Buying a second circuit feels like buying protection. Often it isn't. If your two links share a duct, a building entry, a local exchange, a stretch of physical route, or an upstream carrier, a single failure can take both down at the same moment, and you've paid twice to be protected once. That's the difference between dual and diverse. Dual means two connections. Diverse means two connections with nothing in common to fail. Real resilience lives in the second one.
The two words get used interchangeably, and that's where the risk starts. Dual is about quantity: two circuits instead of one. Diverse is about independence: two circuits that don't share a single point of failure. You can easily have dual without diverse, and it's more common than most teams realize. Two links from two different providers can still ride the same duct into your building, terminate at the same local exchange, or even run over the same underlying carrier's fiber with two different logos on the invoice. On paper you have redundancy. In the ground, you have one cable. The useful question isn't whether you have two links, but what those two links have in common.
That false diversity hides in the places you can't see on an order form:
Not one of these shows up on the order form, which is exactly why diversity so often breaks without anyone noticing.
Because the failures that actually cause outages tend to hit shared infrastructure, and they aren't rare. The Uptime Institute reports that outages tied to fiber and connectivity incidents have more than doubled against their recent historical average, and that third-party providers now account for the majority of publicly reported outages. When those outages hit, they are costly: ITIC's research puts the price of a single hour of downtime above $300,000 for more than 90% of mid-size and large enterprises.
Now look at what those failures have in common. A contractor's backhoe through a duct, a flooded exchange, an upstream carrier's outage: each one takes down everything riding that shared element, and a second circuit on the same element goes down with the first. Duplicating a service over shared infrastructure duplicates the service, not the protection. One event, both links, at the same time. The uncomfortable math is that you can pay for two connections and still be running a single point of failure.
Network engineers have a name for this: a shared risk link group. It's the set of links that would all fail together because they depend on the same physical thing, whether that's a conduit, a building, or a power feed. Two circuits in the same shared risk group are not diverse, however different they look on paper. The whole point of diverse design is to make sure your primary and your backup don't sit in the same group, because redundancy you haven't checked for shared risk is a line item, not a safeguard.
Genuine diversity means removing shared points of failure one layer at a time. Four checks do most of the work:
Yes, and increasingly it's the simplest way to guarantee real diversity. Because cellular uses a completely different medium, it can't share a duct, an exchange, or a building entry with your fixed line, which removes the most common causes of correlated failure in a single step. Modern 5G is also fast enough to carry real work during a fixed-line outage, not just keep a card reader alive, and it can be deployed in days rather than the weeks or months a second fiber build can take. The more layers of independence you can stack, the fewer ways a single event can take you fully offline.
The best way to do this is to ask, and to get it in writing. Since the risky shared elements are invisible on an order form, treat "diverse" as unproven until a provider confirms the specifics: is the second path physically diverse end to end, who owns the underlying network for each circuit, where does each one enter the building, and are those commitments written into the SLA rather than implied. If a provider can't answer, that's your answer. The shortcut around the whole interrogation is medium diversity, because a wireless path sidesteps most of these questions: it can't share the physical route in the first place. Diversity is an operational discipline, not a one-time purchase, and it's worth confirming before you need it.
The organizations that stay online in a crisis aren't the ones with the most connections. They're the ones with the fewest shared points of failure. That's a sourcing and design question before it's a hardware one.
It's the principle we work from at Pure IP. Because we source connectivity across many carriers and countries rather than reselling a single network, a second path can be genuinely independent rather than the same fiber with a different label. Our managed SD-WAN best practices recommend diverse active/active links, so the traditional backup circuit proves itself every day instead of sitting idle and being exposed during an outage. For inbound calls, HyperNetwork adds carrier-diverse routing for your most important numbers, so one carrier's problem doesn't ruin the voice network.
If you're not sure whether your current setup is dual or genuinely diverse, talk to our team and we'll help you identify the potential failure points before they create a massive outage.