Data products

Connectivity Intelligence: broadband availability for every U.S. location.

Broadband availability, provider competition, and BEAD funding status for every U.S. census block and H3 cell: provider-reported filings transformed into a governed, versioned, quality-checked dataset and delivered where your analysts and applications already work. Connectivity Intelligence is the data product line of GoldenTech KSA. The Snowflake Marketplace listing is live with a free one-state sample, and national coverage is being published release by release.

The attribute set

What one row tells you about a block or a cell

Every row answers the questions that come up when someone asks "what broadband is available here?", and carries the provenance to show where the answer came from.

BEAD status

Served, underserved, and unserved counts and percentages, so you can see whether an area is eligible for BEAD funding and how much of it qualifies.

Speed

The fastest advertised download and upload, and the median best reliable offer, so you know both the headline and the typical service.

Technology mix

The share of locations reachable by fiber, cable, DSL, licensed fixed wireless, and satellite: what is physically available.

Competition

Average number of providers at 25/3, 100/20, and 1000/100 Mbps, the share of locations with a real choice, and the dominant provider.

Keys and provenance

The geography ID, the FCC release and revision, and the BEAD rules version on every row, so two releases can be told apart before their numbers are compared.

Context note

One deterministic, plain-language sentence generated from the row, which a person or an AI agent can quote as it stands.

Example context note from the sample dataset: "1 of 2 locations served (1 underserved, 0 unserved); fastest advertised 1200/1000 Mbps; median best reliable 750/22 Mbps; fiber 50.0%, cable 100.0%, DSL 50.0%, licensed fixed wireless 0.0%, satellite 50.0%; 50.0% of locations have 2+ reliable 100/20 providers; BEAD rules 2026.09, FCC data as of 2025-12-31"
How it is built

From raw provider filings to a governed connectivity product

Four stages, each checkpointed, timed, and recorded in the release registry.

Source

FCC Broadband Data Collection fixed availability files: provider-reported availability for every serviceable location, in the public domain, published for two as-of dates a year plus corrections.

Classify

Versioned BEAD rules mark each location served, underserved, or unserved from reliable low-latency offers at 100/20 and 25/3 Mbps. One rule set, one version stamp on every row.

Aggregate

One row per 2020 census block and per H3 resolution-8 cell, with the same columns in both tables, plus metadata, a quality report, and lineage beside the data.

Deliver

Snowflake Marketplace tables with column comments, bulk Parquet files, a REST API, and a read-only MCP server so an AI agent can call the same data.

Only three public keys leave the FCC files: the FCC location ID, the census block GEOID, and the H3 index. No address, no coordinate, and no CostQuest Fabric data is ingested, stored, or redistributed at any stage. Block and H3 geographies carry the same availability signal and join cleanly to any geocoded record.
Quality and governance

Every release ships its own evidence

A quality gate runs on the real output of every build. A blocking failure stops the release before export, and the result ships with the data as a quality report table, so a reviewer or a buyer can read it without asking us.

  • 31 structural checks on every build. Unique keys, valid block and H3 IDs, percentages in range, status counts that sum, medians under maxima, release and rules stamps, a documented schema, and a Fabric-free column allow-list.
  • Reconciliation within 1% of the FCC's own totals. National figures are checked against the FCC provider summaries for the same as-of date, with per-state deviations reported beside them.
  • Statistical warnings. A move of more than 5 points in a state's served share since the previous release, or a state that disappears, raises a warning for review.
  • Immutable, revisioned releases. Releases are labelled by FCC as-of date and never rebuilt; an FCC correction becomes a new revision such as 2025-12-31.2.
  • A queryable release registry. FCC file IDs, SHA-256 checksums of every shipped file, the pipeline commit, timestamps, and the quality summary, loaded as a RELEASES table.
Illustration of a network of connected locations
Kept current

Current with every FCC publication and correction

The FCC publishes two as-of dates a year, June 30 and December 31, typically four to six months later, then re-publishes each as challenge corrections land. We track every change.

Weekly detection

Every Monday the FCC file listing is fingerprinted and compared with the release registry. A change opens a tracking issue.

10 business days

A new FCC as-of date is built, checked, reviewed by a person, and published within 10 business days.

5 business days

An FCC correction is published as a new revision within 5 business days. Publishing freezes the label forever.

"Newest label for this as-of date" and "what changed between revisions" are queries on the RELEASES table, not questions you have to ask us.

Delivery

One product, four surfaces your team already uses

The same descriptor, quality report, and lineage record ship through every surface, so what you evaluate is what you deploy.

Snowflake Marketplace

Block and H3 tables with column comments, plus metadata, quality, and release tables. The listing is live with a free one-state sample for evaluation. A Team tier and an Enterprise tier with redistribution rights cover the national data as each release is published, and the RELEASES table shows exactly what is available today.

Bulk Parquet

The same tables as Parquet files for Databricks, BigQuery, or your own warehouse, with the data dictionary and quality report alongside.

REST API

Query by block, cell, or point, with an OpenAPI document generated from the same handlers that serve the MCP tools.

MCP server for AI agents

Seven read-only tools: describe_dataset, list_releases, get_block_metrics, get_h3_metrics, h3_for_point, bead_status_summary, and get_quality_report. Every description states grain, release semantics, rules version, and limitations. Coordinates are converted to an H3 index in memory and never stored.

Joins to the records you already hold. Any address your geocoder resolves to a census block or a coordinate can be matched to its connectivity context by block GEOID or H3 index, so the dataset works as an attribute set on customer, property, or asset records.
Who uses it

Who asks "what broadband is available here", and why

Internet service providers

Overbuild analysis and market entry: where competitors are, where fiber is absent, and where 100/20 is not yet met.

State broadband offices and BEAD consultants

Eligibility maps and grant applications need served, underserved, and unserved counts per geography, with the rule set that produced them.

Real estate and proptech

A listing attribute: fastest available speed, fiber yes or no, and provider choice for the location a buyer is looking at.

Utilities and infrastructure planners

Fiber build planning against existing coverage, competition, and public funding eligibility.

Data platforms and enrichment vendors

An attribute set keyed to geographies you already resolve, with the metadata, quality, and lineage records a partner review expects.

Analysts and AI agents

A plain-language context sentence on every row, and MCP tools designed for a model to call safely, with caveats it should repeat.

Licensing

Ready for a flow-down review

  • Every shipped field derives from U.S. Government public-domain data: the FCC Broadband Data Collection fixed availability files.
  • A policy block ships with each release: license of the derived data, redistribution terms per tier, source licenses, owner, update service levels, and retention.
  • Non-exclusive terms. Marketplace tiers are purchased through Snowflake; bulk files, API access, enterprise redistribution, and custom cuts are licensed directly with us.
  • Questions about the schema or a specific use are answered by the published data dictionary and quality report before you buy. Read our data product FAQ or ask us.

Evaluate Connectivity Intelligence on your own records

Ask for the free one-state sample, a walkthrough of the quality report, or a licensing conversation for your use case. We reply within one business day.