Buying connectivity for the places the network does not reach

Plenty of New Zealand operations run from somewhere fibre is not, or is not yet, or would take four months and a trenching quote to become. Marinas, orchards, remote plant, construction sites that will be gone before the install date.

The instinct is to ask what the fastest available option is. Usually the more useful questions are what the site actually needs, and what happens when the link drops.

Start with what the site does when the link is down

This is the question that determines the design, and it gets skipped.

If the answer is that staff cannot dispatch, cannot scan, cannot take payment or cannot open a gate, then you are not buying connectivity, you are buying availability, and one link of any speed is the wrong shape regardless of how fast it is.

If the answer is that people are mildly inconvenienced and work catches up later, a single cellular connection is fine and the money is better spent elsewhere.

Most sites are somewhere between, and the useful output is a short list: these three functions must keep working, everything else can wait.

Bandwidth is usually not the constraint

For operational sites, raw speed is less often the problem than three other things.

Latency and jitter matter for anything interactive, and for voice. A link with good throughput and poor consistency feels broken to a user even when the graph looks fine.

Upload matters more than most sales conversations suggest. Camera footage, photo evidence from the field, backups and cloud application traffic are all outbound. Plans marketed on download can be badly lopsided.

Contention and coverage at the specific spot decide what you actually get. Cellular performance in a shed with a steel roof, or in a gully, bears no relationship to the coverage map. This is worth testing on site before committing, with the equipment in the position it will live in, because the difference between a good and a poor antenna placement can be larger than the difference between two carriers.

Redundancy that is genuinely independent

Two links from the same carrier down the same road is one link with extra billing. Independence is what you are paying for: different medium, different path, ideally different carrier.

Cellular as failover for a fixed line is a common and sensible pattern. Failover should be automatic and tested, and someone should know when it has happened, because a site quietly running on a metered backup link for three weeks is an expensive way to find out the primary is down.

The parts people forget

Power. A connection that survives a cut but a router that does not is not resilience. Modest battery backup on the network equipment is cheap and turns a whole class of outages into a non-event.

Physical security of the equipment. On a yard or a wharf, the cabinet matters. Weather, dust, forklifts and curiosity are all real.

Who is on the same network. Site connectivity usually ends up carrying cameras, controllers and operational equipment as well as staff devices. Those do not belong in the same segment as a guest phone, and it is far easier to design that at install than to retrofit it later.

Contract term against site life. A three-year term on a site that closes in nine months is a cost nobody accounted for. Match the commitment to the operation, and ask about relocation before you need it.

How we approach it

We are a One NZ partner, so we can supply and manage connectivity, and we would rather say plainly that this is something we sell. It also means the connection is managed alongside the network, the devices and the security on it rather than as a separate account with a separate helpdesk, which is the actual value: when a site goes quiet, one party is responsible for finding out why.

Before any of that, the question is still what the site cannot do without. The answer usually costs less than expected, and gets spent in a different place than first assumed.