locationInstantly surface every place in the country that behaves like your best one.

This is the flagship PDI move — the query above in action. Give it one location you know performs well, and it ranks every other place in the country by how closely its real-world profile matches. No guesswork, no gut feel.

2

Turn sparse data into a complete pictureHave real numbers for a handful of locations and nothing for the rest? PDI fills in the gaps.

Every location in the country already has a fingerprint, whether or not you have data on it. That means a handful of known locations is enough to estimate the rest with confidence — using the real-world behavior of comparable places, not a blind average.

3

Expand into new territory with confidenceCarry what you've learned in one region straight into one you've never entered.

Train on the markets you already know, then use PDI's fingerprint to project those learnings onto an entirely new region — before you've opened a single location there.

4

Sharpen the pictureTurn broad, county-level data into finer, hyper-local detail.

A coarse statistic — like a single county-level figure — gets refined down toward the neighborhood level using PDI's fine-grained fingerprint, revealing local detail the original data never had.

5

Add a time dimensionCombine PDI with your own historical data to project trends forward.

Paired with your data over time, PDI's fingerprint of today becomes a forward-looking view of where things are headed next.

05 / use cases

What customers do with it

Four questions come up again and again — and PDI answers every one of them from a single fingerprint, without ever exposing a raw number from inside the vector itself.

Find my twin regionsWhat markets behave like my successful locations, so I can replicate what's working — nationwide, not just nearby?

Score a new siteWhat performance should I expect at a candidate location I haven't opened yet?

Name the vibe of an areaIs this a nightlife corridor, a commuter suburb, a retiree community — for tailoring marketing that actually lands.

Boost an existing modelAdd the fingerprint as extra features to a sales or demand model you already trust — and watch accuracy climb.

Case studies span retail demand planning, public-health forecasting, and disaster response. Retail is where PDI shines brightest — it captures real behavioral signal that demographic data alone simply can't see.

industries

Retail & QSR Real estate Financial services Insurance Public health Government & planning Franchising Hospitality & travel Logistics & delivery Telecom CPG & consumer goods Automotive Media & entertainment

06 / privacy and explainability

All the power of real behavioral data, none of the exposure

PDI gets its power from signals no other provider can access — and delivers that power without ever exposing the people behind it. Raw search queries and mobility traces are never in the data at all. To turn a match into a story you can bring to a stakeholder, pair it with Places Insights to describe what's actually there, or an LLM to narrate the pattern in plain language.

privacy-safe by design

no raw search queries or mobility traces — ever
no individual-level or device-level data anywhere in it

bring it to life

pair a PDI match with Places Insights to show what's there
or with an LLM to narrate the pattern in plain language

About PDI data

330-dim fingerprint Privacy-safe encodingprotects the underlying data Similarity score

Similarity score Add the storypair with Places Insights or an LLM A clear, shareable insight

07 / vs. Places Insights

Two different questions, two different answers

Places Insights describes the built environment — what's actually there. PDI describes population behavior — how people actually act around it. Neither one replaces the other; together, they see more of the picture than either could alone.

	Population Dynamics Insights	Places Insights
What it describes	Population behavior	The built environment
Shape of the data	330-dim embedding per cell	Named attributes per place
Grain	S2 cell (~4–5 km²)	Individual place / POI
Human-readable	No — vector only	Yes — rating, hours, category, price
Best for	"Find twin regions," ML features	"What's near this site," competitive density
Aggregation threshold	None observed	Yes — results under 5 places omitted
Release stage	Preview / pre-GA	Generally available

The pairing is the point: PDI finds where to look, Places Insights explains what's there once you've found it.

08 / coverage and access

National coverage today, backed by Google's reach

PDI already covers all 50 states plus DC and Puerto Rico, with the richest detail concentrated exactly where your customers are. It's available now through Google as an early-access offering, delivered the same straightforward way as the rest of the Maps Platform — through BigQuery sharing, on infrastructure already trusted at global scale.

coverage

all 50 states, DC, and Puerto Rico represented
richest detail where your customers are, growing over time

access

available today through early access, with onboarding via Google's Maps Platform team
same delivery model as Places Insights — no separate infrastructure to stand up

Already built to grow beyond the US — additional countries ship on the identical schema, so international coverage is a configuration change, not a rebuild.

PDI overview docs

coverage today

50 states Washington, D.C. Puerto Rico

on the roadmap

Additional countries Deeper rural coverage