How Accurate Are ISP Coverage Maps?
An ISP coverage map is a sales asset that happens to be shaped like a dataset. It is not fabricated, but it is built from optimistic inputs, and every ambiguity in it resolves in the provider's favour. Knowing exactly which assumptions are baked in tells you how much to discount.
What a coverage map is actually showing
There are three distinct things people call a coverage map, and they fail differently.
- Wired footprint maps shade the areas where a provider believes its plant passes homes. The underlying question is "does our network reach this location?", answered by matching addresses against an internal serviceability database.
- Mobile coverage maps are the output of radio propagation models: tower locations, antenna heights, transmit power, spectrum band, and a terrain model, run through software that predicts signal strength. Almost none of it is drive-tested at your address.
- Regulatory filings feed the FCC National Broadband Map, using the standard that a provider may report a location if it does serve it or could within ten business days without extraordinary construction cost.
Where the errors come from
| Source of error | What it does | Direction |
|---|---|---|
| "Up to" speeds | Reports the top tier sold, not the typical result | Overstates |
| Serviceability databases | Address matching errors put homes inside or outside footprints incorrectly | Both |
| Construction charges | "Available" homes turn out to need a paid line extension | Overstates |
| Node and sector congestion | Evening slowdowns are never modelled | Overstates |
| Propagation modelling | Predicts outdoor signal; walls, foliage and terrain detail are approximated | Overstates |
| Capacity limits | Sign-ups closed in a sector still show as covered | Overstates |
| Update lag | New builds missing; retired plant still shown | Both |
The consistent theme is that coverage maps answer "is service possible here?" while customers are asking "what will I get here?". Those questions diverge most exactly where it matters most: at the edge of a footprint, in rural areas, and in congested urban nodes.
Coverage is not capacity. This is the most common misreading of a mobile map. A strong-signal shading means the radio reaches you. It says nothing about how many other people are sharing that sector at 8pm, which is what actually determines your speed.
Measured data is not perfect either
Honesty cuts both ways. SignalScout is built on Ookla Open Data — real speed tests run by real people, aggregated into tiles about 610 metres across, 1.63 million fixed tiles and 424,000 mobile tiles in the 2026 Q1 release. That is a genuinely different kind of evidence, but it carries its own biases:
- Self-selection. People run speed tests for a reason, usually because something feels slow or because they just upgraded. Both ends of that are over-represented.
- The test measures the whole path. A large share of tests run over Wi-Fi, sometimes on old devices, so results reflect the household setup as much as the line. This pushes medians below true line capability.
- Plan mix, not plan availability. A tile median tells you what neighbours bought, not the best product sold on the street.
- Sample size. Dense urban tiles carry hundreds of tests; rural tiles may carry a handful. Always check the test count.
- Quarterly cadence. A build completed after the data vintage will not appear yet.
- No provider attribution. SignalScout does not identify which ISP produced which result, and does not currently list providers at all.
Two biases pointing opposite ways
Coverage maps overstate. Measured medians understate. That is genuinely useful, because it brackets reality. If a provider claims 1 Gbps and the measured tile median is 600 Mbps, those are consistent — a fast network with mixed plan take-up. If the provider claims 1 Gbps and the measured median is 22 Mbps with a 3 Mbps upload, the claim is doing something other than describing your street.
A practical triangulation routine
- Measured first. Read the tile: download, upload, latency, test count, plus the neighbouring tiles and the mobile numbers.
- Regulatory second. The FCC National Broadband Map for provider names and filed technologies. The filed-versus-measured explainer covers how to read it sceptically.
- Provider third. Address checkers on each candidate ISP site. Compare your address with a neighbour's to find build boundaries.
- Human fourth. A local forum or a neighbour. This is the highest-quality data source available and it is free.
- Commit last. Place an order and demand an install date and a written quote for any construction. Availability that cannot survive an order was never availability.
What good scepticism looks like
Trust a coverage map on the question "who might sell me service here?" and distrust it on "how fast will it be?". Trust measured data on "what is realistically happening on this street?" and distrust it on "what is the best plan I could buy?". Where they agree, you can relax. Where they disagree by an order of magnitude, that disagreement is the finding — and it is usually the coverage map that needs to explain itself. If you are weighing specific properties, the pre-move checklist turns this into a short routine you can run before signing anything.
Common questions
Are mobile coverage maps based on real measurements?
Mostly no. Carrier coverage maps are generated by propagation models using tower locations, antenna configuration, spectrum band and terrain data, which predict outdoor signal strength rather than measure indoor experience. Measured speed-test data is a useful counterweight because it reflects what devices actually achieved.
Why is measured speed lower than the plan I pay for?
Speed tests measure the whole path, including Wi-Fi, the device, and the home network, so results routinely come in below the line's capability. Aggregated tile medians also mix households on cheaper plans with those on premium tiers, which pulls the median down relative to the fastest product sold on the street.
If both sources are biased, which should I trust?
Use them together, because their biases run in opposite directions. Coverage and availability data overstate what is possible, measured data understates what is achievable, and the real answer for an address usually sits between the two.
How big an area does one measured tile cover?
Ookla aggregates results into tiles of roughly 610 metres per side at the equator, shrinking with latitude. That is small enough to distinguish neighbourhoods and often individual streets, but large enough that a tile can blend several technologies together.
Check a real address
See the measured download, upload and latency for any US address, plus a Connectivity Score ranked against every county in the country.