Every mobile network publishes a map of where it works, and almost nobody reads one correctly. The colours look like a photograph of reality — this area has signal, that one does not — and they are nothing of the kind. A coverage map is the output of a calculation, run against a set of assumptions that are rarely printed next to the legend, and those assumptions decide what the colours are allowed to mean.
That matters most to a visitor, because a visitor is the person least able to check. You cannot walk the valley before you book the house in it. This article is about what a coverage map is actually claiming, where its claim quietly stops, and the two or three questions that turn a picture into a decision you can act on.
Key takeaways
- A map is a prediction — computed from transmitter data and terrain, not driven and recorded.
- It is drawn outdoors, at about head height — so it says nothing about your room, the metro or the underpass.
- Covered means a minimum, and the minimum is far below what a video call needs.
- It is one network's map, and a visiting phone is not always on that network.
A Coverage Map Is a Prediction, Not a Photograph
Nobody drives every road. A coverage map is produced by feeding a propagation model the things the operator knows — where its masts are, how much power each one radiates, how the antennas are tilted, what the terrain does between there and here — and asking it to estimate the signal at every point on a grid. The estimates are then cut into bands and painted.
So the map is a model plus a list of assumptions, and the assumptions are the interesting part. Where a regulator requires operators to publish maps, it usually has to write those assumptions down, which is the only reason outsiders can see them at all. The United States rule for broadband coverage filings, 47 CFR §1.7004, requires maps drawn with a "cell edge coverage probability of not less than 90%" and a "cell loading of not less than 50%".
Read that twice. Even the definition of "covered" has a probability inside it, and a guess about how busy the network is.
A coverage map answers one narrow question: would a test phone, standing outdoors, probably work here?
What the Colours Are Actually Claiming
The legend words vary — excellent, good, fair, variable — and none of them are standardised. Each band is a threshold on a predicted signal level, chosen by whoever drew the map. Two operators' greens are not the same green, and the same operator's green can move when it re-runs the model with better building data.
The probability is the part most readers skip. If a shaded square is drawn at ninety per cent cell edge confidence, the map is saying that in roughly one location in ten inside that square, a phone would fall short. That is not a defect in the map. It is what the map means, stated honestly, and it is why a single bar of solid colour across a whole valley is a summary rather than a promise.
Every Map Is Drawn at One Height, Outdoors
The model was run for a phone standing in the open. It was never run for the bottom of this staircase, and it has no way to tell you what happens there.
The same rule is explicit about where the phone is standing. Maps "must depict outdoor coverage, to include both on-street or pedestrian stationary usage, and in-vehicle mobile usage", with signal levels as they would be measured at "the industry standard of 1.5 meters above ground level".
1.5 m
the height above open ground a coverage model assumes you are standing
That single number explains most of the disappointments. A sunken courtyard, a staircase down to a bar, a railway underpass, a metro platform: none of them are one and a half metres above open ground, and none of them were modelled. Terrain grids are coarse too, so a gorge narrow enough to swallow a signal can be narrower than the cell the map is drawn in. The physics of why a hillside and the city on the horizon disagree is covered in more detail alongside the band question, and it is the same physics that makes the map's height assumption load-bearing.
The Map Stops at the Front Door
An outdoor map contains no estimate of what your walls do. Getting from the street into a building costs signal, and the cost swings enormously with what the building is made of: rendered stone and reinforced concrete are expensive, and modern energy-efficient glazing is worse than either, because the metallic coating that keeps the heat out is very good at keeping radio out with it. Basements, lift lobbies and underground car parks are effectively a different country.
This is why one apartment can be perfect at the window and useless in the bathroom, on a street the map paints a confident single colour. No better map fixes it, because the thing that varies is on the wrong side of the front door. What you can influence is which network the phone is allowed to attach to when it goes indoors, and whether calls can hand over to Wi-Fi when the signal thins — both of which are decided by the profile on the device rather than by the building. If that mechanism is new to you, how a travel eSIM attaches to a network is the shorter way in.
Covered and Usable Are Two Different Questions
Colour on a map is a threshold, and a threshold is a floor rather than an expectation. The same American rule spells its floors out: 4G areas are drawn where a user should expect at least "5 Mbps and user upload speed of 1 Mbps at the cell edge", and the 5G tiers sit at 7 Mbps down and at 35 Mbps down. Those are honest engineering minimums. They are also the difference between a map application that works and a video call that does not.
The loading assumption compounds it. Fifty per cent is a modelling convention, not a forecast of the afternoon you turn up: the map has no idea it is August, that the beach below the mast is full, and that a few thousand other phones are attached to the same cell. Capacity is the thing coverage maps are structurally worst at showing, because it is the one variable that changes hour by hour while the colours stay exactly where they were.
So decide what you need before you look. Messages and directions survive almost anything the map calls covered. A working day, a video call home or a hotspot for a laptop needs a different answer, and the map will not volunteer it.
The Map Belongs to One Network, and a Visitor May Be on Another
Every footprint you find belongs to whoever drew it. A country has several networks, they were built in different orders, and they do not fail in the same places — one may have taken the motorways and the coast, another the cities and the rail corridors. A map is therefore an answer to "where does this operator reach", which is only your answer if you are on that operator.
A visiting phone frequently is not. Which network it attaches to abroad follows the agreements behind the plan on the device, and the answer can be a single partner network, or several with an order of preference, or a different partner from one region to the next. It can also change during the trip without anything visible happening on screen. Before a map is worth reading, it is worth knowing what is actually sold for the country you are going to — because that decides which of the country's maps is yours.
One further trap: a domestic footprint can include a layer a visitor never gets, or a generation an older handset cannot use at all. A country that has retired one of its network generations will still draw a full map, and it will be a full map of something your phone may not be able to join.
Crowdsourced Maps Measure Where People Already Went
On a measured map this quay is blank. That is a statement about where people walk with their phones, not about whether there is signal on it.
The alternative to a predicted map is a measured one, built from readings contributed by real handsets. It has an enormous advantage — it is evidence rather than arithmetic — and a bias that is easy to miss: it only knows the places its contributors went, at the times they were awake, on the devices they happened to be carrying.
That sample follows roads, railways, town centres and tourist routes, because that is where people are. It thins out on footpaths, thins further on water, and stops entirely at the edges of the map's own audience. Time of day is baked in the same way. Regulators know this, which is why the American challenge process only accepts consumer tests taken "outdoors between the hours of 6:00 a.m. and 10:00 p.m. local time", and only counts an area as contested when the negative results are spread across locations and separated by at least four hours — a single bad reading is noise, not evidence.
| Aspect | Predicted | Measured |
|---|---|---|
| Built from | A propagation model | Readings from phones |
| Sample | Every square, evenly | Where people go |
| Position | Outdoors, 1.5 m up | Pocket height, anywhere |
| Time | None — it is a model | Whenever someone tested |
| Blank means | Predicted too weak | Nobody measured |
The honest use of the two together is to believe the overlap. Where a modelled map and a measured one agree, you have something close to a fact. Where they disagree, you have a question worth asking a human being who lives there.
Three Questions That Turn a Map into a Decision
A map becomes useful the moment you stop asking whether a place is covered and start asking three narrower things.
Whose map is this, and will I be on it? Find out which operator drew it, then find out which network your plan actually puts you on. If those are different, the picture is interesting rather than relevant.
What does the colour promise, and is that what I need? Look for the legend's thresholds. If the strongest band is defined at a handful of megabits, treat it as messages and maps, not as an office.
Where will I really be standing? Indoors, below ground, in the lee of a hill, at eight in the evening in the busiest week of the year. Every one of those sits outside what the map modelled, and together they explain nearly every "but it said full coverage" story anyone tells.
Common Questions
Why Does the Map Show Full Coverage Where I Have No Signal?
Almost always because you are not where the map is. The model assumed open ground at head height, and you are indoors, one floor down, or behind a hill too small for the terrain grid to notice. A ninety per cent confidence band also allows for the location that falls short — that is a stated part of the claim rather than an error in it.
Is a Crowdsourced Map More Trustworthy than the Operator's?
It is better evidence and a worse sample. Real readings beat a model in the places people go, and tell you nothing at all about the places they do not. Use the operator's map for shape and the crowdsourced one for confirmation, and treat a disagreement between them as the interesting result.
Does an Operator's Map Show What My Phone Will Get Abroad?
Only if your phone ends up on that operator, and that is decided by the agreements behind your plan rather than by the map. A visitor is often carried by one partner network in a country with three, so two thirds of the coverage a resident sees may be irrelevant to the trip.
Are Blank Areas on a Map Really Dead Zones?
On a modelled map, blank means the prediction fell under the threshold, which is a real if approximate claim. On a measured map, blank usually means nobody has been there with a phone reporting results. The two look identical and mean completely different things.
The Short Version
A coverage map is a model of an outdoor world at head height, painted in bands that carry a probability and a minimum speed inside them, drawn by one operator about its own network. None of that makes it dishonest. It makes it narrow, and knowing exactly how narrow is the difference between a map that misleads you and a map that tells you something.
Read the legend before the colours. Check whose network it is. Then ask yourself where you will genuinely be standing, and how fast the week actually needs to be.
See what is available for the country you are going to Coverage, plans and live prices, per destination, before you decide what the map is worth.
