kvlak limn
FCC Broadband Data Collection / Signal diagnostic Snapshot Jul 3, 2026

What they say
vs what you get

The national broadband map is built from what providers say they offer. A block is “served” when a company reports it could connect you - at the fastest tier it advertises, not the speed that actually arrives at night. This is a diagnostic of where that gap runs widest.

Real Top-line federal figures are sourced. Illustrative County rankings and per-provider speeds model documented patterns - marked throughout. See Methodology.

Three optimisms

01 / Why the map says you're covered

Nobody has to lie for the broadband map to overstate reality. The overstatement is built into three definitions - each defensible on its own, compounding into a picture rosier than the one you meet when you actually try to buy service. Every claim in this section is documented.

  1. Fault 01

    “Could serve” counts as served

    A location is listed as covered when a provider says it could begin service within ten business days of a request - even if nobody there subscribes, the build-out quote runs into the thousands, or a line was never actually installed.

    Real The BDC availability standard is provider self-report of where service is or could be made available, not where it is built, sold, and running.

  2. Fault 02

    Advertised speed, not measured speed

    Filings carry each provider's maximum advertised tier. What actually arrives at 8 p.m. on a congested node is a different number entirely - measured by a separate program, never shown on the coverage map.

    Real The FCC's Measuring Broadband America program exists precisely because the availability filings report advertised, not measured, throughput.

  3. Fault 03

    One qualifying spot can flag a whole area

    The retired Form 477 marked an entire census block “served” if a single location in it could get service. The BDC moved to location-level reporting in 2022 - an improvement - but the data is still self-reported and the underlying Location Fabric still misses units.

    Real Form 477's block-level overstatement is the documented reason Congress ordered the more granular Broadband Data Collection.

The gap, once someone measured it
FCC · 2019 reported lacking 25/3
claimed
~21M
Microsoft telemetry not using broadband speeds
measured
~120M

In 2019 the FCC reported ~21 million Americans lacked 25/3 access. Microsoft, reading its own service telemetry, estimated ~120 million-plus were not actually using the internet at broadband speeds - the single most-cited illustration of the availability-versus-reality gap.

The coverage gap

02 / Who gets left off

The share of each community without even the older 25 / 3 Mbps floor. Urban America is effectively wired; the gap opens in rural counties and widens again on Tribal lands, where terrain, jurisdiction, and the thinnest build-out returns compound. Raising the bar to today's 100 / 20 Mbps benchmark pushes every one of these numbers higher.

Without service Widest gap Signal bars = served strength · FCC 2019 Broadband Deployment Report (2017 Form 477 data)
The same figures, in a table
CommunityWithout 25 / 3 Mbps
Urban 1.5%
Rural 22.3%
Tribal lands (rural) 27.7%

Real Shares are the FCC's own reported figures at the 25 / 3 Mbps benchmark (FCC 2019 Broadband Deployment Report (2017 Form 477 data)); the ordering and the size of the Tribal gap are well documented. Because Form 477 counted a block as served on a single qualifying location, even these understate the true gap.

76%
Illustrative of BDC filed locations clear 100 / 20 Mbps (modeled)

Modeled illustrative distribution - no raw BDC export is loaded. Drop bulk availability CSVs into data/raw/ and run `npm run data` to replace these with real filed-location aggregates.

One option, or none

03 / Single-provider country
~83M Monopoly market

Americans with access to only one wired broadband provider - or none - at the current benchmark. Where a single company files, the “advertised price” is simply the price: no competitor to undercut it, and the coverage map still reads as green.

34%
Illustrative of BDC filed locations have one provider or none (modeled)

Modeled illustrative distribution - no raw BDC export is loaded. Drop bulk availability CSVs into data/raw/ and run `npm run data` to replace these with real filed-location aggregates.

Illustrative The entries below are archetypes, not named counties with fabricated provider counts - each dramatizes a documented single-provider pattern. Swap in real BDC location data to name and rank the actual counties (see Methodology).

  1. 01

    A ranching county in the Mountain West

    Interior West / Geostationary satellite

    Terrestrial providers file the county as “served” from a tower miles away. The only line you can actually order is a satellite dish with a 600 ms round-trip.

  2. 02

    A Mississippi Delta parish

    Deep South / One cable incumbent

    A single cable company holds the whole footprint. No competitor files here, so the advertised price is simply whatever the incumbent lists.

  3. 03

    An Appalachian hollow

    Central Appalachia / DSL over aging copper

    The map shows 25 Mbps DSL. On a copper loop this long the line syncs at a fraction of that - but it still counts as covered.

  4. 04

    A high-desert Tribal community

    Southwest / Fixed wireless, line-of-sight

    Service depends on an unobstructed path to a distant tower. Where terrain blocks it, the household is counted served and gets nothing.

  5. 05

    An exurban new-build subdivision

    Sun Belt fringe / One provider, waitlisted

    The developer wired a single provider. It sits on the map at gigabit - with a months-long install backlog.

Advertised vs measured

04 / The gap, by technology

The map counts a line as broadband on the strength of its advertised tier. But the distance between “up to” and “what arrives” depends almost entirely on the wire. Fiber keeps its promise; long copper and geostationary satellite are where the paper number and the lived experience diverge most.

Illustrative Bars encode the documented gap tier for each technology, not a per-provider measurement - the empty stretch is the gap, no fake precise percentage implied. The FCC's Measuring Broadband America program is the real source of measured-vs-advertised throughput.

  1. DSL (copper)

    large gap
    They advertise
    Up to 25 Mbps
    What you actually get
    On long rural copper loops the line syncs well below the “up to” number - frequently an advertised availability the household can never actually reach.

  2. Satellite - geostationary

    large gap
    They advertise
    25 - 100 Mbps
    What you actually get
    The download tier can be met, but a 600 ms+ round-trip and hard monthly caps make it unusable for video calls or gaming. “Available” on paper, marginal in practice.

  3. Fixed wireless

    moderate gap
    They advertise
    Up to 100 Mbps
    What you actually get
    Highly variable - terrain, weather, and distance to the tower decide it. A clear line-of-sight home is fine; a blocked one gets a fraction.

  4. Cable (DOCSIS)

    small gap
    They advertise
    Up to 300 - 1,000 down
    What you actually get
    Download usually lands near the advertised tier off-peak; upload is a small fraction of it, and both dip on congested nodes at night.

  5. Satellite - low-earth-orbit

    small gap
    They advertise
    50 - 200 Mbps
    What you actually get
    Latency drops to 20 - 60 ms and speeds land closer to advertised, but shared capacity means slowdowns where too many subscribe. It reshaped rural options after 2021.

  6. Fiber (FTTP)

    minimal gap
    They advertise
    300 - 1,000 Mbps
    What you actually get
    Delivers at or above the advertised tier, with 5 - 15 ms latency. The gap here is availability, not speed - most homes still can't get it run to the door.
The same table, plainly
TechnologyAdvertisedGap
DSL (copper) Up to 25 Mbps large
Satellite - geostationary 25 - 100 Mbps large
Fixed wireless Up to 100 Mbps moderate
Cable (DOCSIS) Up to 300 - 1,000 down small
Satellite - low-earth-orbit 50 - 200 Mbps small
Fiber (FTTP) 300 - 1,000 Mbps minimal

Methodology

05 / What's real, what's illustrative

This is a v1 curated snapshot, not a live query of the FCC map. The raw Broadband Data Collection is enormous - hundreds of millions of location-by-provider rows plus a separate Location Fabric - and isn't wired into this build yet. So the page hand-assembles a small set of well-documented public figures and clearly labels everything modeled rather than measured. The honest line runs down the middle of the page: Real where a claim is sourced, Illustrative where it dramatizes a documented pattern.

Source

The real feed is the FCC Broadband Data Collection (broadbandmap.fcc.gov), which since 2022 has published provider-filed availability at the location level, plus a public challenge process. Speed reality checks come from the FCC's separate Measuring Broadband America program, and the top-line deployment counts from the FCC's annual Section 706 / Broadband Deployment Reports. src/lib/source.ts is the documented swap-point; scripts/build-data.ts is the real→derived stub; HANDOFF.md has the exact download instructions.

How the FCC defines “availability” - and why it runs optimistic

A location is counted as served when a provider reports it could supply service there, at its maximum advertised speed, typically within ten business days of a request. Three consequences follow, and they are the spine of this page: “could serve” is treated as “served”; advertised speed stands in for measured speed; and the legacy block-level method (Form 477) let a single qualifying address flag a whole census block. Moving to location-level BDC reporting narrowed the third gap but not the first two. Not a conspiracy - just what the definition literally counts.

What is real

  • The current fixed-broadband benchmark of 100 / 20 Mbps, raised in 2024 from the 25 / 3 Mbps floor set in 2015.
  • The FCC's own estimate that roughly 24 million Americans still lack a 100 / 20 Mbps fixed option.
  • The FCC-versus-Microsoft gap: ~21M reported unserved at 25/3 in 2019 against Microsoft's ~120M-plus not actually using broadband speeds.
  • The community shares - Urban, Rural, Tribal - are the FCC's reported figures at the 25 / 3 Mbps benchmark (FCC 2019 Broadband Deployment Report (2017 Form 477 data)). The ordering and the size of the Tribal gap are well established.
  • The oft-cited estimate that ~83M Americans have one wired provider or none (Institute for Local Self-Reliance, 2020 analysis of single-provider markets).

What is illustrative

  • The single-provider archetypes are composites of documented patterns, not named counties with fabricated provider counts.
  • The per-technology advertised-versus-measured bars encode the documented gap tier, not per-provider measurements. Measuring Broadband America supplies the real numbers.
  • The Compare page is a static example pairing at a hypothetical rural address - a preview of the address-lookup tool, not a live query.

What you're not seeing

Price. Installation quotes that make a “served” address unbuildable in practice. Data caps and throttling. The households the Location Fabric misses entirely. Availability answers “could someone sell you a line here,” which is a narrower question than “can you get affordable, usable internet” - and it is the only question the map is built to answer.

Snapshot generated 2026-07-04. v1 curated snapshot, not a live query of the BDC. Top-line federal figures - the 100/20 benchmark, the ~24M-unserved estimate, the FCC-vs-Microsoft gap, the 25/3 community shares - are real and sourced. County rankings, per-provider speeds, and example pairings are illustrative of well-documented patterns and are marked as such. Source attribution: Federal Communications Commission · Broadband Data Collection · Measuring Broadband America · Section 706 reports.