US Census API Data

One catalog, 548 endpoints, most of the quantitative picture of American business and population. The US Census API is the Census Bureau's programmatic front door: Decennial Census counts for 2000, 2010 and 2020, the American Community Survey from 2005 onward, County Business Patterns and ZIP Code Business Patterns reaching back to 1986, Nonemployer Statistics from 1997, and the Economic Census benchmarking the whole economy every five years across 2002-2022. Each endpoint publishes its own variable dictionary defining labels and concepts, which is why the same query pattern answers across programs.

For the security alarm industry the load-bearing program is County Business Patterns filtered to NAICS 561621 - Security Systems Services (except Locksmiths): establishments, employment and annual payroll by county, state, metro and ZIP, cut by employment size class. Nonemployer Statistics sits beside it counting the sole operators - the one-truck installer with no payroll - and their revenue. Together they draw the supply side of the market with no survey panel, no vendor estimate and no extrapolation in sight. Of the 1,744 datasets Datadory catalogs, this record scores 9/10 against a mean of 7.81. Get a sample of this dataset and see the 561621 cells for your own territories.

Aggregate cells, not company records:

# one row per geography x employment-size class - NAICS 2017 = 561621, Security Systems Services
naics2017_label : Security systems services          name : <geographic area>
empszes_label   : <size class, e.g. 5 to 9 employees>
estab           : <number of establishments in the cell>
emp             : <paid employees, flagged where suppressed>
payann          : <annual payroll, $1,000s>

# the same spine switched to the nonemployer side
sector          : <NAICS economic sector>             lfo : <legal form of organization>
nestab          : <nonemployer establishments>
nrcptot         : <sales, value of shipments, or revenue, $1,000s>

Three things to notice. First, the grain is deliberate: one row per geography-by-size-class cell keeps a full national sweep small enough to hold in memory while staying sharp enough to separate one suburb from the next. Second, the size-class column is the segmentation - it is what lets a market model distinguish the micro-installer from the regional firm without any proprietary panel. Third, there is no firm underneath: these are aggregates by design, which is precisely what makes the coverage total.

The rows above show record shape; a requested sample pins every value to live records.

Four jobs, one row: industry identity, geographic identity, the employer measures, and the nonemployer measures. The core columns every security-alarm analysis touches are defined in the table above - the NAICS filter, the geographic identifiers, establishment counts, employment, payroll, the size classes, and the nonemployer establishment and revenue pair.

Additional fields on request. The wider catalog runs deeper than the business-count core: Economic Census subject-series measures at the five-year benchmarks, ZIP-level columns where the vintage publishes them separately, ACS and Decennial population and housing variables, and the suppression-flag conventions that tell you why an employment cell came back blank. Each folds into your delivery once named, so the dictionary you build on is the dictionary you tested.

Geography - the full United States, nested: national roll-ups, states, counties, places and CBSA metros, with ZIP Code Business Patterns extending the same spine to the neighborhood grain. One frame serves a corridor plan and a country plan.

Temporal - depth varies by program and that is the point of having all of them: County Business Patterns annually back to 1986, Nonemployer Statistics from 1997, the Economic Census every five years across 2002-2022, ACS from 2005, Decennial Census at 2000, 2010 and 2020. Long-run structural work and current-vintage comparisons draw from one catalog.

Granularity - aggregated counts by NAICS x geography x employment-size class, with nonemployer revenue by NAICS x geography. No firm-level records anywhere, by statute as much as by design: Title 13 forbids using the outputs to identify individual persons or businesses, which is why the aggregates can be total rather than sampled.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel, the cadence and the scope; the field dictionary above travels unchanged through all three. Rows arrive flattened to one observation per geography-size-class cell, types resolved against the program's own documentation, NAICS labels attached alongside codes so a 561621 filter reads in plain English in your BI layer. Suppressed cells ship with their flags rather than silently dropped, so a missing employment figure stays a known unknown instead of a zero poisoning a sum.

Cadence changes are a settings conversation, not a re-integration project, and a sample scoped to your geographies and vintages comes first either way.

• Market researchers and consultants run the baseline sizing: establishments and payroll on the 561621 line, segmented by size class, become the supply-side skeleton of every residential and commercial security market deck.
• Investors and quant researchers read four decades of counts for structure: whether the installer base is consolidating, where monitoring density concentrates, and what roll-up targets look like in the cells.
• Sales and growth teams rank territories by density of security-systems businesses rather than generic business counts, then set recruiting quotas against cells that actually exist.
• Competitive intelligence teams overlay competitor footprints on the establishment map and read white space as geography instead of anecdote.
• Data scientists and ML engineers treat GEO_ID as the entity spine joining business counts to permits, crime baselines and household demographics inside one schema.

Market researchers get the cleanest denominator in the pack: a census, not a sample, so the top-of-funnel number survives scrutiny. Investors get forty years of structural memory in one consistent schema. Sales and growth teams get territory math grounded in where the trade actually operates. Competitive intelligence gets the context layer that says whether thin rival coverage reflects opportunity or absence. The common thread: everyone here needs the shape of the whole industry rather than any single company's story - exactly the question this catalog was built to answer.

It is the denominator the rest of the shelf argues with. The FBI Crime Data Explorer supplies demand-side pressure - burglaries and thefts by jurisdiction; this record supplies the firms positioned to answer it. The National Crime Victimization Survey counts the crimes nobody reported; this record counts who would have answered the phone. SBA Small Business Profiles gives the state-level small-business context, while IRS Statistics of Income puts tax-based profit numbers on the sole-proprietor slice this record counts structurally. Head-to-head treatments live at the SEC EDGAR Full-Text Search vs US Census API comparison.

Three things, all knowable upfront. First, these are aggregates: if your model needs firm-level records, say so now, because the program publishes none - though the size-class and nonemployer splits recover much of what analysts go looking for. Second, temporal depth differs by program, and suppressed cells appear wherever a geography is dominated by one employer; name your vintages and the sample tells you exactly where the blanks fall and how they were flagged. Third, variable names move between vintages - the exact column names for your target years get pinned against that vintage's own documentation during sampling, and that confirmation travels with your delivery. A scoped sample settles all three before anything is built on top.