Corporate VoIP and Hosted PBX
Reliable voice services for multi-user and multi-site organisations.
Design and deploy business telephony with network readiness, resilience, quality and support considerations.
Discuss this requirement →Start with the environment, risk and required outcome.
Design and deploy business telephony with network readiness, resilience, quality and support considerations. Namandla Consultants begins with evidence and recommends a proportionate engagement rather than a generic technology package.
Connected expertise for the wider environment.
Clear steps. Visible decisions.
How business VoIP is actually put together
Three architectures, and the choice is usually made for the wrong reason.
Hosted PBX, also called cloud PBX
The call platform runs in the provider’s data centre. You have handsets, softphones and an internet connection. The provider owns the platform, the upgrades and the resilience.
Suits most organisations. Low capital cost, predictable monthly cost and it scales in either direction without hardware. The dependency is your connectivity, which makes the failover design the important part of the specification rather than an afterthought.
On-premises PBX with SIP trunks
The call platform is a server or appliance on your site, connected to the carrier over SIP trunks rather than physical lines. You own it and you maintain it.
Still the right answer in specific cases: sites with poor or expensive connectivity where internal calling has to survive an outage, contact centres with heavy customisation, and environments with integration into systems that need local presence. It carries capital cost, a maintenance burden and a hardware refresh cycle.
Microsoft Teams as the phone system
Teams becomes the phone, with external calls delivered through Direct Routing, Operator Connect or Microsoft’s own calling plans. No handsets required, though handsets are supported.
Compelling where Teams is already the collaboration platform, which in South African corporate environments it usually is. It removes a separate system to maintain and it puts calling into the application people already have open. The trade is that call handling features are less deep than a mature PBX, and contact centre requirements usually need an add-on.
| Hosted PBX | On-premises PBX | Teams calling | |
|---|---|---|---|
| Capital cost | Low | High | Low |
| Monthly cost | Per extension | Trunk and maintenance | Per licensed user |
| Survives internet outage | No, without failover design | Internal calls yes | No |
| Feature depth | High | Highest | Moderate |
| Suits | Most multi-site business | Contact centres, poor connectivity | Teams-centric organisations |
Direct Routing or Operator Connect
Where Teams is the platform, the connection to the public network is the decision.
Operator Connect is a carrier arrangement configured through the Teams admin centre. Simple, fast and limited to participating carriers.
Direct Routing connects Teams to any SIP carrier through a session border controller, either your own or a provider’s. More work to set up, and it gives you carrier choice, least cost routing, analogue device support and the ability to keep an existing carrier relationship. For most South African organisations with existing carrier contracts and a mix of number ranges, Direct Routing is the practical route.
What decides call quality
Call quality is a network property, not a phone system property. Four things determine it and they are all measurable.
Bandwidth
A G.711 call uses roughly 87 kbps in each direction once overhead is counted. Opus and G.729 use materially less at some cost to fidelity. Thirty concurrent calls on G.711 is about 2.6 Mbps each way, which is not large, and it must be available consistently rather than on average.
Latency
One-way delay above roughly 150 ms produces the talk-over effect where both parties start speaking at once. Beyond 300 ms the call is difficult.
Jitter
Variation in packet arrival. Above about 30 ms the jitter buffer cannot compensate and you get the choppy, digitised effect. Jitter is usually the culprit when a call is intermittently poor rather than consistently poor.
Packet loss
Above one percent is audible on most codecs. Above three percent is unacceptable. Loss on a link that shows plenty of spare bandwidth almost always means a buffering or duplex problem rather than a capacity problem.
Quality of service, which is where most deployments fail
Voice is small packets at a constant rate and it is intolerant of delay. Bulk data is large packets in bursts and it does not care. On an unmanaged network they compete and voice loses, because a file transfer will fill any available buffer.
The fix is straightforward and it is routinely omitted.
Mark the traffic
DSCP EF for voice media, AF31 or CS3 for signalling, applied at the handset or the access switch.
Trust the marking
Switches and routers must be configured to honour DSCP rather than rewriting it to zero, which is the default behaviour on a lot of hardware.
Queue it
Strict priority queuing for the EF class on every egress interface, particularly the WAN link where the bottleneck usually sits.
Separate it
A voice VLAN keeps voice traffic off the data broadcast domain and makes the policy simple to apply.
Without this, a VoIP deployment works during the pilot with four users and degrades as it scales, which is the pattern behind most of the “we tried VoIP and went back” stories in this market.
Wireless voice needs specific attention. Handsets and softphones on Wi-Fi need WMM, proper roaming configuration and access point density planned for voice rather than for coverage. A wireless network that is fine for laptops will drop calls during roaming between access points.
The South African specifics
Number porting
Geographic and non-geographic numbers port between licensed operators. It works and it takes time, typically two to four weeks depending on the donor operator and the accuracy of the paperwork. The single most common cause of delay is a mismatch between the account details held by the donor and those submitted on the port request. We check that before submitting rather than after a rejection.
Keep the numbers
Your published numbers appear on signage, vehicles, stationery, directories, Google Business Profile and every document you have issued for a decade. Port them. Changing them costs far more than the porting fee in ways that are hard to see until afterwards.
The carrier is licensed, we are not
Electronic communications services in South Africa require an ICASA licence. Namandla designs, deploys and manages the platform and the network. The licensed service is provided by the carrier, and we work with your existing carrier or help you select one. Any provider presenting themselves as both the integrator and the licensed operator should be able to show you the licence.
Least cost routing
With SIP trunking, calls can be routed through different carriers by destination. Mobile termination, national and international rates vary between carriers, and routing by destination is a genuine saving on high-volume estates. It adds configuration complexity and it is worth it above a certain call volume, not below it.
Failover
VoIP over a single fibre link has a single point of failure. The standard arrangement is an LTE or secondary fibre backup with automatic failover, plus a divert rule at the carrier that sends inbound calls to a mobile number or a voicemail service when the trunk is unregistered. That last part is a five-minute configuration that a surprising number of deployments never had, and it is the difference between an outage that customers notice and one they do not.
Power
Handsets are powered over Ethernet from the access switch. If the switch is not on a UPS, the phones are not either. This is worth checking at the design stage rather than discovering it.
What Namandla delivers
Assessment and design
Current call volumes, number inventory, contract positions and network readiness. Architecture selection with the reasoning written down. Failover design.
Network preparation
Voice VLAN, QoS marking and queuing across switches and routers, PoE budget verification, wireless voice configuration where softphones or wireless handsets are in scope.
Deployment
Platform configuration, call flows, auto-attendants, hunt groups, ring strategies, call recording where required, handset provisioning and user setup.
Teams calling
Direct Routing or Operator Connect, session border controller configuration, voice routing policy, dial plan and emergency calling.
Number porting
Managed end to end, including the paperwork that causes the rejections.
Integration
CRM click-to-dial, call recording, contact centre platforms where required.
Migration
Parallel running through cutover, so calls do not stop while numbers move.
Ongoing management
Platform administration, call quality monitoring against MOS, carrier liaison and moves, adds and changes under the managed services agreement.
Onsite work is Gauteng. Configuration, porting and management run nationally.
Common questions about Corporate VoIP
How much does business VoIP cost in South Africa?
Hosted VoIP is priced per extension per month, with call costs billed separately or bundled into a minute allocation. Handsets are a once-off cost or rented. The saving against traditional lines comes mainly from call rates and from removing line rental rather than from the platform itself.
The variables that move the figure are extension count, whether handsets are purchased or rented, call volume and destination mix, whether call recording is required and whether failover connectivity has to be added. Compare quotations on the total including call spend at your actual volumes, because a low per-extension price with high call rates is common and it is not a saving.
Will VoIP work on our internet connection?
Usually yes. Voice needs very little bandwidth, roughly 87 kbps per concurrent call on G.711. What it needs is consistency. Latency, jitter and packet loss matter far more than raw speed.
A well-provisioned fibre line handles voice comfortably. The failures we see are almost always internal: no quality of service configuration, voice sharing a VLAN with bulk data, or a wireless network not built for roaming voice. Assessing the network before deployment takes a day and it prevents the most common cause of a disappointing rollout.
What is the difference between hosted PBX and SIP trunking?
Hosted PBX means the whole phone system runs in the provider’s data centre and you connect handsets to it. SIP trunking means you keep a PBX on your premises and the carrier delivers lines to it over the internet instead of over copper.
Hosted suits organisations that would rather not own a phone system. SIP trunking suits organisations with an existing PBX that still has life in it, or with requirements that need a local system. Both remove physical lines and both reduce call costs. The difference is who owns and maintains the platform.
Can we keep our existing phone numbers?
Yes. Geographic and non-geographic numbers port between licensed operators in South Africa. Allow two to four weeks and expect the timeline to be driven by the donor operator.
Port them rather than changing them. Published numbers appear on signage, vehicles, stationery, directories, your Google Business Profile and every document issued over the life of the business. The cost of changing a number is mostly invisible until afterwards. The most common cause of a rejected port is a mismatch between the account details held by the donor and those on the request, which is worth verifying before submission.
Can we use Microsoft Teams as our phone system?
Yes, through Direct Routing, Operator Connect or Microsoft calling plans. Where Teams is already your collaboration platform it removes a separate system to maintain and it puts calling into the application people already have open.
The trade is feature depth. Teams handles standard business calling well and it is less capable than a mature PBX for complex call flows, and contact centre requirements generally need an add-on platform. For most corporate environments that is not a constraint. For a service desk or a sales floor it might be, and it is worth testing against your actual call handling before committing.
What happens to our phones if the internet goes down?
Without failover, hosted VoIP stops. With it, calls continue over a secondary link. The design has two parts and the second is often missing.
The first part is connectivity failover, an LTE or secondary fibre link with automatic switchover. The second is a carrier-side divert rule that sends inbound calls to mobile numbers or voicemail when the trunk is unregistered. That rule is quick to configure and it means an outage is invisible to the person calling you. We configure both as standard.
Do we need new handsets?
Not necessarily. Any SIP-compatible handset can usually be reprovisioned to a new platform. Handsets from a proprietary system generally cannot, and traditional analogue and digital handsets cannot.
Analogue devices that have to stay, such as lift phones, gate intercoms, fax lines and alarm dialers, connect through an analogue telephone adapter. Those devices are the ones most often forgotten in a migration and then discovered when someone is stuck in a lift, so we inventory them explicitly at survey stage.
Is VoIP call quality as good as a normal phone line?
On a properly configured network, better. Wideband codecs carry more of the voice frequency range than a traditional line does, so calls between compatible endpoints sound noticeably clearer.
The qualifier is real. Quality is determined by the network, and a deployment without quality of service configuration will sound worse than what it replaced under load. This is the single biggest determinant of whether an organisation is happy with VoIP two years later, and it is entirely within your control.
How long does a VoIP migration take?
Four to eight weeks for a typical corporate site, with number porting the longest single item. Network preparation and platform configuration run in parallel with the port.
The sequence that works is: assess, prepare the network, build and test the platform, submit the port, run in parallel with the old system live, then cut over. Parallel running matters. Cutting over at the same moment the numbers port removes your ability to fall back, and the fallback is what makes the cutover low risk.
Can VoIP integrate with our CRM?
Yes. Click-to-dial from a contact record, screen pop on inbound calls matching caller ID, and call logging against the record are standard on most platforms, either natively or through an API.
Depth varies. Native integrations for common CRM platforms generally work out of the box. Anything custom needs development against the platform API and should be scoped separately rather than assumed. Ask for a demonstration against your specific CRM rather than a compatibility list.
Do you provide the phone lines yourself?
No. Electronic communications services in South Africa require an ICASA licence and the licensed service is provided by the carrier. We design, deploy and manage the platform and the network underneath it, and we work with your existing carrier or help you select one.
That separation is worth understanding when comparing providers. Some integrators resell carrier services and some carriers offer integration. Both models work. What matters is knowing who is accountable for the line when a call fails, and having that written down before it happens.
When is VoIP the wrong choice?
When your connectivity is genuinely unreliable and cannot be made redundant, when your site has a small number of extensions and an existing system that works, or where a specific regulatory requirement mandates a physical line.
For a five-person office with a working phone system and no growth plans, migrating is effort for a marginal saving. The case gets strong at multiple sites, at meaningful call volumes, where staff move between locations, or where the existing system needs replacing anyway. If you are in the first group, we will say so.