Every CDN publishes a world map covered in dots, and every dot is technically true. This guide is the interrogation method: what a dot does and does not claim, the four questions that separate presence from performance, and the checks that turn a marketing map into shortlist evidence.
What a dot actually claims
A dot on a coverage map asserts, at minimum, that the provider has equipment in a facility in that metro. It does not assert that your traffic will be served from there. Between the dot and your user sit routing policy (many networks serve some markets from regional hubs despite local presence), capacity tiers (a two-server presence and a multi-terabit fortress are the same dot), commercial gating (some POPs serve only certain plans or products), and peering reality — a POP without dense local eyeball-network interconnection can be slower than a better-connected POP a country away. None of this is dishonesty; it is the difference between a facilities list and a delivery promise, and the map only ever shows the first. Note too that several majors have stopped publishing per-city lists entirely, which makes independent checking more necessary, not less.
The four questions per dot
For each city that matters to your audience, interrogate the dot with four questions. Is it always-on for my traffic? Ask directly whether your product tier and workload serve from that POP under normal routing, or only under load/failover. What is its capacity class? Vendors rarely give numbers but will usually characterize tiers (full-stack metro vs embedded cache vs partner rack). Who does it peer with locally? The eyeball networks carrying your users are the question — a POP peering with the two ISPs that carry 80% of your national audience beats three dots that don't. What happens when it fails? The next-best serving location for that market tells you your degraded-mode latency, which for thin regions is the number that actually hurts.
Desk checks that cost nothing
You can verify a surprising amount before any benchmark. Public looking-glass and traceroute services in your target countries show where a provider's hostname actually resolves and terminates from local networks. Community measurement platforms and published IX participant lists reveal whether the provider is present at the exchanges your users' ISPs use — the peering-density signal that regional surveys like our Africa field piece lean on. And your own diaspora works: a colleague, customer or friend in the target city running a single traceroute and a timed download gives you a real data point per minute spent. Log every check; three cities of desk evidence routinely reorders a shortlist before the formal benchmark begins.
When the map matters and when it doesn't
Coverage maps discriminate hardest at the edges of the market: emerging regions, archipelago geographies, and any audience concentrated outside the North Atlantic corridor — exactly where the regional comparisons in this library show the field diverging. For an audience concentrated in well-served metros, nearly every serious vendor's map passes, and your decision should weight the factors maps cannot show: cache behavior, operations, price, support. The failure mode this guide exists to prevent is the middle case — a map that looks global, an audience with a meaningful tail, and a contract signed before anyone checked whether the dots in the tail serve your traffic. Check the tail first; then let the benchmark and trial settle what the desk cannot.