Business Continuity: What Happens to Your Phones in an Outage

"The phones are down" is four different events wearing one sentence: your building lost power, your internet failed, your provider had an incident, or your building itself is inaccessible. Each breaks a different link in the chain, and a continuity plan is just deciding, in advance, what should happen in each case. Cloud systems changed the defaults here profoundly, because the system itself no longer lives in the place that flooded.

Failure by failure

• Building loses power: desk phones and network gear die locally; the phone system, in a data center, keeps running. Calls follow failover rules to mobile apps and cell numbers instantly. With PoE phones on a UPS-backed switch (how that works), desks can survive short outages entirely.
• Internet fails: same shape: desk phones lose registration, the cloud keeps answering, and calls reroute to apps on cell data and forwarding targets. Callers hear normal ringing, not dead air.
• Provider incident: the failure you cannot route around locally, which is why it belongs in provider selection: multi-region automatic failover, owned infrastructure, and a status page you can actually read. Ask before you sign, not during.
• Building inaccessible (fire, flood, storm): identical to power loss from the system's perspective, but for weeks. Staff work from apps; menus get a temporary greeting; the address on file for E911 gets updated for wherever people land.

The settings that make failover real

Failover is configuration, not magic. Five things to set before the bad day: per-extension failover targets (which cell, after how many seconds), an emergency routing profile you can flip on in one action (a "storm mode" greeting and simplified menu), mobile apps installed and logged in before they are needed, at least two people with admin access who are not in the same building, and voicemail-to-email so messages remain visible when nobody is at a desk.

Testing, the step everyone skips

A continuity plan that has never run is a hypothesis. Twice a year: unplug the office router for ten minutes and observe what callers actually experience; flip the emergency profile on and off; have someone answer a forwarded call on the app from home. The test costs twenty minutes and reliably finds one broken assumption, an app logged out, a forward pointing at a departed employee's cell, a greeting recorded by someone who left.

What legacy setups cannot do

An on-premise PBX inverts every failure: power loss kills the system itself, internet loss may spare it but strands its lines, and building loss takes the whole nervous system. Much of the continuity case for hosted systems (what those are) is simply that the brain stopped living in the building that has the emergencies.