Hosted PBX is a cloud-based business phone system where call routing runs in redundant data centres instead of hardware on your premises, delivered as a monthly subscription. A PABX (Private Automatic Branch Exchange) is that on-premises hardware, a physical box in your comms room. VoIP (Voice over Internet Protocol) is the technology carrying calls as data packets rather than analogue signals over copper. In short: VoIP is the transport, PABX is the legacy hardware, hosted PBX is the cloud service replacing it.
Unlike a traditional PABX, hosted PBX is built to survive site power and fibre failures through carrier redundancy and automatic failover, the kind of protection that matters in a market where load shedding has, in recent years, taken on-premises systems offline entirely for extended periods. Even with the grid currently stable, that risk hasn't structurally disappeared, and resilience built in now is insurance against it returning.
The Six Components of a Hosted PBX System
A hosted PBX is not one product. It has six parts, and providers differ materially on each:
- Call routing platform, cloud software deciding where calls go. Lives in the provider's data centres.
- Carrier interconnects, physical routes to the public telephone network. Diversity here determines whether one carrier outage takes you offline.
- Endpoints, desk phones, softphones, mobile apps, video room systems. What users touch.
- Site connectivity, the fibre or LTE link between premises and platform. Most common single point of failure.
- Failover layer, UPS, LTE backup, platform-level call-redirect rules. Governs behaviour when components 3 or 4 fail.
- Data and integration layer, recordings, call detail records, API access. Determines what you can do with the data later.
Evaluate a hosted PBX across all six components. Most buying mistakes come from comparing endpoint feature lists while ignoring interconnects, failover and data access.
What Hosted PBX, PABX and VoIP Actually Mean
These three terms describe different layers of one system: PABX is the hardware, VoIP is the transport, hosted PBX is the cloud service replacing the hardware. Confusion in buying conversations comes from treating them as competing products.
Analyst tracking of the local UCaaS market shows cloud-delivered telephony taking the majority of new business phone deployments in South Africa, with on-premises PABX shipments in structural decline.
We regularly see procurement documents asking for "VoIP or PABX quotes" as though those are alternatives. They aren't. You can run VoIP into an on-premises PABX, and run a hosted PBX over a traditional-looking desk phone. The real question is where call-routing intelligence lives, and what happens when that location loses power.
PABX vs Hosted PBX: The One-Line Difference
A PABX is a physical box in your comms room routing calls between extensions and out to the public network. Hosted PBX moves that routing logic into redundant data centres, delivered as a subscription.
Everything follows from that. On-premises means you own the hardware, maintenance contract, licensing and the failure. Hosted means you own configuration and outcome, the provider owns infrastructure.
One practical consequence: adding 40 extensions to an on-premises PABX usually means an expansion card and a scheduled engineer. On a hosted platform it's a configuration change made the same afternoon.
Where VoIP Fits: It's the Pipe, Not the Product
VoIP carries voice as data packets rather than analogue signal over copper. It's the transport layer. SIP (Session Initiation Protocol) is the signalling standard setting up and tearing down calls.
"We're moving to VoIP" is like saying "we're moving to TCP/IP," true, but it doesn't describe what you're buying. What you're buying is a system: hosted PBX, UCaaS, or SIP trunking into existing hardware.
The distinction matters at contract stage. A SIP trunk supplier sells connectivity. A hosted PBX provider sells the whole telephony estate. Conflating them is how businesses end up with a modern trunk feeding a PABX from 2011.
What "Video PBX" Adds That Voice-Only Systems Can't
A video PBX treats video as a native media type on the same extension as voice, not as a separate conferencing product bolted alongside the phone system.
Concretely: extension 2041 rings on desk phone, laptop softphone and mobile. If a caller escalates to video mid-call, it happens on the same session with the same recording policy, routing rules and reporting. One directory, one dial plan, one system to administer.
The alternative, what most South African SMEs run today, is a hosted PBX for calls plus a separate video conferencing subscription. Two vendors, two directories, two invoices, no shared call detail records.
PABX is hardware, VoIP is transport, hosted PBX is the service. The decision that matters is where call-routing intelligence lives and what happens when that site loses power.
Why Uptime Is the Only Spec That Matters in South Africa
Uptime is the correct primary evaluation criterion here because every other feature is worthless during the hours your system is offline. Feature matrices assume availability. In this market, availability has historically been the variable, and even during the current period of grid stability, the infrastructure decisions that protected businesses through 2022 and 2023 remain the right ones to have in place.
Eskom's own published figures show load shedding peaked at 24,638 gigawatt-hours in 2023, with the Energy Availability Factor as low as 54.69% that year. At Stage 4, a typical site historically lost power roughly 4 to 6 hours per day across two or three windows, 60 to 90 business hours per month. No call queue, IVR tree or CRM integration compensates for a phone system dark a third of the working month.
What Happens to Your Calls at Stage 6
Without failover design, calls to your main number simply fail. Callers hear a ring tone terminating nowhere, or a network error. Your business appears closed.
With hosted PBX the routing platform stays online, it sits in a data centre with generator backup, not your comms room. But that solves half the problem. If your site loses internet and power, handsets are unreachable even though the platform is healthy.
The gap between "the platform is up" and "customers can reach us" is where most hosted PBX deployments quietly fail. Closing it is a design decision, not a product feature.
The Failover Stack: Three Layers, Three Failure Modes
- Power. A UPS sized for router, switch and PoE handsets, not the whole office. A 1500VA unit typically carries a small switch stack and several phones through a 2.5-hour window. For a 200-seat site: UPS at the network core, battery backup on the fibre ONT, reception and critical desks prioritised.
- Connectivity. An LTE or 5G failover router taking over automatically when fibre drops. Voice is light on bandwidth, so LTE backup, inadequate for general office use, handles voice comfortably.
- Routing. Platform rules redirecting calls to mobile numbers, a secondary site, or a mobile app when a location goes unreachable. Requires no on-site power at all, the most reliable of the three.
A site with all three rarely misses a call. A site with none misses most of them.
Geographic Redundancy and Carrier Diversity
Platform redundancy means call routing runs across more than one data centre, and voice traffic reaches the public network through more than one carrier.
Innovo's hosted PBX platform runs on Huawei Cloud infrastructure, engineered for high availability across multiple regions.* Innovo Networks' enterprise partnerships with Vodacom and Liquid Telecoms exist precisely so traffic routes across independent national networks rather than a single point of failure.
Ask any provider directly: which carriers do you interconnect with, and what happens to my calls if one has a national incident? Vague answers are informative.
* Innovo's Video PBX has run on Huawei Cloud's Johannesburg region for eight years at a measured 99.99% availability. See our full uptime report for what's independently verifiable and how the figure is evidenced.
How to Read an SLA, and What "99.99%" Actually Permits
An SLA percentage tells you how much downtime is contractually acceptable, not how much you'll experience.
- 99.9% permits roughly 8.7 hours of downtime per year
- 99.95% permits roughly 4.4 hours
- 99.99% permits roughly 52 minutes
That looks decisive until you read exclusions. Most SLAs exclude scheduled maintenance, third-party carrier failure and last-mile connectivity. Those exclusions cover the majority of real-world South African outages.
Two questions: does the SLA cover the last mile to my premises, or only the platform? And is the remedy a service credit or a response commitment with defined timelines?
A hosted PBX is only as available as its weakest layer. Platform uptime means nothing without site power, connectivity and routing failover designed alongside it.
South Africa's First Video PBX: Voice and Video on One Platform
When video runs natively on the PBX rather than alongside it, you administer one system instead of two, and every policy set for voice applies automatically to video.
Workplace communication research consistently finds the average organisation running multiple overlapping communication platforms, with material duplicated spend and administrative overhead.
This is the architectural reason we built Innovo's video PBX as a single platform rather than integrating third-party conferencing. The value isn't video itself, video conferencing is commodity. The value is eliminating a parallel system.
One Extension, Any Media
A single extension follows the user across devices: desk phone, mobile app, desktop softphone, boardroom video endpoint, all answering the same number.
For an IT manager this collapses a persistent burden. Onboarding is one account, not three. Offboarding is one deprovisioning action, not a scavenger hunt. Reporting covers all communication in one set of records.
Why Bolted-On Video Creates Two Systems to Manage
Separate video platforms create four recurring problems: duplicated user directories drifting out of sync, inconsistent recording and retention policies, split reporting making utilisation impossible to assess, and two support relationships when something breaks.
The last is most expensive. When a video call drops, the conferencing vendor blames the network, the network provider blames the application, and your IT manager spends a morning proving which is right.
Bandwidth Reality on a South African Link
Video needs roughly ten to twenty times the bandwidth of voice, so size links on peak concurrent video sessions, not headcount.
Working figures: 100 kbps per concurrent voice call, 1 to 2 Mbps per concurrent HD video session. A 200-seat site at 20% voice concurrency needs about 4 Mbps, trivial on modern fibre. Add ten concurrent video sessions and you need another 10 to 20 Mbps, with QoS configured so video can't starve voice during congestion.
Native video on the PBX removes an entire parallel platform from your estate. The saving is architectural and administrative, not just licensing.
Frequently Asked Questions
What is the difference between PABX and Hosted PBX?
A PABX is physical hardware at your premises routing internal and external calls. Hosted PBX moves that functionality into the cloud, delivered over VoIP. The practical difference: no on-site hardware to maintain, and no single point of failure sitting in your comms room.
Will my phones work during load shedding?
Only if designed for it. Hosted PBX keeps call routing online in redundant data centres, but you still need UPS or LTE failover for on-site handsets and internet. Properly configured, calls automatically redirect to mobile numbers when the site drops.
Next: what migration actually costs and takes. Read our companion guide on migrating off a legacy PABX, real Rand pricing, and 12 questions to ask any provider.
Get a Quote