How to read a mobile coverage map without being misled

The maps operators publish are usually generated rather than surveyed. Arcep, the French telecoms regulator, which publishes operators' maps, describes them as theoretical: produced by modelling that accounts for the location of masts, their transmission power and the surrounding geography. The mobile layer of the FCC's National Broadband Map in the United States is built the same way, from propagation models that operators run with certain common settings for consistency.

A model can be good and still be wrong about the place you care about. The FCC is explicit that, because its map is based on propagation modelling, real experience varies with the device you use, the capacity of the cell site and the terrain. It adds that the maps on operators' own websites may rest on different parameters and assumptions, including service provided through roaming agreements, and so may not match its own.

A legend compresses a probability into a flat block of colour, and several conditions hide behind it. The FCC's mobile coverage describes where a user should be able to connect outdoors or in a moving vehicle and reach certain upload and download speeds; it does not cover indoor service at all. Arcep's public site grades coverage in four tiers, from very good coverage to no coverage, which is more candid than a single shade and still a statement about typical conditions rather than a promise for one address.

For eSIM buyers, the most important condition is whose network the colour describes. A local operator's map shows its own network. A travel plan that roams onto that network uses the same masts under a different agreement, and the map was never drawn to describe that agreement. Before trusting a colour, find out:

Some regulators do not take the maps on trust. Arcep checks operators' maps through regular field audits and scores each one by the share of its tests that succeed in areas the operator reports as covered. When it raised the reliability it requires from 95 per cent to 98 per cent in 2020, operators changed the way they produced their maps, and in some places the coverage they showed became smaller than in earlier editions.

Arcep's own example shows one operator's 4G coverage shrinking in some areas between two quarters, presumably to meet the stricter requirement, while new deployments added coverage elsewhere. That is the useful lesson. A colour reflects the assumptions of whoever drew it, and the same network can look more or less generous depending on how cautious the drawing is required to be.

Crowd-sourced measurements answer a different question: not where signal should be, but where real phones actually connected. Germany's Bundesnetzagentur runs a dead spot checker app whose users record, with their own devices and their own mobile contracts, whether they had any coverage and which technology was available. Its map covering 2023 and 2024 drew on some 250 million test points from around 300,000 users.

Measured data has limits of its own, and the regulator says so: its map gives no conclusive information about a particular operator's locally available technology, and it is only as detailed as the number of tests submitted in each region. Others use public measurements as a check on the models. People who think the FCC's mobile coverage is wrong can take speed tests with its app, and the results are aggregated into challenges that operators must respond to; Arcep, too, invites testing apps, including crowd-sourced ones, to add their findings to its public site.

A review is a third kind of evidence, and the narrowest: one traveller, one phone, one plan, one set of places. Read it against the map rather than instead of it. If a reviewer reports weak service somewhere an operator's map shows solid colour, check whether they were indoors, on a different technology, or on a plan that roams onto that network, because each of those is something the map was never describing.

The order that works is simple. Use the mobile coverage map to learn which networks should reach your route, measured tests to see whether they did for real devices, and reviews to learn how a particular plan behaved on those networks. When all three agree, you have something close to evidence. When they disagree, the disagreement is the most useful thing on the page.