Signal strength, signal quality and network load - and the second and third matter more than people expect.

Strength alone is misleading. A strong signal that is heavily interfered with, or received from a cell serving hundreds of other users, performs far worse than a moderate clean signal on a quiet cell. Cellular routers report both strength and quality metrics, and the quality figures are the ones that explain inconsistent performance.

Network load is invisible from the router and varies by time of day. A link tested at nine in the evening may behave quite differently at nine in the morning.

Band matters too: lower frequency bands penetrate buildings far better while carrying less capacity, and higher bands do the reverse. A router locked to the wrong band, or one that keeps switching, performs erratically.

The practical method is to survey at the actual installation position with the actual antenna, at more than one time of day, before committing.

Frequently the difference between an unusable link and a reliable one - it is the highest-value component in the installation.

The stub antennas supplied with routers are omnidirectional and low gain, designed for convenience in good coverage. In marginal coverage they are inadequate.

An external antenna helps in three ways: it gets outside the building, so the walls and any metal cabinet stop attenuating the signal; it can be mounted high and clear of obstructions; and a directional type concentrates its sensitivity toward the serving mast, adding useful gain in that direction.

Aiming matters for directional antennas, and the router's own signal reporting is the tool - adjust and watch the figures rather than guessing from a map. Modern routers use two or four antennas for MIMO, and all of them must be connected and spaced correctly or throughput suffers even with a strong signal.

Keep the cable runs short: loss in antenna cable at these frequencies is substantial, and a long thin cable can waste the gain the antenna provided.

Because coverage and outages are carrier-specific, and a second carrier is the only genuine diversity a cellular connection can have.

Two SIMs from different networks let the router fail over when one carrier has an outage, when a mast is congested or under maintenance, or when coverage changes. Since carrier incidents are a real proportion of cellular downtime, this converts a single dependency into two.

It also helps where the router moves. A vehicle or temporary site crossing coverage boundaries benefits from being able to attach to whichever network is present.

The alternative is a multi-network or roaming SIM, which achieves something similar with one SIM by allowing attachment to several networks under one account. It is simpler to administer and often the better answer for fixed sites; dual physical SIM gives more explicit control and separate billing.

Configure the failover conditions and the fallback behaviour deliberately, including when to return to the primary, or a router can end up running months on the expensive backup.

Not by default. Most cellular addresses are private and translated by the carrier, so inbound connections are impossible without arranging otherwise.

The usual answer is for the router to establish an outbound tunnel to a central point, and for all access to arrive through that. This works on any SIM, needs no special plan, and is how most remote monitoring is done.

Where inbound access is genuinely required, carriers offer fixed public addresses or private APNs, both usually at extra cost and requiring a business account. A private APN also puts the SIMs onto a private network reaching the organisation directly, which is the cleanest arrangement for a fleet of remote sites.

Be cautious about exposing management interfaces on a public cellular address even where one is available. They are scanned constantly, and a router with a public address and a weak password is found within hours.

The outbound-tunnel approach avoids the whole question and should be the default.

By monitoring it on the router, alerting before the limit, and being deliberate about what is allowed to use the link.

Cellular data is metered, and overage charges are significant. The failure mode is well known: a fixed line goes down overnight, the router fails over, a backup job or a software update runs over the cellular link, and the bill arrives.

Routers used for backup should therefore know they are on backup and restrict traffic accordingly - policy that permits business-critical applications and blocks bulk transfer, updates and backups while the cellular path is active.

Set usage counters and alerts on the router as well as relying on the carrier's portal, since the carrier's figures lag.

For fleets, a pooled data plan across all SIMs is usually far more economical than individual allowances, because usage between sites averages out.

Also check what the router itself uses. Frequent probes, telemetry and cloud management traffic add up across a year on a small allowance.

Where the coverage exists and the throughput is needed, yes - but LTE remains adequate for most backup and remote-site duties.

5G offers higher throughput and lower latency where the network supports it. For a primary connection at a site with no fixed line, or one carrying video or many users, that is a real difference.

For a backup link carrying business-critical traffic only while the main circuit is down, LTE is usually sufficient and the hardware is cheaper.

Coverage is the deciding factor. Much 5G is deployed on higher frequency bands with less building penetration, and a site with marginal indoor coverage may perform better on LTE. Routers supporting both fall back automatically, which is the sensible default.

Also weigh the lifespan. A router installed today may be in service for many years, and LTE-only hardware will have a shorter useful life as networks re-farm spectrum. For new installations that are hard to reach, the extra cost of a 5G-capable unit is usually justified on that basis alone.

Where the antenna can be placed well - which usually means the router follows the antenna, not the other way round.

The instinct is to install the router in the comms cabinet with everything else. That is frequently the worst position: a metal enclosure in the middle of a building attenuates heavily, and long antenna cables lose much of what is left.

The better approach is to establish where the antenna needs to be - typically outside, high, with a clear line toward the serving mast - and then mount the router as close to it as practical, extending the Ethernet rather than the antenna cable. Ethernet can run 90 metres with no loss; antenna cable cannot.

Where the router must sit in a cabinet, use external antennas mounted on or outside the cabinet and keep the cables short and of appropriate quality.

For industrial and outdoor positions, choose a DIN-rail or IP-rated model with a wide temperature range, and consider PoE-powered variants that need only one cable to the mounting position.