A video meeting freezes just as a customer joins. Calls become clipped while a large cloud backup runs. Staff can still browse the web, so the connection appears to be working - but the services that keep the business moving are not. Knowing how to prioritise network traffic prevents this familiar problem by ensuring critical applications receive the network capacity and low latency they need.
Why prioritisation matters
For most organisations, traffic prioritisation is not about giving every application more bandwidth. It is about making deliberate decisions when demand exceeds available capacity, whether that happens on an internet circuit, between sites, or across a busy office Wi-Fi network. The result should be clearer voice calls, more dependable cloud access and fewer productivity issues that are difficult to trace.
Start with business impact, not application names
The first question is not which traffic generates the most data. It is which traffic causes the greatest operational impact if it is delayed, lost or interrupted. A VoIP call uses relatively little bandwidth, for example, but is highly sensitive to delay, jitter and packet loss. A large file transfer may use far more capacity but can usually wait a few seconds without affecting the user.
This distinction matters because prioritisation is most valuable during congestion. When a circuit has ample spare capacity, queues are rarely needed. When a link is busy, however, the network needs clear rules for deciding what is sent first and what can wait.
A sensible starting point is to speak with department leads as well as IT. Finance may rely on a hosted accounting platform at month end. A contact centre may depend on call quality throughout the day. A warehouse may need uninterrupted access to stock systems, scanners or labels. These are business requirements, not just technical preferences.
How to prioritise network traffic in practice
Most businesses use Quality of Service, commonly called QoS, to classify traffic and apply a defined treatment. In simple terms, QoS identifies a type of network traffic, marks it where appropriate, and places it into a queue with rules for bandwidth, delay and priority.
A practical policy normally separates traffic into a small number of classes rather than trying to create a rule for every application. Too many classes are difficult to maintain and make faults harder to diagnose. Four or five well-defined categories are generally more effective:
Voice is often placed in a priority queue, but that does not mean it should be allowed to consume unlimited bandwidth. A well-designed policy reserves enough capacity for the expected number of simultaneous calls and places a firm limit around the queue. This protects call quality without allowing a misclassification or unusual event to crowd out other essential services.
Business-critical applications may receive guaranteed bandwidth or preferential treatment, while bulk activity is shaped to a set rate. Shaping is particularly useful for cloud backup and large uploads. If they are permitted to fill the upstream connection, they can affect calls, video and interactive cloud applications even when download speeds look healthy. Expert network consultancy can help you design classes that reflect how your business actually works.
- Real-time communications, such as hosted PBX, VoIP and video conferencing, need the lowest possible delay and jitter.
- Business-critical applications, such as line-of-business platforms, secure remote access and transaction systems, require predictable performance.
- Standard business traffic, including email, ordinary web use and collaboration tools, should perform well but can tolerate short delays.
- Background and bulk traffic, such as software updates, cloud backups, synchronisation and large downloads, can use remaining capacity.
- Unwanted or high-risk traffic should be restricted or blocked through appropriate security controls, rather than merely deprioritised.
Classify traffic reliably
Classification is the foundation of QoS. It can be based on several factors: application recognition, source or destination network, port and protocol, device type, or an existing traffic marking such as DSCP. The right method depends on the environment.
Application-aware firewalls, SD-WAN platforms and managed routers can often recognise common SaaS, voice and collaboration traffic without relying only on port numbers. This is useful because many modern applications use encrypted web traffic and frequently changing cloud addresses. However, application recognition is not perfect, so policies should be tested against real usage.
DSCP markings can preserve priority information as traffic passes through a managed LAN and WAN. They are valuable where the organisation controls the path end to end. They should not be treated as a guarantee across the public internet, where markings may be ignored or overwritten. The key is to apply QoS at points you manage, especially at the edge of an internet connection or on links between sites.
Put QoS where congestion actually occurs
Prioritisation has to be applied at the constrained link. For a single office, this is often the WAN edge or internet firewall, particularly on the outbound path. For a multi-site organisation, congestion may occur on site-to-site links, local breakout connections or cloud security gateways.
This is one reason a business should understand its true available bandwidth. A dedicated internet access circuit provides uncontended, SLA-backed capacity, making it easier to plan and enforce traffic policy. FTTP can offer excellent performance too, but actual service characteristics, contention and upstream capacity should be considered before supporting a voice-heavy or cloud-dependent operation.
SD-WAN adds another useful layer. It can identify applications, steer selected traffic over the most appropriate connection and react to packet loss, latency or circuit failure. A voice call can be directed over the healthiest path while lower-priority data uses another link. With suitable backup connectivity, the business can maintain essential services during an outage rather than treating every connection failure as a full stop.
QoS cannot manufacture capacity. If a site regularly has more critical traffic than its connection can support, increasing bandwidth or adding a secondary circuit is the proper remedy. Prioritisation buys order under pressure; it is not a substitute for network design.
Protect voice, video and cloud services without starving everything else
Real-time services deserve special attention because users notice their failures immediately. Voice packets need consistent delivery, not huge amounts of bandwidth. Packet loss, jitter and excessive latency can make calls sound broken even when a speed test reports a fast connection.
Calculate the number of concurrent calls the business genuinely needs, allow headroom for signalling and other real-time traffic, then reserve an appropriate portion of capacity. Video conferencing may need its own class or may sit below voice, depending on how essential it is to operations. In a small office, treating all collaboration traffic equally may be sufficient. In a busy contact centre, voice should usually take precedence.
Cloud applications introduce a different challenge. Not all SaaS traffic is equally time-sensitive. An interactive CRM or remote desktop session should be treated differently from an overnight file synchronisation task. Where business applications share an internet connection with backups, updates and guest access, separate VLANs or SSIDs can improve control as well as security.
Guest Wi-Fi should never compete freely with core operations. Give it a defined rate limit and keep it logically separate from the corporate network. The same approach can apply to personal devices, streaming and non-essential downloads where they create avoidable contention.
Monitor, test and adjust the policy
A QoS policy should be treated as an operational control, not a one-off configuration. Monitor bandwidth use, top applications, latency, jitter, packet loss and queue drops over time. The goal is to identify whether a problem is caused by congestion, a failing circuit, poor Wi-Fi coverage, an application issue or an incorrectly classified flow.
Test during the moments that matter. A policy that appears sound at 10am may fail when calls, backups, Teams meetings and cloud usage peak together. Controlled testing can include placing simultaneous calls while running a managed backup job, transferring a large file between sites or switching traffic to a backup connection.
Review the policy when the business changes. New cloud software, a larger contact team, a new site or a move to hybrid working can alter demand quickly. Rules that once protected the business may become too restrictive or no longer reflect what staff rely on.
Avoid the common mistakes
The most common error is prioritising everything. If every application is marked critical, none receives meaningful preference when the network is busy. Keep the highest-priority class small, justified and protected by sensible limits.
Another mistake is relying on bandwidth tests alone. They do not show whether traffic is being queued correctly under load, nor do they reveal intermittent loss or jitter. Equally, avoid giving an application priority simply because a supplier requests it. Check its real business dependency and test the outcome.
Finally, do not overlook security. A device or user that is compromised can create substantial traffic and disrupt operations. Endpoint, identity and email security, network security segmentation and clear access controls all reduce the chance that unwanted activity consumes the capacity reserved for the business.
The best traffic policy is one employees never have to think about: calls remain clear, critical systems respond when they should, and background work stays in the background. That level of reliability comes from matching QoS rules, connection capacity and backup design to how the organisation actually operates.
