VoIP latency and jitter: acceptable limits

For business calls that sound professional, keep one-way latency under 150 milliseconds, jitter under 30 milliseconds, and packet loss under 1 percent. The 150 ms figure is the formal one: ITU-T Recommendation G.114 sets it as the upper limit for high-quality real-time voice, and treats delay approaching 400 ms as generally unacceptable. Miss any one of the three and callers hear it, even when your speed test looks fine.

That last point is the one that wastes the most time. A business runs a speed test, sees 300 Mbps down, and concludes the network cannot be the problem. Bandwidth is rarely the problem. A single call in the standard G.711 codec uses about 80 kbps including packet headers, so a 300 Mbps connection has thousands of times the raw capacity a small office needs. What breaks calls is not how much data the line can carry, it is whether small packets arrive on time, evenly spaced, and in one piece.

The three numbers, and where they come from

Two of those thresholds are engineering design targets rather than standards, and it is worth being clear about which is which. The 150 ms latency limit is a published ITU recommendation. The jitter and packet loss figures are the targets network engineers design to, and the numbers behind them come from what the technology can actually compensate for: a jitter buffer can smooth out variation up to roughly 100 ms, and the G.729 codec starts producing audible errors well before loss reaches a full percent.

What is acceptable jitter for VoIP?

Under 30 milliseconds is the working target for business calls, and under 20 ms is comfortable. Jitter is the variation in arrival time between packets, not the delay itself. Voice packets leave in an even stream roughly every 20 ms, and if they arrive uneven, the receiving device has to guess how to reassemble them. Above about 100 ms of variation, the jitter buffer can no longer compensate and callers hear gaps and robotic artifacts.

This is why jitter, rather than latency, is the usual culprit behind the complaint that calls are "choppy". Latency that is high but steady mostly makes people talk over each other. Jitter is what turns words into fragments.

What is the maximum latency for VoIP?

ITU-T G.114 recommends no more than 150 milliseconds of one-way delay for high-quality real-time voice, and describes delay approaching 400 ms as generally unacceptable. For international calls, particularly over satellite links, up to 300 ms is treated as workable. Note that these are one-way figures: most consumer test tools report round-trip time, so a 200 ms ping is roughly 100 ms each way and is within the limit.

Below 150 ms, a conversation feels normal. Between 150 and 300 ms, people start unconsciously interrupting each other, because the natural pause that signals "your turn" arrives late. Past 400 ms it stops feeling like a conversation and starts feeling like a radio handover.

How much packet loss is acceptable for VoIP?

Under 1 percent, and ideally zero. Cisco's design guidance is blunt about it: the G.729 codec requires packet loss far below 1 percent to avoid audible errors. Unlike a file download, a voice call cannot ask for a lost packet again, because by the time it arrived it would be too late to play. Lost audio is simply gone, and the codec conceals the gap as best it can.

Small amounts of loss sound like clipped word endings, which people usually blame on the other person's phone. Loss above about 3 percent is unmistakable and makes a line unusable for anything that matters.

What causes jitter and latency on a business network?

In small offices the cause is almost always one of five things, roughly in order of how often it turns out to be the answer:

• Wi-Fi. By far the most common. Wi-Fi shares airtime, and a laptop syncing a large folder can starve a phone that needs a tiny packet every 20 ms. Wired connections eliminate more call quality problems than any other single change.
• No QoS on the router. Without traffic prioritization, voice packets queue behind video uploads, backups and software updates. Every business-grade router can prioritize voice; most are shipped without it configured.
• Upload saturation. Calls need upstream capacity, and most business broadband is asymmetric. A cloud backup running at full tilt during business hours will wreck calls while the download speed test still looks perfect.
• Old or cheap network hardware. Consumer routers with small buffers introduce jitter under load, and a failing cable or switch port causes intermittent loss that shows up as calls being fine in the morning and bad after lunch.
• Distance and routing. Physical distance to the provider's nearest point of presence sets a floor on latency that no amount of local tuning removes.

How do you test VoIP latency and jitter?

Use a VoIP-specific test rather than a generic speed test, because the metrics you need are the ones a speed test does not report. A proper test sends a stream of small packets shaped like a real call and reports latency, jitter, packet loss and a mean opinion score.

The important part is not running the test, it is running it repeatedly. Test from a wired connection and again over Wi-Fi, and test at 9am, at midday and mid-afternoon. Call quality problems are usually load problems, and a single clean test at 7pm proves nothing about how the line behaves when twelve people are working. If the results are good on cable and bad on Wi-Fi, you have found it. If they degrade at a predictable time each day, look for a scheduled backup or sync job.

It is also worth knowing that most businesses find out about a call quality problem the slow way. Customers rarely call back to report that a line sounded bad; they say so where other people can read it, which is why watching for complaints about your business appearing publicly often surfaces a phone problem days before anyone internally notices one.

How do you fix VoIP latency and jitter?

Work through these in order. The first two solve the large majority of cases in a small office.

1. Put the phones on cable. Move desk phones and heavy call users to wired Ethernet. If a wired connection is genuinely impossible, put voice devices on a dedicated 5 GHz network with as few other devices on it as you can manage.
2. Turn on QoS and prioritize voice. Configure the router to give voice traffic priority over everything else. This costs nothing and is the single most effective change after cabling.
3. Move bulk transfers out of business hours. Schedule backups, large syncs and system updates for the evening. Cap their bandwidth if the tool allows it.
4. Check upload, not download. Confirm you have enough upstream capacity for your concurrent call count with headroom, and remember that 80 kbps per call means the requirement is small but must be genuinely available.
5. Replace the weak link. Swap suspect cables, test a different switch port, and replace consumer-grade routers with business equipment that has proper buffering.
6. Escalate to the ISP with evidence. If wired tests still show high jitter or loss with the network otherwise idle, the problem is upstream. Bring timestamped test results rather than a description.

If you want the capacity side of this worked through properly, including how many concurrent calls a given connection supports and what headroom to leave, our guide to how much internet speed VoIP needs covers the arithmetic.

When it is the provider, not your network

Sometimes the network is clean and the calls are still bad. The signs are reasonably clear: wired tests come back well inside the thresholds, the problem affects calls to particular destinations rather than all of them, or it appears at the same time for everyone regardless of what they are doing. In those cases you are looking at the provider's own network, their carrier interconnects, or the codec they have selected.

This is a fair question to put to any vendor before you sign, and the answer tells you something. Ask what codecs they use, whether they publish network status, and what their support process looks like when a customer reports quality problems. A provider who treats call quality as your problem alone is telling you how the relationship will go. If you are still choosing, our comparison of business phone service providers covers what to ask, and the Ooma vs RingCentral comparison shows how differently two vendors can package the same underlying service.

A realistic standard to hold yourself to

Do not chase perfection. Aim for one-way latency comfortably under 150 ms, jitter in the teens or twenties, and packet loss that rounds to zero on a wired connection during a busy hour. A network that hits those numbers under load will carry business calls that sound like a phone call rather than a video conference, and it will keep doing so as you add people.

The reason to measure at all is that call quality problems get blamed on the wrong thing for months. Staff assume the phone system is bad, the phone system gets replaced, and the new one sounds exactly the same because the office Wi-Fi was always the culprit. Fifteen minutes of testing before you switch vendors is the cheapest diagnostic available, and it occasionally saves you the cost of switching at all. If the numbers are clean and the calls are still poor, then the system itself is worth changing, and a VoIP phone service built for business call volumes is the right place to look.