Legacy PBX to SIP trunk migration

01Inventory the numbers before touching anything

Organisations consistently have more numbers than they think, and the surplus is where migrations go wrong. Direct dial ranges, published numbers on old marketing material, a fax line in accounts, a lift emergency phone, an alarm dialler, a number that forwards to someone's mobile for reasons lost to history.

Every one of those needs finding before the old service is cancelled, because a number that is not ported is lost, and recovering it afterwards is at best slow and frequently impossible. The port request is also the long-lead item, so the inventory is genuinely the first task rather than a preliminary.

02Analogue devices do not move cleanly

Fax machines, alarm diallers, lift phones and older card terminals were designed for a circuit-switched line and do not tolerate packet networks well. They are sensitive to the compression, jitter and packet loss that voice calls tolerate easily, and they fail in ways that are hard to detect because nobody is listening.

Each needs a decision rather than an assumption: replace it with an IP-native equivalent, keep a dedicated analogue line where the device is safety-critical, or accept a gateway with uncompressed codecs and test it properly. Lift and alarm lines in particular usually warrant keeping a separate circuit, because the failure mode is a safety one and the saving is small.

03Emergency calling needs an address, not a number

On a traditional line, the emergency service knows the location because the circuit has a fixed installation address. SIP has no such property. The address associated with each number has to be registered deliberately and kept current, and if it is wrong the call routes to the wrong dispatch area.

This gets harder with remote and hybrid working, where a handset or softphone may be anywhere. Establish the policy explicitly: which devices may place emergency calls, what address is registered against each, how a moved device is updated, and what users are told about calling emergency services from a softphone. It is the part of the migration with the most serious consequence and the least attention.

• Register a correct service address for every number that can dial emergency services.
• Establish a process for updating the address when people or sites move.
• Decide whether softphones outside the office may place emergency calls at all.
• Tell users what the policy is; assumption is not an acceptable control here.
• Test emergency routing during commissioning through the proper test procedure.

04Quality problems are network problems

Voice is unusually intolerant of jitter and loss. A network that carries file transfers and video calls without complaint can produce audibly poor voice, because the failure threshold for real-time audio is far lower and users notice immediately.

Assess the network before migrating rather than diagnosing after. That means measuring jitter, loss and latency on the paths voice will actually take, including the internet connection at each site, and implementing prioritisation for voice traffic. Quality problems found after cutover are painful because everybody experiences them at once and the obvious culprit is the new phone system.

05Run both in parallel, then port

The safest sequence establishes the new platform alongside the old one, tests it with real users on temporary numbers, and only then ports the production numbers across. Porting is the irreversible step and it should happen last, once the system it is porting into is known to work.

Port in stages where the provider permits it. A small non-critical group first proves the process and the routing, and it means a problem affects a handful of people rather than the whole organisation. Porting everything in one window is faster and leaves no room to discover anything.

06Rebuild the call flows deliberately

Old PBX configurations accumulate two decades of changes, and the call flow that exists is rarely the one anyone would design. Migration is the opportunity to rebuild it intentionally, and also the point at which undocumented behaviour that people quietly depend on disappears.

Walk the actual flows with the teams who use them before rebuilding, particularly out-of-hours routing, holiday handling, overflow and escalation to mobiles. The person who knows why calls divert a particular way on Thursdays is usually not in the project, and their absence is discovered on the first Thursday after cutover.