Why bandwidth alone cannot determine whether an internet connection is suitable for enterprise operations.
A Johannesburg business receives two proposals for a 200 Mbps fibre connection. Both advertise the same download speed. One provides shared broadband access with standard support arrangements; the other specifies symmetrical capacity, defined performance commitments and an enterprise service-level agreement. On the quotation, the principal difference may appear to be price. In operation, the difference can extend to cloud application performance, voice quality, fault management and the business’s ability to maintain essential services.
This is the central question in enterprise connectivity in South Africa: not simply how fast a connection can transfer data, but whether its technical and contractual characteristics correspond to the work the business requires it to perform.
A 200 Mbps service is not automatically more suitable than a 100 Mbps service. Nor does the word *fibre* establish how capacity is allocated, how traffic is routed, what happens during congestion or how a fault will be managed.
For businesses procuring connectivity, the correct starting point is therefore an operational specification rather than a speed package.
Bandwidth: capacity must follow the application
Bandwidth describes the rate at which data can be transferred across a connection, usually expressed in megabits per second (Mbps). It is an essential specification, but it does not describe the complete behaviour of a network.
Consider an office with 50 employees. Some are accessing cloud accounting systems, others are participating in video meetings, while automated backups, file synchronisation and software updates generate additional traffic.
The number of employees alone does not establish the required bandwidth. What matters is the combination of applications, their traffic patterns, the number of simultaneous users and the direction in which information travels.
An enterprise should distinguish between sustained traffic and short-duration peaks. It should also identify operations that cannot tolerate congestion, such as business calling, remote desktop sessions or time-sensitive transactions.
Microsoft’s network guidance for Teams illustrates the distinction. Video meeting bandwidth requirements vary according to meeting type, resolution, participant behaviour and network conditions. A business cannot reliably determine its connection requirements from a single universal bandwidth allowance per employee.
A practical procurement assessment begins with measured utilisation where available, supported by an inventory of critical applications and expected growth.
The objective is not to purchase the largest bandwidth figure. It is to secure sufficient capacity, with appropriate headroom, for the organisation’s actual workload.
Symmetrical speeds: why upload capacity matters
Many internet packages advertise their download speed prominently while allocating substantially less capacity to uploads.
This arrangement may be adequate for environments where most traffic consists of receiving information. It can become restrictive when employees regularly send large files, synchronise cloud storage, host video meetings or transfer data between business locations.
A connection advertised as 200/50 Mbps provides up to 200 Mbps downstream and 50 Mbps upstream. A symmetrical 100/100 Mbps service provides equal capacity in both directions.
Neither is universally superior. Their suitability depends on the workload.
A design practice transferring large project files to cloud storage may place greater demands on upstream capacity than a retail branch primarily accessing a centrally hosted application. A company running multiple concurrent video meetings must account for traffic in both directions.
Upload congestion can also affect interactive applications even when download capacity appears sufficient.
For this reason, enterprise connectivity quotations should always state upload and download speeds separately. Where symmetrical capacity is required, it should be an explicit contractual specification rather than an assumption.
Contention: what is actually shared?
Two connections with identical advertised speeds can behave differently when multiple customers or applications compete for network capacity.
Contention occurs where traffic shares a finite resource. This may arise within an access network, at an aggregation point, across an ISP’s upstream capacity or elsewhere along the route to a destination.
Shared access is not inherently unsuitable for business use. Properly engineered shared networks can deliver excellent performance. The procurement issue is whether the service provides adequate capacity under the conditions in which the enterprise will use it.
An uncontended or dedicated bandwidth commitment can offer greater predictability at the specified service boundary. It does not, however, eliminate congestion elsewhere on the public internet or guarantee the performance of an external application.
Businesses should therefore ask where capacity is committed, which portions of the service remain shared and whether any minimum throughput or performance commitments are enforceable.
Terms such as uncapped, unshaped and unthrottled should also be interpreted separately.
Uncapped generally concerns data-volume limits. Unshaped concerns traffic treatment according to application or protocol. Unthrottled concerns deliberate speed restriction. None of these terms, by itself, establishes an uncontended connection or guarantees application performance.
The distinction becomes especially important when comparing ordinary business broadband with Dedicated Internet Access.
Latency, jitter and packet loss: the performance behind the speed
Bandwidth determines how much information a connection can carry over time. It does not establish how quickly individual packets travel or how consistently they arrive.
Three additional measurements are particularly important.
Latency is the time taken for information to travel between network points, commonly assessed through round-trip time. Excessive latency can make interactive applications feel unresponsive even when bandwidth is plentiful.
Jitter describes variation in packet delay. Real-time voice and video applications are sensitive to inconsistent packet arrival because they depend on a reasonably continuous stream of information.
Packet loss occurs when transmitted packets fail to reach their destination. Depending on the application and protocol, this can result in retransmissions, reduced throughput, audio interruptions or degraded video.
A speed test reporting 500 Mbps cannot establish whether a business’s voice calls will remain clear, its remote sessions responsive or its cloud applications stable.
The relevant measurements must be taken towards representative destinations, at appropriate times and under realistic operating conditions.
This is also why network performance should be assessed across the complete path. A poorly configured local network, congested wireless access point, overloaded firewall or inefficient VPN route can undermine an otherwise well-provisioned internet service.
Increasing the access speed does not necessarily correct these problems.
Read More: Enterprise Systems Integration
Access technology: fibre is not the complete specification
South African enterprises can obtain connectivity through several access technologies, including fibre, licensed or unlicensed fixed wireless, microwave links and mobile networks.
The appropriate choice depends on the location, available infrastructure, required capacity, service characteristics and commercial constraints.
Fibre is often attractive where substantial, stable capacity is required. Yet the presence of fibre at a building does not reveal whether the service uses shared passive optical infrastructure, dedicated Ethernet access or another network arrangement.
Fixed wireless and microwave connections can provide practical alternatives where fibre construction is difficult, costly or unavailable. Their performance depends on engineering factors such as radio conditions, interference, line of sight, spectrum use and equipment configuration.
Mobile connectivity can support particular operational requirements, although performance may vary with radio coverage, network loading and the conditions of the service.
An enterprise should not select an access technology solely from a general hierarchy of perceived quality.
It should establish whether the proposed technology can meet the required performance and service commitments at the specific premises.
Coverage and installation: availability must be verified at the address
A network coverage map indicates the potential availability of infrastructure. It is not necessarily confirmation that a particular building can be connected under the advertised commercial terms.
Actual service delivery may depend on spare network capacity, available ports, the route into the premises, building access permissions, wayleaves, civil works or additional installation costs.
This is particularly relevant to businesses occupying industrial premises, commercial parks, multi-tenant buildings or sites outside established fibre footprints.
A procurement process should distinguish between a preliminary coverage indication and a confirmed service feasibility assessment.
It should also establish who is responsible for the installation, customer-premises equipment, handover point and subsequent maintenance.
Where several parties participate in delivery, the enterprise needs a clear understanding of which organisation will coordinate the work and manage service escalation.
A nominally available connection has little operational value until the installation requirements, delivery programme and responsibilities are established.
Service-level agreements: the contractual meaning of reliability
A service-level agreement, or SLA, defines specified service commitments and the conditions under which they apply.
Depending on the product, these may include service availability, fault reporting arrangements, response times, restoration targets, performance measures and remedies for failures.
These terms should not be treated as interchangeable.
A four-hour response commitment does not necessarily mean the service will be restored within four hours. Similarly, an availability commitment may apply to a particular network segment or measurement period rather than every application used by the customer.
Even a 99.9% monthly availability target permits approximately 43 minutes of unavailability in a 30-day month. Whether that is acceptable depends on the operational consequences of an outage and the precise contractual exclusions.
A service credit may provide a financial remedy, but it cannot recover lost production, missed transactions or interrupted customer communications.
Enterprise buyers should examine the SLA alongside the technical specification, particularly its measurement boundaries, exclusions, escalation process and restoration obligations.
Reliability is not established by the presence of an SLA alone. It depends on what the agreement actually commits the provider to deliver.
The hidden boundary between connectivity and business operations
A company’s internet connection is only one component of its communications environment.
Between an employee and a cloud application may sit the local network, wireless infrastructure, firewall, DNS services, VPN configuration, internet access provider, upstream networks and the application provider’s own infrastructure.
A failure at any one of these points may appear to the employee as an internet problem.
This distinction matters commercially because it determines who can diagnose and resolve the incident.
An enterprise that purchases connectivity without considering its internal network and application dependencies may discover that the access circuit is performing correctly while employees remain unable to work.
The procurement specification should therefore identify not only the external connection but also the technical boundary at which responsibility passes between the network provider, the organisation’s IT environment and third-party services.
For businesses with multiple sites, the assessment must additionally consider how locations communicate, where traffic exits to the internet and which services depend on central infrastructure.
The result is a connectivity specification based on the business’s actual operating architecture.
From broadband purchasing to enterprise connectivity planning
South Africa’s expanding fibre infrastructure gives businesses more connectivity options, but a wider choice of advertised speeds does not remove the need for technical evaluation.
Enterprise connectivity should be selected according to the applications it supports, the traffic it must carry, the locations it must reach and the operational consequences of degraded or unavailable service.
Bandwidth remains important. So do symmetrical capacity, contention, latency, packet loss, access technology, confirmed coverage and the terms under which faults are managed.
Taken together, these characteristics determine whether a proposed service is suitable for the business that will depend upon it.
ML Communications works with enterprises to assess connectivity requirements, coordinate appropriate network services and integrate external connections with the communications infrastructure they support.
Read More: How a Second Internet Connection Strengthens Business Continuity
For organisations evaluating new connections, upgrading existing services or connecting additional premises, the most useful first step is a technical assessment of their requirements.
Scan the QR code to call and request an Enterprise Connectivity Assessment from ML Communications
[…] Read More: Enterprise Systems Integration Assessment […]