Federal Employment Records (Origin-Destination)
U.S. Federal Statistical Program
Workplace-residence flows matched to federal data.
Research & Methodology
How we rebuilt
the population map.
The technical foundations, validation methodology, and data architecture behind Ether Data's spatio-temporal intelligence layer.
Data provenance
Government statistical records, not device signals
Every feature traces back to federal statistical programs. No mobile panels, no SDK data, and no inferred device locations.
Federal Census Records (Community Survey)
U.S. Federal Statistical Program
Demographics, income, education, housing, and commuting priors.
Government Labor Statistics (Quarterly Census)
Federal Labor Statistics Agency
Establishment-level employment counts by industry sector.
Federal Time Use Survey
Federal Labor Statistics Agency
Time-use curves by occupation and sector for hourly presence.
Spatial architecture
Proprietary spatial grid
We use a hexagonal tiling system because hexagons tile without distortion, have uniform neighbor distances, and integrate natively with modern analytics stacks.
Resolution
~460m
Edge length per cell
Manhattan coverage
3,360
Complete borough coverage
Features per cell
1,000+
Across 12 attribute domains
Attribute domains
Core identifiers & totals
Age & sex distribution
Race & ethnicity
Education & schooling
Income, poverty & inequality
Housing stock & tenure
Housing structure types
Household & family structure
Labor force & employment
Commuting & travel behavior
Transportation & vehicles
Mobility, migration & language
Product layers
Three intelligence surfaces
Each layer builds on the previous one: presence, attraction, and hourly composition.
01
Daytime Population Proxy
Employment assignment and commute redistribution onto H3 cells.
Validation:
02
Ether's Gravity Model
Hub scores, commuter vectors, and economic adjacency mapping.
Structural priors:
03
Ether's Temporal Engine - Hourly Composition
Sector-level workforce curves per cell by hour.
Lift above baseline:
Validation
Tested against ground truth
We validate on observed outcomes, not internal consistency alone.
7-day holdout prediction
Structural priors only
NYC temporal validation
492,316 observations
Restaurant revenue prediction
No transaction training data
Daytime population accuracy
Validated on high-density zones
Integration
SQL-first. BigQuery-native.
One join key connects Ether data to your existing spatial tables.
SELECT
cell_id,
daytime_population,
top_sector,
gravity_score,
finance_worker_pct,
peak_hour
FROM ether.population_intelligence
WHERE metro = 'NYC'
AND hour_local = 12
AND finance_worker_pct > 0.15
ORDER BY gravity_score DESC
LIMIT 100;