Most businesses know they should have a backup connection. Far fewer could tell you what would actually happen if the main line went down at half past nine on a Tuesday. Automatic failover is the bit that makes a backup useful: the switch happens on its own, nobody rings IT, and the first person to notice is usually the one whose video call went briefly odd. Here is how it works, what it needs, and why so many failover setups quietly do not.
The 30-second version
- Failover is a second connection that takes over by itself when the first one fails. Nobody plugs anything in.
- The router does the work, watching the main line and rerouting traffic when it stops answering.
- A mobile SIM is the usual backup because it does not share anything with your fixed line.
- The switch takes seconds to a couple of minutes. Calls in progress usually drop, everything else reconnects.
- Untested failover is the most common failure. If it has never been tried, assume it does not work.
On this page
What automatic failover actually is
A backup connection on its own is not failover. Plenty of businesses have a spare router in a cupboard, or a mobile hotspot somebody can share in an emergency. That is a contingency plan, and it depends on somebody noticing, finding the kit and knowing what to do, usually while the phones are ringing.
Automatic failover removes the person from the middle. One device holds both connections, watches the primary one constantly, and moves traffic across the moment it stops working. When the main line comes back, it moves everything back.
The difference in practice
With a spare router, an outage costs you the time it takes somebody to notice, fetch it and set it up, which is rarely less than twenty minutes and often much more. With automatic failover it costs you the length of the switch, which is usually under a minute.
How the switch actually happens
The router does not simply wait for the cable to be unplugged. Most outages are not that obvious: the line is still physically up, but nothing is getting through. So the router tests the connection continuously, normally by pinging a known address out on the internet every few seconds.
When enough of those tests fail in a row, it declares the primary down and sends traffic out of the backup instead. The thresholds matter more than people expect:
| Setting | What it does | Why it matters |
|---|---|---|
| Test target | The address the router pings to prove it is online | Pointing it at your own provider can hide a fault further out |
| Test interval | How often it checks | Shorter means faster detection and slightly more background traffic |
| Failure count | How many failed tests before it switches | Too low and a brief wobble triggers a pointless switch, too high and you sit offline waiting |
| Recovery count | How many successful tests before it switches back | Stops it flapping between the two lines while the main one is unstable |
| Failback behaviour | Whether it returns to the main line automatically | Usually yes, but not always immediately, and not always at a convenient moment |
Sensible defaults get you most of the way. The point of knowing these exist is that if your failover switches when it should not, or sits there while everyone complains, these are the numbers somebody needs to look at.
What it needs
- A router that supports two connections. Either a dual WAN router with two sockets, or a router with a SIM slot built in. This is the part most standard broadband routers cannot do.
- A second connection that is genuinely separate. A mobile 4G or 5G data SIM is the usual choice because it shares no cable, no cabinet and no exchange with your fixed line.
- An allowance sized for a bad week, not a bad hour. A backup that runs out of data halfway through an outage is not a backup.
- Somewhere with signal. The router is often in a comms cupboard, which is frequently the worst place in the building for mobile reception. An external aerial solves this and is cheap.
Why the SIM should be on a different network path
If your fixed line and your backup SIM both end up on the same infrastructure, a fault in the wrong place takes out both at once. Our guide to true network diversity covers where backups secretly overlap.
How fast is the switch, and what breaks during it
Detection usually takes between ten seconds and a minute, depending on how the tests are set. The switch itself is close to instant once the decision is made. So in practice you are offline for somewhere between fifteen seconds and two minutes.
What that feels like depends on what people are doing:
| What is happening | What the user sees |
|---|---|
| Email, files, browsing | A pause, then it carries on. Most people do not notice |
| A call in progress | Usually drops and has to be redialled |
| Video meetings | Drops out and reconnects, often without leaving the meeting |
| Card payments | A failed transaction that needs retrying |
| VPN and remote desktop | Disconnects and reconnects, sometimes needing a login |
| Anything with a fixed IP | May stop working entirely, because the address changes on the backup |
That last one catches people. If a supplier reaches your building on a fixed IP address, or you allow access somewhere based on your IP, that stops working the moment you fail over to a mobile connection, unless the SIM has been set up with a fixed IP of its own.
Three things that quietly stop failover working
- It shares a path with the main line. A second broadband circuit from a different supplier that runs down the same duct into the same cabinet is not diversity, it is two ways of losing the same fault. A mobile SIM avoids this almost entirely.
- The SIM has quietly died. Backup SIMs sit unused for months. They get disconnected for inactivity, run out of allowance, or the contract lapses, and nobody finds out until the day it matters. This is the single most common failure we see.
- It has never been tested. Failover that has not been tried since installation is a theory. Settings drift, firmware updates change defaults, and somebody may well have unplugged something in the cupboard to plug in a Christmas tree.
The one we find most often
An empty backup SIM. It is also the easiest to prevent: we monitor the SIMs we supply, so a backup line that has gone quiet shows up on our side rather than yours.
How to test it without upsetting everybody
Testing failover properly means causing an outage, which is why it rarely happens. Do it deliberately, at a quiet time, and tell people first.
- Pick a quiet slot and warn the team that calls may drop for a minute.
- Unplug the main line at the router, not at the wall, so you are testing the router rather than the socket.
- Time how long it takes before traffic flows again, and note what broke.
- Actually use it for ten minutes. Send an email, make a call, take a test payment. A connection that comes up but cannot do the job has not passed.
- Plug it back in and time the failback, because that can be the disruptive half.
- Write down what happened and repeat it twice a year, or after any change to the setup.
Not sure whether your failover would actually work?
Tell us what you have and we will check it properly: whether the backup shares a path with your main line, whether the SIM is still alive, and what would really happen on the day. It costs nothing to find out.