IdiomaMarket Atlas works only with public, documented data sources (with optional paid enrichments) and surfaces every estimate alongside its margin of error and confidence flag. This page is the full reference — written so a buyer, a planner, or a grant reviewer can read it end-to-end.
ACS variable registry
The Census tables IdiomaMarket Atlas uses.
The atlas defaults to the most recent published ACS 5-year Detailed Tables and stores every variable's estimate and margin of error. Supportive tables (age, household type, foreign-born) are available when a project category needs them; none of them carry individual-level data.
Table
Subject
Use in the atlas
B16001
Language spoken at home by ability to speak English
Universe: population 5 years and over. Distinguishes Spanish-at-home and Spanish speakers who speak English less than very well.
B03001
Hispanic or Latino origin by specific origin
Provides the origin segments the atlas uses for Mexican, Salvadoran, Cuban, Dominican, Puerto Rican, Guatemalan, Honduran, Colombian, Venezuelan, and other communities.
B19013
Median household income
Used as the household income proxy where aggregate income is missing or unreliable.
B19025
Aggregate household income
Used as the primary income-pool input for buying-power estimates when available.
B01003
Total population
Denominator for share calculations and reliability checks.
B11001
Households
Denominator for household-based metrics and buying-power estimates.
B25003
Tenure
Owner / renter segmentation used in category-specific scoring where relevant.
Section 01
Data sources
Every layer in the atlas is traceable to a public, documented source. Paid enrichments are optional and labeled.
ACS 5-Year Detailed Tables (B16001 language, B03001 Hispanic origin, B19013 / B19025 income, B01003 population, B11001 households, B25003 tenure, plus supportive age and household tables).
TIGERweb block-group geometries from the U.S. Census Bureau (centroids, areas, and polygons).
Google Places (Nearby + Text Search) for high-quality competitor and amenity POIs, with OpenStreetMap / Overpass as a fallback.
Optional: BEA regional income, BLS Consumer Expenditure category coefficients, IRS SOI ZIP-code AGI, customer-uploaded store and competitor datasets, and licensed paid enrichments (SafeGraph, Data Axle, Placer.ai, Claritas, Esri Business Analyst) where the client provides the license.
Section 02
Block-group analysis
Block groups are the finest geography ACS Detailed Tables publish, and they're roughly 3× more granular than tracts.
ACS estimates and margins of error are pulled per variable per block group for the configured ACS year.
Counts and rates are computed with explicit denominators (population, population 5+, households) drawn from ACS B01003, B16001's universe, and B11001 respectively.
An origin index is calculated as a cell's origin share divided by the project-area average share, multiplied by 100 — so a cell with index 200 has twice the project-area share.
Block-group reliability is flagged using coefficient-of-variation thresholds derived from the published margins of error.
Section 03
H3 allocation and smoothing
Block-group polygons don't tile cleanly, and they vary wildly in size. H3 fixes both — it's a uniform hexagonal grid that aggregates predictably across scales.
Default resolution is H3 r8 for metro analysis. Resolutions r7 and r9 are available for regional and dense-urban views respectively.
Block-group estimates are allocated to intersecting H3 cells using area-weighted allocation: each cell receives the source block group's count times the share of the block group's area that the cell covers.
Optional smoothing applies a k-ring weighted average (center 1.0, ring-1 0.5, ring-2 0.25) to produce a continuous opportunity surface. Smoothed layers are clearly labeled as smoothed, not raw Census estimate.
Finer H3 resolution does not invent finer source data. The atlas displays smoothing labels and confidence flags so users don't mistake interpolation for additional precision.
Section 04
Buying-power proxy
Buying-power values are modeled estimates, not observed sales — and the atlas labels them that way everywhere they appear.
Income pool: estimated_segment_income_pool = aggregate_household_income × target_segment_share (or households × median_household_income × target_segment_share where aggregate income is unavailable).
Category spend: estimated_category_spend = estimated_segment_income_pool × category_spend_coefficient, with coefficients derived from BLS Consumer Expenditure Survey averages and stored admin-editable.
Coefficients exist for grocery, restaurants, telecom, legal services, healthcare, education, remittance-related proxy, home services, and a general retail bucket.
Provider interface: PublicProxyBuyingPowerProvider is the MVP default. PaidProviderBuyingPowerProvider is the swap-in point for SafeGraph, Placer.ai, Claritas, or Esri BAO license-holders.
Section 05
Competitor density
Competitor presence is fetched per project from Google Places (primary) or OpenStreetMap (fallback), then aggregated to H3 cells.
Google Places Nearby and Text Search drive primary coverage. Searches are tiled across the project geography and deduplicated by Place ID, with per-plan budget caps.
OpenStreetMap / Overpass tags (e.g., shop=supermarket, office=lawyer, amenity=place_of_worship) provide fallback coverage where Google rate limits or coverage matters.
Team-tier users can upload competitor CSVs (with name, address or lat/lng, and category). Addresses are geocoded subject to per-plan limits.
Density per H3 cell feeds the whitespace score, with category-specific interpretation curves (saturation for CPG, validation-plus-whitespace for legal, gap-plus-presence for ministry).
Section 06
Opportunity scoring
The composite opportunity score is transparent and tunable. Every cell ships with an explanation of how it scored.
Components, each normalized 0–100: origin / segment concentration, segment scale, language need, buying-power proxy, competitor whitespace, category fit, and a confidence adjustment.
Default weight presets vary by buyer type (CPG, media, immigration law, religious, nonprofit). Weights are displayed on each solution page and tunable inside a project.
Confidence adjustment multiplies the final score: 1.0 high, 0.9 medium, 0.75 low, 0.6 directional only.
Per-cell explanations describe why it scored where it did and what the user should do next — surfaced in the UI and in the report PDFs.
Section 07
Reliability, margins of error, and limitations
Fine spatial resolution improves usability, but ACS estimates remain survey-based and include uncertainty.
Every demographic estimate is paired with its published margin of error. Coefficient of variation drives the cell-level confidence flag.
Low-population cells, sparse origin groups, and high-MOE estimates surface explicit warnings, and the atlas suppresses or hides values that would mislead.
Smoothed surfaces are visually distinguished and labeled so they are not mistaken for raw Census estimates.
Buying-power values are modeled, not observed. The atlas never claims ACS measures brand preference or affinity.
Section 08
Ethical use
The atlas is an aggregate demographic-analysis tool. It is not, and should not be used as, a system for identifying individuals.
All data is aggregate. The atlas does not expose individual-level records and is not designed to target individuals.
Origin and language are separate questions. Both are used because they support different planning decisions; neither is treated as a proxy for the other.
The atlas does not infer immigration status. It does not estimate undocumented populations. It does not predict legal needs at the individual level.
Reports and exports always carry methodology and limitations language so downstream readers understand what the data does — and does not — claim.
Out of scope
What IdiomaMarket Atlas does not do.
Buyers should be confident about what they can — and cannot — claim from the data. These are the boundaries the product holds everywhere it generates output.
Does not expose individual-level data of any kind.
Does not infer immigration status, undocumented populations, or legal need at the individual level.
Does not claim ACS data proves brand preference or affinity.
Does not predict guaranteed sales — buying-power values are modeled estimates only.
Does not treat “Hispanic” as one homogeneous market.
Does not hide high margins of error — they are surfaced in flags and reports.
Bring this to the call
Want a sample H3 surface for your metro?
Request a demo and we'll come ready with a sample opportunity surface, a ranked target-list preview, and the buyer-specific scoring model for your use case.
We choose the stack around the job: browser maps for decision surfaces, spatial databases for reliable data, Python and GIS libraries for analysis, and annotation tools when imagery needs reviewer-ready labels.