Why VoIP Call Quality Goes Bad (and How to Find the Culprit)

Start with what the calls sound like

The receptionist sounds like a robot. The customer heard every third word. Calls drop at ten in the morning, or the audio only works one way, or there’s an echo that wasn’t there last month.

Every one of those is a specific, diagnosable symptom, and each points at a different culprit. The phone system itself causes almost none of them. When a hosted phone system misbehaves, the cause is usually the path the call travels: your Wi-Fi, your router, your cabling, your internet connection, or occasionally your provider’s carrier. The hosted VoIP explainer makes this point in one line: a hosted system depends on the same data path your computers use. This post is the long version, written so you can find your actual culprit instead of rebooting phones and hoping.

A phone call is a stream, not a download

The thing that makes voice different from everything else on your network: a call is a live stream of small packets, sent continuously, that all have to arrive on time and in order.

A standard-quality call (the G.711 codec most desk-phone systems still default to) sends 50 packets per second and uses roughly 90 Kbps of bandwidth in each direction, headers included.¹ That’s nothing. A 10 MB email attachment moves about as much data as seven minutes of talking.

Most of what your network carries is tolerant of delay. A web page that loads 200 milliseconds late is still the same web page. An email that arrives out of order gets reassembled and nobody knows. Voice has no such patience. The receiving phone throws away any packet that arrives late, because the moment it described is already over. There is no re-sending a syllable. Video meetings and remote desktop sessions live under the same rules, which is why an office with garbled calls usually has choppy Zoom and laggy remote sessions too.

So when someone says “our internet is fast, the phones should be fine,” both halves can be true and unrelated. Speed is how much the pipe moves. Call quality is whether the pipe moves a trickle of packets on a tight schedule. Different properties, different failures, and a speed test only measures the first one.

The four numbers that decide every call

Every call-quality complaint traces back to one of four measurements. Each has a target and a distinct sound when it goes wrong.

Two details worth pausing on.

Jitter is the one nobody’s heard of and the one we find most often. Your phone keeps a small buffer to smooth out irregular arrivals. When packets arrive erratically enough to overflow that buffer, your phone stitches the audio back together with pieces missing, and the voice on the other end turns metallic. Wi-Fi is the leading manufacturer of jitter in small offices.

Packet loss is the most misleading, because 1% sounds like a rounding error. But loss rarely spreads itself out politely. It arrives in bursts: forty packets gone in one moment of router congestion, which is most of a second of someone’s sentence. That’s why a connection can test at “99% delivered” and still produce calls where the customer says you keep cutting out.

Where the problem usually lives

You don’t need to become a network engineer. You need to know the five or so places these failures hide, in roughly the order we find them.

Wi-Fi carrying calls that should be wired. The most common culprit we see. Wireless is a shared medium: every laptop, phone, tablet, and smart TV in range competes for airtime, and the retransmissions that make Wi-Fi feel reliable for web browsing are what create jitter for voice. Desk phones belong on ethernet. Softphones and mobile apps are legitimately useful (sometimes they’re the whole product), but they inherit the airwaves they ride on.

The router your internet provider handed you. Consumer-grade routers treat every packet identically. Your call and a YouTube video are the same to them, first come, first served. The fix is QoS (quality of service), which is the router recognizing voice packets and letting them skip the line. More on that below, because it’s the one setting in this post that does the most work.

A saturated upload. Business internet is usually asymmetric, especially on cable: fast down, slow up. Your voice leaves through the upload side, and so does the cloud backup that kicks off mid-morning, the file sync, the security camera feed. When the upload pipe fills, outbound voice packets queue behind bulk data and arrive late or not at all. This is the classic “calls are only bad at certain times” pattern, and it’s invisible in a speed test run at 7 a.m.

Physical-layer decay. A patch cable with a damaged conductor, a switch port going bad, a cable run pinched behind a filing cabinet years ago. These produce packet loss that no configuration change will fix, and they’re why “have you tried a different cable” is a real diagnostic step and not tech-support theater.

The path beyond your building. Sometimes the problem is the internet provider’s line or a congested route between them and the phone platform. This is the least common culprit and the first one everybody blames. It’s also the only one on this list you can’t fix yourself, which is why you rule out the other four before spending an afternoon on hold with the carrier.

QoS, in one paragraph

QoS is a rule in your router that says: voice goes first. When the connection is busy, voice packets jump ahead of the backups and file syncs leaving your office, whose extra half-second of delay no one will ever notice. A router can only prioritize what it sends, so inbound congestion is a different fight, which is one more reason the upload side decides more calls than the download side does. A network with correctly configured QoS can carry clean calls on a modest connection. A fast connection without QoS produces bad calls every time the pipe gets briefly full. The catch: it has to be configured, on equipment capable of it, by someone who knows what voice traffic looks like. It’s off by default, and the router in your closet right now probably isn’t doing it. It’s included setup work on any competent installation, not an upsell, and it’s a fair question to ask any provider: “what QoS configuration do you do during rollout?” A specific answer means they’ve done this before. A pause means voice will wait in line with everything else.

Test it before you blame anything

All of the above is theory until you measure your own network. We built a free VoIP network test that runs from your browser, in your office, and measures the numbers in the table: latency, jitter, packet loss, and up/down bandwidth, plus an estimated call-quality score. It takes about two minutes, and you don’t need to hand over an email address to see your results.

One note on reading the result: a browser test, ours included, measures latency as round-trip time, so compare it against roughly double the one-way target in the table, or lean on the verdict the test gives you. Jitter and packet loss read the same either way.

Three ways to make the result mean something:

1. Run it during your worst hour, not your quietest. If calls fall apart mid-morning, a 7 a.m. test will cheerfully tell you everything is fine. Test when the network is doing what it does when calls are bad.
2. Run it twice: once wired, once on Wi-Fi. If the wired result is clean and the wireless one isn’t, you’ve just diagnosed the most common problem in small-office VoIP without opening a ticket.
3. Save the result. A screenshot of jitter and packet-loss numbers turns “our calls sound bad” into evidence. That’s useful whether you send it to us, your current provider, or your internet carrier.

Good answer: “Before we quote anything, we test your network during business hours and walk you through what we find.” Any provider worth shortlisting should be willing to say this, us included.

Red flag: A provider gives you a firm price before testing anything about your network. The calls will ride that network. Someone should want to see it first.

What “managed” means for call quality specifically

This is the part where we describe how we handle it, so you can compare it against whoever you’re evaluating. The fully-managed post covers the whole service model; this is the call-quality slice of it.

Before installation, a network review is part of the rollout’s first week: what the internet connection is, what the router can and can’t do, what’s wired and what isn’t, where QoS needs to be configured. Quality problems are cheapest to fix before the phones arrive, and most of the ugly surprises in VoIP deployments are just skipped homework from this phase.

After installation, call-quality issues come through the same support model as everything else. You call or email, a human takes the intake, and an engineer who can read a jitter graph calls you back, typically within 30 to 60 minutes during business hours. Since we run the phone platform, we can see call metrics from our side and correlate them with what your network test shows from yours. We usually know which of the five culprits it is by the end of that first callback. The configuration work (QoS, codec settings, call routing) is part of the service; when the fix is physical, like a new router or a cable run, it’s quoted separately, and you’ll know before anything is ordered.

And when the data says the problem is upstream (your carrier’s line, not your network and not our platform), we say so. We can’t fix another company’s wiring. What we own is the evidence: call metrics from the platform side, your test results from the office side, and a carrier ticket written so it can’t bounce. That’s what keeps you from getting stuck between two vendors each pointing at the other, which is where unmanaged VoIP customers spend a depressing amount of their lives.

Start with the one-minute version

If your calls are bad right now, or you’re pricing a new system and want to know whether your network is ready for one: run the network test. Wired and wireless, during a busy hour. It’s the same first step we’d take anyway.

Then, if you want someone to read the results with you, call us at 844-767-1924 or send us the details. We’ll tell you what the numbers mean and what we’d fix, in plain terms, whether or not you become a customer.

Sources