On-site PBX or hosted service: the link decides, not the invoice
The comparison usually turns on cost per extension and on the feature list. Those are often a draw. What actually decides is what happens to the phones when the connection drops.
Enterprise telephony tenders are almost always settled on two columns: monthly cost per extension, and the feature list. On both, an on-site exchange and a hosted service end up looking alike. The first concentrates the spend up front and the second spreads it; the everyday features have been the same on either side for years.
The question that decides appears in neither column. It is this: what happens when the connection drops?
The link decides, and it decides alone
A hosted service places the call logic off site. The handsets are terminals and nothing more. When the link is down they can no longer reach the service that registers them, and telephony disappears entirely — outside calls, but also internal calls between two offices in the same building.
An on-site exchange keeps its call logic locally. Losing the link costs the site its outside calls; it does not stop the site working internally, and a modest backup line is then enough to keep essential outbound calls available.
Where connectivity is stable and genuinely redundant, this difference does not matter. Everywhere else it is decisive, and it is the first thing to establish before comparing prices.
Three situations, three answers
A single, well-served urban site. The hosted service is usually right. It removes equipment to maintain, a maintenance contract and a skill to keep in house. The condition is redundant connectivity whose route diversity has been verified rather than merely purchased.
A branch network. The answer is rarely uniform. A single numbering plan and central administration are worth having across the estate, but branches with fragile connectivity gain from keeping local call capability. A mixed design, with the core hosted and exposed sites holding an autonomous relay, costs more to design and less to run.
A remote site. A border post, a technical station, a distant mining site. There, telephony is a safety measure before it is a working tool, and it has to outlive the loss of the link. Local equipment is not up for discussion; how it connects to the rest of the network is.
What to establish before choosing
Four things determine the answer, and none of them is a telephony feature.
Measured availability of the connection over the past twelve months, not the figure in the contract. Whether a second path exists, whose infrastructure carries it, and where it physically runs. Power for the terminal equipment, including the switch that powers the handsets. And what the organisation is prepared to lose during an outage, expressed in functions rather than in minutes.
A tender that does not ask these four questions is comparing offers that answer different needs.
Reversibility is arranged at the start
Changing your mind costs little if three points were settled when you chose.
The numbers belong to the organisation, and portability is in writing. The numbering plan, hunt groups and diversions are documented outside the tool that hosts them. The carrier connection uses a standard link that other providers know how to take over.
None of these costs anything to obtain at signature. Each becomes a negotiation on the day you want to leave.
