VoIP call quality problems — the crackle, the robotic warble, the half-second delay that makes everyone talk over each other, the call that simply drops mid-sentence — are almost never a reason to abandon internet telephony and crawl back to the old phone line. In the overwhelming majority of UK business cases they are network faults: jitter, latency, packet loss, a misconfigured router or a SIP trunk that was never tuned. This guide walks you through a systematic diagnostic and fix process so you can find the actual cause instead of guessing, and know exactly when a problem has moved beyond a quick setting change into escalate-to-a-specialist territory.
By the end you will be able to measure the three numbers that decide whether a voice call sounds good — jitter, latency and packet loss — against the thresholds the ITU publishes; configure Quality of Service so voice traffic is protected from everything else on your line; spot the router and firewall settings (SIP ALG, NAT timers, port ranges) that quietly wreck calls; diagnose the classic SIP trunking pitfalls that cause one-way audio and dropped calls; and plan bandwidth so a busy afternoon never starves the phones. Every figure here is anchored to the UK 2026 reality: the Openreach PSTN switch-off, Ofcom rules, real leased-line and broadband behaviour, and the kind of managed VoIP support a UK provider will actually put in front of you.
What actually determines VoIP call quality?
Voice over IP turns your speech into a stream of tiny packets — typically one every 20 milliseconds — and sends them across the same network that carries your email, cloud apps and backups. Traditional phone lines gave every call a dedicated circuit, so quality was fixed the moment the call connected. VoIP shares the road, which means quality is decided moment to moment by how cleanly those little packets arrive. Three measurements capture almost everything that matters, and a fourth summarises the result.
Latency is the one-way delay between you speaking and the other person hearing it. The ITU-T G.114 recommendation puts the comfortable ceiling at 150 ms one way; beyond about 300 ms round trip, people start stepping on each other’s sentences. Jitter is the variation in that delay — packets that were sent evenly but arrive in bursts. A jitter buffer smooths small amounts, but sustained jitter above roughly 30 ms produces the choppy, underwater sound everyone recognises. Packet loss is exactly what it sounds like: packets that never arrive. Because voice cannot wait for retransmission, anything above 1 percent is audible as clipped or missing syllables.
The industry rolls these into a single score called MOS (Mean Opinion Score), a 1–5 rating where 4.0 and above is considered good business quality and 4.3 is the practical ceiling for a compressed codec. When someone says “the calls sound bad”, what they mean, measured properly, is that one of latency, jitter or loss has pushed the MOS down. Fix the network condition and the score — and the experience — comes straight back up.
Before you change a single setting, get a baseline. Most hosted VoIP portals expose per-call MOS, jitter and packet-loss figures, and free tools will run a 60-second line test. Without a number to measure against, you cannot tell whether a change helped or you simply got lucky on a quiet line.
Two ways to fix VoIP call quality — and when each is right
Broadly, there are two remediation paths once you know voice is degraded. The first is the in-house route: you work the diagnostics yourself, adjust QoS and router settings, and lean on your provider’s support desk for the SIP-side pieces. The second is a managed VoIP arrangement, where a provider owns the network path, the QoS policy and the SIP trunk end to end and is contractually on the hook for quality. Neither is automatically correct — it depends on how complex your estate is and how much call quality costs you when it slips.
Self-managed troubleshooting
In-house IT tunes the network
Managed VoIP provider
Provider owns quality end to end
A single-site office of fifteen people with a clean fibre line and a handful of handsets can usually self-manage: the fixes in this guide will resolve most faults. A multi-site business, a contact centre, or anyone running voice over a contended broadband line alongside heavy cloud traffic is generally better served by a managed arrangement, because the failure modes span the LAN, the trunk and the carrier — and diagnosing across all three under time pressure is a specialist job. Our guide to what to look for in a managed IT support agreement covers how to hold a provider to a genuine quality SLA rather than a vague “best efforts” clause.
VoIP call quality readiness — where most UK businesses sit
Before diagnosing a specific fault it helps to see the common pattern. Across UK small and medium businesses, the network foundations for good voice are frequently half-built: the line is fast enough, but nothing prioritises voice, the router still has SIP ALG switched on, and no one has ever measured jitter. The grid below maps the usual state of play so you can spot which parts of your own setup are most likely to be the culprit.
VoIP call quality by the numbers — the thresholds that matter
These four figures are the goalposts. Measure your line against them first; every fix in this guide is aimed at pulling one of these back inside its target. They are drawn from ITU-T recommendations and standard carrier acceptance criteria, and they apply whether you run a hosted PBX, a SIP trunk into an on-premise system, or a cloud contact centre.
A useful rule of thumb: latency and jitter are usually WAN or QoS problems, while packet loss can live anywhere — a failing cable, an overloaded Wi-Fi cell, a saturated upload, or congestion out on the carrier network. That is why the diagnostic order in this guide moves outward from the handset to the trunk: you rule out the cheap, local causes before you start blaming the internet.
The most common causes of poor VoIP call quality
When we audit a business reporting bad calls, the root cause distribution is remarkably consistent. The chart below reflects the relative frequency with which each issue turns out to be the primary culprit — useful for deciding where to point your diagnostics first. Note how few cases are genuinely “the internet is too slow”; raw bandwidth is rarely the real problem.
The percentages overlap because faults compound — a line with no QoS and a thin upload will fail the moment someone starts a large file sync. But the ranking holds: the single highest-yield fix in UK offices is almost always turning on proper QoS, and the second is disabling SIP ALG. Together they resolve the majority of “jitter and latency” complaints without spending a penny on more bandwidth.
Network readiness benchmarks for clean voice
How does a voice-ready network actually score? These benchmarks reflect where a well-configured UK SME line sits after remediation versus the typical starting point. Use them as a target profile: the goal is to push every row toward the right before you conclude the problem is beyond your control.
Voice-readiness profile (post-remediation targets)
The last two rows — continuous monitoring and a failover path — are where even otherwise-tidy businesses lag. If you cannot see the numbers in real time you will always be diagnosing after the fact, and if a single fibre bearer carries all your voice you have no answer when it drops. Our comparison of leased lines versus business broadband covers how to build that resilience into the underlying connection.
The systematic diagnostic path — from handset to trunk
Random fiddling is how call-quality problems drag on for weeks. Work the path in order, outward from the desk, confirming or clearing each layer before moving to the next. The timeline below is the exact sequence we use on a triage call; each step either finds the fault or definitively rules that layer out.
Decision gauge — fix it yourself or escalate?
Not every fault is worth chasing in-house. This readiness gauge summarises how well-placed a typical business is to resolve voice problems without specialist help, based on the state of its monitoring, QoS and SIP knowledge. A low score is not a failure — it is a signal that escalation will be faster and cheaper than trial and error.
Score yourself honestly: award points for having real-time MOS visibility, a working QoS policy, a wired handset estate, and someone who understands SIP registration and RTP. Below about 50, escalate early; the diagnostic evidence a managed provider needs is exactly what a low-readiness business struggles to produce, and the fault will outlast your patience before it outlasts theirs.
What fixing VoIP call quality actually costs
Most of the highest-impact fixes cost nothing but time — QoS configuration and disabling SIP ALG are free settings changes. Costs appear only when the underlying network needs upgrading or when you bring in ongoing management. The table below sets out realistic UK 2026 figures so you can weigh a one-off spend against the recurring cost of degraded calls.
| Remediation | Typical UK cost | Fixes | When it is worth it |
|---|---|---|---|
| QoS configuration | £0 (or a few hours’ support) | Jitter, latency under load | Almost always — do this first |
| Business-grade router | £150–£600 | SIP ALG, NAT, weak queuing | Consumer router struggling with voice |
| Managed PoE switch | £250–£900 | Cabling, VLAN, PoE budget | Handsets share cabling with data |
| Upload/bandwidth upgrade | £10–£150 per month | Contention, thin upload | Calls collapse under real load |
| Leased line + failover | £200–£600 per month | Contention, resilience, SLA | Voice is business-critical |
| Managed VoIP support | £5–£12 per seat / month | End-to-end quality, monitoring | Multi-site or no in-house skills |
Frame the recurring numbers against the cost of poor calls: a sales line that drops customers, a support desk that has to ask people to repeat themselves, an office that quietly stops trusting the phones and reverts to mobiles. Those costs are real even though they never appear on an invoice, which is why the free fixes are the ones to exhaust before anyone signs off a spend.
How much of the problem is genuinely solvable?
The reassuring headline for anyone tempted to give up on VoIP: the vast majority of call-quality faults we see are network conditions inside your control or your provider’s, not inherent flaws in internet telephony. Very few cases turn out to be unfixable without changing carrier or platform.
The remaining slice — genuinely saturated last-mile infrastructure, a carrier with a persistent routing fault, or a location where fixed connectivity simply cannot deliver a stable path — is where a change of line or provider is warranted. But you reach that conclusion only after the systematic path above has cleared every other layer, not as a first reaction to a bad afternoon.
The VoIP call quality troubleshooting checklist
Work through this in order. It condenses the diagnostic path into concrete actions you can run today; each one either fixes the fault or removes a layer from suspicion. Print it, keep it by the comms cabinet, and record the reading at each step so you build the evidence trail an escalation will need.
- Capture a baseline: MOS, jitter, latency and packet loss from the portal or a line test, on an affected call.
- Classify the fault: one-way or two-way, constant or under load, one handset or all of them.
- Swap the handset and its patch cable; move it to a known-good switch port.
- Take voice off Wi-Fi and off any powerline adapters — wire every handset to PoE.
- Confirm QoS tags RTP as EF (DSCP 46) and the router enforces a priority queue for it.
- Disable SIP ALG on the router or firewall — then re-test.
- Open the RTP media port range (commonly 10000–20000 UDP) and set NAT keepalive to 30–60 seconds.
- Check for double NAT (two routers in series) and collapse it to a single NAT boundary.
- Measure upload headroom during a real backup or large sync; confirm calls survive peak load.
- Verify codec negotiation and registration stability on the SIP trunk; watch for re-registrations.
- Raise a carrier ticket with your captured figures if the trunk or carrier is implicated.
- Escalate to a managed VoIP specialist with the full evidence pack if the fault survives every step.
Re-test after every single change, not in batches. If you flip five settings at once and calls improve, you will never know which one mattered — and you will carry four unnecessary changes forward that may bite you later.
Real-world example — a Leeds firm that nearly ditched VoIP
A 40-person professional services firm in Leeds was on the verge of ripping out its hosted phone system after months of dropped calls and complaints about robotic audio on the sales line. Calls were fine first thing and fell apart by mid-morning. The instinct was to blame the platform; the reality was mundane. Their consumer-grade router still had SIP ALG enabled, no QoS policy existed, and the nightly cloud backup had crept into working hours after a schedule change — so every morning the backup saturated a thin upload just as the office got busy.
The fix took an afternoon and no new hardware beyond a business router: SIP ALG off, a QoS policy tagging voice as EF with a priority queue, the backup window moved back to overnight, and the handsets moved off a shared Wi-Fi extender onto wired PoE. Jitter dropped from the high 40s to under 15 ms and the dropped-call complaints stopped. The lesson was not that their VoIP was bad — it was that nothing had ever protected voice from the rest of the traffic.
We were a week from signing a contract to go back to traditional lines. It turned out the whole problem was a setting we did not know existed and a backup running at the wrong time. Two changes and the phones just worked again.
Stories like this are the norm, not the exception. The pattern — a fast-enough line, no prioritisation, and one greedy background task — accounts for a large share of the “VoIP is unreliable” verdicts we are asked to review. Our guide to network administration best practices covers building the QoS and change-control habits that stop this recurring.
Common VoIP call quality mistakes to avoid
These are the errors that turn a quick fix into a long saga. Most are assumptions — that bandwidth is the problem, that the platform is at fault, that a consumer router is “good enough”. Recognising them early saves days.
- Throwing bandwidth at a jitter problem. Adding a faster line does nothing if there is no QoS — voice still competes with everything else on the way out.
- Leaving SIP ALG enabled. This “helpful” router feature rewrites SIP packets and is a leading cause of one-way audio and dropped calls. Turn it off.
- Running handsets over Wi-Fi. Wireless adds jitter and loss that no jitter buffer fully hides. Wire them to PoE wherever possible.
- Ignoring the upload figure. Broadband is advertised on download; voice lives on the upload, which is often a fraction of it and the first thing to saturate.
- Changing five things at once. Without re-testing between changes you cannot attribute the fix, and you accumulate risky config.
- Blaming the carrier first. Only about one in five faults is genuinely carrier-side. Clear your own LAN, QoS and router before you raise a trunk ticket.
- No monitoring. If you cannot see MOS and jitter in real time, every diagnosis is a guess and every fix is unverifiable.
- Forgetting the PSTN switch-off deadline. With the UK migration to all-IP voice, putting off a proper VoIP setup only compresses the time you have to get it right.
Double NAT is the fault people find last and should find early. Two routers in series — often an ISP box behind another ISP box — breaks SIP registration and media in ways that look like random drops. Collapse to a single NAT boundary before you spend a day chasing the trunk.
At-a-glance summary
| Factor | Target / action |
|---|---|
| One-way latency | Under 150 ms (ITU-T G.114); aim under 100 ms |
| Jitter | Under 30 ms; aim under 20 ms sustained |
| Packet loss | Under 1%; aim under 0.5% |
| Target MOS | 4.0 or above for business calls |
| Highest-yield fix | QoS: tag RTP as EF (DSCP 46) + priority queue |
| Second fix | Disable SIP ALG on the router/firewall |
| Bandwidth per G.711 call | ~85–100 kbps each way, including overhead |
| RTP media ports | Open the range (often 10000–20000 UDP) |
| NAT keepalive | 30–60 seconds; avoid double NAT |
| Handset connection | Wired PoE, never Wi-Fi or powerline for voice |
| Diagnostic order | Handset → LAN → QoS → router → bandwidth → trunk |
| When to escalate | Fault survives every layer — hand over the evidence pack |
| Share solvable in-house | ~89% via network / configuration changes |
How Cloudswitched keeps your calls clean
Cloudswitched designs, tunes and manages VoIP for UK businesses so call quality is engineered in rather than debugged after the fact — the right QoS policy, a voice-ready router and switch estate, correctly configured SIP trunks, and continuous monitoring of jitter, latency and MOS so problems are caught before your team notices them. When a fault does appear, we work the same systematic path set out here, with the evidence and carrier relationships to resolve it quickly.
Struggling with jitter, latency or dropped calls?
Our specialists diagnose VoIP call quality end to end — LAN, router, QoS and SIP trunk — and tune it so the phones simply work.
Talk to a VoIP SpecialistFrequently Asked Questions
Why do my VoIP calls keep dropping?
Dropped calls almost always trace to one of a few network faults: SIP ALG rewriting packets on the router, a NAT keepalive that expires so registration lapses, double NAT breaking the media path, or packet loss on a saturated upload. Work the diagnostic path outward from the handset — disable SIP ALG, check for two routers in series, and measure upload under load. In most UK offices the cause is a configuration setting, not the VoIP platform itself.
What causes jitter and latency on VoIP?
Jitter — the uneven arrival of voice packets — and latency — the delay before the other person hears you — are usually caused by voice traffic competing with everything else on the line because no Quality of Service policy protects it. A backup, a large upload or heavy cloud use will spike both. Tagging voice as EF (DSCP 46) with a priority queue is the single most effective fix, well ahead of buying more bandwidth.
How much bandwidth does a VoIP call need?
A single G.711 call uses roughly 85–100 kbps in each direction once IP, UDP and RTP overhead are counted; the compressed G.729 codec drops that to around 30 kbps. The critical figure is upload, not download, and it must hold for every concurrent call at peak. Ten simultaneous G.711 calls need roughly 1 Mbps of reliable, prioritised upload — which is why thin or contended upload is such a common culprit.
What is QoS and do I really need it?
Quality of Service is a set of rules that tells your router to treat voice packets as more urgent than data, so a file transfer cannot delay your call. Yes, you need it: without QoS, voice is just another stream fighting for the line, and quality collapses the moment anything else gets busy. Proper QoS network configuration — tagging RTP as EF and giving it a priority queue — resolves the majority of jitter and latency complaints.
Should I disable SIP ALG on my router?
In almost all cases, yes. SIP ALG (Application Layer Gateway) tries to help SIP traffic cross NAT by rewriting packets, but its implementations are frequently buggy and cause one-way audio, failed registration and dropped calls. Most VoIP providers explicitly recommend turning it off and letting the phone system handle NAT traversal. It is one of the first settings to check when troubleshooting SIP trunking problems.
Why can I hear them but they cannot hear me (one-way audio)?
One-way audio is a classic NAT or firewall symptom: the media (RTP) stream is being blocked or misrouted in one direction. The usual causes are SIP ALG interfering, a closed RTP port range, or double NAT. Disable SIP ALG, open the RTP media ports (often 10000–20000 UDP), and confirm there is only one NAT boundary between the handsets and the internet.
Can I run VoIP over standard business broadband?
Yes, and many UK businesses do successfully — but broadband is a shared, contended, download-biased product, so it demands QoS and enough upload headroom to be reliable. For voice that is business-critical, a leased line with a symmetrical, uncontended bearer and an SLA removes the contention risk entirely. The right choice depends on how much a bad call costs you.
How do I measure VoIP call quality?
Most hosted VoIP portals report per-call MOS, jitter, latency and packet loss, which is the fastest route to a baseline. Free online line tests and tools such as a continuous ping to your SIP host will also reveal jitter and loss. The key is to measure on an affected call, under the conditions that trigger the fault, so you capture the problem rather than a quiet-line snapshot.
What is a good MOS score for business calls?
MOS (Mean Opinion Score) runs from 1 to 5. For business VoIP, aim for 4.0 or above; the practical ceiling for a compressed codec is around 4.3, and traditional PSTN quality sits near 4.4. Anything drifting toward 3.5 or below will be noticeably degraded. Watching MOS continuously, rather than only when someone complains, is what separates a proactive setup from a reactive one.
When should I escalate to a managed VoIP provider?
Escalate when the fault survives the full diagnostic path — handset, LAN, QoS, router and bandwidth all cleared — or when the problem spans multiple sites, a contact centre, or a SIP trunk you cannot see into. Hand over your captured measurements, packet captures and the timeline of changes; that evidence turns weeks of back-and-forth into a same-day resolution and lets a specialist hold the carrier to account.
Does the UK PSTN switch-off affect my VoIP plans?
It reinforces them. The UK is migrating all voice services to IP as the traditional PSTN and ISDN networks are withdrawn, so getting your VoIP setup properly configured now is time well spent rather than a job to defer. A clean, QoS-protected, well-monitored VoIP deployment is exactly what the all-IP future requires — the troubleshooting discipline in this guide is the same discipline that keeps an all-IP estate healthy.
Related reading
Make dropped calls and choppy audio a thing of the past
Cloudswitched tunes and manages VoIP for UK businesses end to end — QoS, routers, SIP trunks and continuous monitoring — so your team stops thinking about the phones.
Talk to a VoIP Specialist
CloudSwitched
London-based managed IT services provider offering support, cloud solutions and cybersecurity for SMEs.
