VoIP Quality of Service: QoS setup and troubleshooting

VoIP Quality of Service, or QoS, prioritises voice traffic when a network is busy. It can reduce delay, jitter and packet loss caused by local congestion, helping business calls remain clear and consistent.

However, QoS cannot repair a faulty broadband service, add missing capacity or improve a poor microphone. Good VoIP quality depends on the complete path: the internet connection, local network, Wi-Fi, equipment, provider and the other caller’s service.

This guide explains how QoS works, which measurements matter and how UK businesses can diagnose and improve VoIP call quality.

What is Quality of Service?

Quality of Service is a group of network techniques used to classify, prioritise and manage traffic. When several applications compete for limited capacity, QoS can give delay-sensitive traffic such as voice priority over transfers that can safely wait.

For example, a large file backup can pause for a fraction of a second without anyone noticing. A voice call cannot. Late or missing speech packets may create gaps, robotic audio or an awkward delay.

QoS policies can identify traffic, mark it with a priority and place it into an appropriate queue. Network devices then apply those rules during congestion.

What is VoIP QoS?

VoIP QoS applies traffic-management rules to the signalling and media used by internet calls. The aim is to protect real-time voice from less urgent network activity.

A business may use QoS on routers, firewalls, switches, wireless access points and a software-defined wide-area network. However, each device must be configured consistently. A priority mark that is removed or ignored later in the path may provide little benefit.

Wayv’s business VoIP service information explains the wider phone-system context.

Why is QoS important for VoIP?

Voice is interactive and time-sensitive. A web page can recover a missing packet or wait for delayed data, but a conversation must continue in real time.

QoS becomes especially useful when:

• many colleagues make calls at the same time
• voice shares a connection with backups, video or large downloads
• several sites use the same wide-area network
• upload capacity is limited
• users report problems only during busy periods

When the network has plenty of spare capacity, QoS may appear to do very little. Its value becomes clearer at the moment queues form and the device must decide which packets leave first.

Which measurements affect VoIP quality?

Latency

Latency is the time data takes to travel from one endpoint to another. High latency creates a noticeable conversational delay, causing people to interrupt or talk over one another.

Measure one-way latency where the platform provides it. A simple ping normally reports a round trip, so it is useful for comparison but does not tell the whole story.

Jitter

Jitter is variation in packet arrival time. Even when average latency looks reasonable, inconsistent arrival can make speech break up.

VoIP devices use a jitter buffer to smooth small variations. A buffer cannot hide unlimited delay, though, and making it too large may add latency.

Packet loss

Packet loss occurs when data fails to reach its destination. Voice codecs can conceal a small amount, but repeated or burst loss causes clipped or robotic audio.

Do not look only at the average. Short bursts can damage a conversation even when the overall percentage seems low.

Available bandwidth

Each simultaneous call needs upload and download capacity. The exact amount depends on the codec and network overhead.

Count the busiest realistic number of calls, not the number of licences. Then allow space for ordinary business traffic and growth.

Mean Opinion Score

Mean Opinion Score, or MOS, is a common way to express perceived voice quality on a scale. Some platforms estimate MOS from network data rather than asking people to score a call.

Use MOS as one indicator alongside latency, jitter, loss and real user reports. Different systems may calculate it differently.

Useful VoIP QoS targets

Targets vary by provider, codec and network design. Nevertheless, these figures are sensible investigation thresholds for many business voice environments:

• One-way latency: aim to stay below about 150 milliseconds.
• Jitter: aim to stay below about 30 milliseconds.
• Packet loss: as close to zero as possible; sustained loss around 1% deserves investigation.

These are not guarantees or universal pass marks. A call can sound poor below a threshold because of burst loss, Wi-Fi interference, faulty equipment or the far-end network. Likewise, some platforms may tolerate different conditions.

Measure during normal and peak periods. An overnight test cannot reproduce the congestion that appears when the whole team starts work.

How does VoIP QoS work?

1. Classification

First, the network identifies traffic by trusted markings, application, protocol, address or another rule. Classify by traffic type rather than assuming every packet from one device deserves priority.

2. Marking

Next, the device can apply a Differentiated Services Code Point, or DSCP, value. Real-time voice media is commonly marked Expedited Forwarding, DSCP 46, when that matches the phone provider’s design.

Signalling may use a different class because it is important but normally less sensitive than the audio stream. Follow the provider’s documented values instead of applying DSCP 46 to everything.

3. Queuing

When an interface becomes congested, the router or switch decides which packets leave first. A low-latency or priority queue can protect voice while fair queuing shares the remaining capacity.

Priority queues need limits. Otherwise, excessive or wrongly marked traffic could starve other applications.

4. Shaping and policing

Traffic shaping smooths bursts by holding packets briefly, while policing drops or re-marks traffic that exceeds a limit. Both tools can support a wider QoS policy, but poor settings may introduce loss or delay.

5. Trust boundaries

A trust boundary defines where the network accepts or replaces a device’s markings. For example, a managed switch may trust an approved phone but re-mark traffic from an ordinary user device.

This prevents an application from claiming unjustified priority. It also creates a consistent policy at the edge of the managed network.

Where should QoS be configured?

Apply QoS where congestion can occur and where the business controls the device. Typical points include:

• the internet-facing router or firewall
• wide-area network links between sites
• managed switches with busy uplinks
• wireless networks carrying voice
• software-defined network policies

The outbound internet link is often the first place to investigate because upload capacity can be much lower than download capacity. The router must know the real usable rate; otherwise, the provider’s equipment may become the uncontrolled queue.

QoS markings may not be honoured across the public internet. Even so, local queuing still helps before traffic leaves the business network.

What can QoS fix?

QoS can help when local congestion causes voice packets to wait behind other traffic. It is particularly effective when problems align with backups, cloud synchronisation, video meetings or large uploads.

In that situation, a well-designed policy can:

• reduce the time voice waits in a queue
• limit the effect of bursty applications
• protect call quality during predictable peaks
• give monitoring clearer traffic classes

A general introductory guide to router QoS is retained from the original article. Business networks, however, should use the phone and network providers’ documented designs.

What can QoS not fix?

QoS is not extra bandwidth and cannot correct every fault. It will not fix:

• an unstable or undersized internet service
• packet loss occurring inside the provider or far-end network
• weak Wi-Fi signal or radio interference
• a damaged cable, port or handset
• a poor microphone, speaker or headset
• incorrect firewall or NAT behaviour
• a codec or platform fault
• a power or broadband outage

If a link is saturated most of the day, prioritisation may simply move the problem to another application. Additional capacity or a separate connection may be the better answer.

How to diagnose poor VoIP call quality

1. Define the symptom precisely

Ask whether the problem is delay, one-way audio, dropped calls, robotic speech, low volume or an echo. Also record who experienced it, which direction was affected and the exact time.

2. Identify the pattern

Check whether faults affect one user, one site, one network, one destination or every call. A pattern often narrows the search faster than changing settings at random.

3. Compare wired and wireless calls

A wired test can show whether Wi-Fi is contributing. If only wireless users are affected, review signal strength, interference, roaming and access-point capacity.

4. Check peak utilisation

Compare call times with upload, download and interface graphs. Look for short bursts as well as sustained use.

5. Review platform data

Use call detail records and platform diagnostics to inspect latency, jitter, loss, codec and endpoint information. Keep examples of good and bad calls for comparison.

6. Test each network segment

Check local ports, switches, firewalls, WAN links and internet paths. A high ping may indicate congestion, distance or another routing issue; this earlier ping explainer is preserved from the original guide.

7. Change one thing at a time

Document the baseline, make a controlled change and measure again. Otherwise, the team may not know which action helped or introduced a new problem.

How to configure QoS for VoIP

Exact commands depend on the network platform, so treat this as a design sequence rather than a universal configuration:

1. Document the voice platform. Obtain the provider’s IP ranges, ports, protocols and DSCP guidance.
2. Measure the link. Record the realistic usable upload and download rate, including peak performance.
3. Estimate concurrent demand. Calculate voice capacity from the codec and busiest call volume.
4. Set the trust boundary. Decide which managed devices may supply trusted markings.
5. Classify and mark correctly. Separate voice media, signalling and ordinary data.
6. Create a bounded priority queue. Reserve enough capacity without allowing voice to consume the whole link.
7. Shape to the usable rate if required. Keep the controllable queue on the managed device.
8. Apply the policy consistently. Check switches, wireless, firewalls and WAN equipment.
9. Test during congestion. Reproduce realistic competing traffic and make test calls.
10. Monitor and review. Update the policy when links, applications or working patterns change.

Businesses without suitable in-house skills should use a qualified network professional. Earlier small-business technology commentary is retained from the original article, but the decision should be based on the risk and complexity of the actual network.

Common VoIP QoS mistakes

Marking all traffic as high priority

If everything is urgent, nothing is prioritised. Reserve real-time treatment for the traffic that genuinely needs it.

Trusting every endpoint

An unmanaged application could mark bulk traffic as voice. Define where markings are accepted and where they are replaced.

Using the contracted speed instead of usable capacity

Throughput can vary, particularly on contended or asymmetric services. Shape and plan against measured, dependable capacity.

Ignoring upload traffic

Voice travels in both directions. A saturated upload can damage outbound speech even when download tests look fast.

Applying settings only to the router

A congested switch or wireless network can still delay packets before they reach the router. Review the full path.

Skipping monitoring

A QoS policy can become outdated as users, applications and circuits change. Regular monitoring shows whether the classes and limits still match reality.

Should voice use a separate VLAN?

A voice VLAN can separate phone traffic logically from ordinary user devices. This can simplify security, addressing, monitoring and QoS policies.

However, a VLAN does not create bandwidth or automatically guarantee quality. The network still needs correct routing, switching, QoS and resilience.

Softphones on computers and mobiles may not fit a traditional voice-VLAN design. Classify their traffic using methods supported by the platform and network rather than relying only on a physical phone port.

Does VoIP work well over Wi-Fi?

Yes, when the wireless network is designed for real-time voice. Nevertheless, Wi-Fi is shared radio capacity and can be affected by interference, weak signals and roaming delays.

Use suitable coverage, channel planning, current security and managed Quality of Service features. Test calls while people move between coverage areas if roaming matters.

For fixed desks with repeated problems, a wired connection is a useful comparison and may be the more dependable option.

VoIP Quality of Service FAQs

Does every VoIP network need QoS?

Not always. A small network with ample spare capacity may work well without complex policies. QoS becomes more valuable when links are shared, constrained or busy.

What DSCP value is commonly used for voice?

Voice media is commonly marked Expedited Forwarding, DSCP 46. Always follow the voice provider’s design and apply trust controls rather than marking every packet the same way.

Can QoS stop dropped calls?

It can help when congestion is the cause. It cannot fix provider outages, faulty equipment, firewall errors or an unstable connection.

Is 150 milliseconds of latency acceptable for VoIP?

About 150 milliseconds one way is a common investigation threshold, not a guaranteed pass mark. Lower and consistent latency usually makes conversation feel more natural.

How much packet loss is acceptable?

As close to zero as possible. Even low average loss can sound poor when packets disappear in bursts, so investigate sustained loss around 1% and use real call evidence.

Does faster broadband remove the need for QoS?

Not necessarily. More capacity reduces the chance of congestion, but busy periods can still create queues. Monitoring shows whether QoS remains useful.

Who should manage VoIP QoS?

The network and voice teams should work together. The voice provider supplies application requirements, while whoever manages the routers, switches and wireless network applies and monitors the policy.

Build quality from end to end

VoIP QoS is most effective as part of a complete call-quality plan. Start with dependable connectivity and a sound local network, then use classification, marking and queuing to protect voice during congestion.

Finally, monitor real calls and review the policy whenever the network changes. Clear evidence is more useful than applying generic settings and hoping for the best.