Skip to main content
VoIPMetrics

Jitter test

Jitter is the reason calls sound robotic even when nothing is technically lost and your speed test looks perfect. It measures how unevenly audio arrives — and it is the early warning sign that shows up before packet loss does.

Measure your jitter

90 closely-spaced probes, about 6 seconds. Jitter needs a dense sample to show its shape, so this fires faster than our packet loss test.

Reading the chart

Each bar is one probe, drawn as its distance from your average response time. A perfectly steady connection produces a flat line along the center. Bars reaching up and down show packets arriving early and late — and that spread, not the height of any single bar, is what your jitter buffer has to absorb. Amber bars mark swings over 30 ms; red marks a probe that never came back at all.

Fix it in this order

  1. 1

    Re-test on a cable

    Wi-Fi is the cause more often than everything else combined. If jitter collapses on Ethernet, your internet service is fine and the work is wireless — position, channel congestion, or an access point too far from the desks that matter.

  2. 2

    Quiet the network and test again

    Backups, cloud sync, and video calls all make timing uneven. If a quiet network tests clean, you have a saturation problem rather than a quality one — and that is what QoS exists to solve.

  3. 3

    Turn on QoS for voice

    Business routers can prioritize voice packets ahead of everything else. For jitter specifically this is the highest-value setting available to you, and it costs nothing.

  4. 4

    Check the time of day

    Jitter that spikes at the same hour every day is congestion, not a fault. Test during your busiest period — that is when your calls actually happen, and a clean midnight result proves nothing.

How this test works

The test sends 90 closely-spaced probes to VoIPMetrics servers and measures the variation between consecutive round-trips — the same statistic a jitter buffer reacts to. It fires faster and takes more samples than our packet loss test, because jitter is a distribution rather than a count: you need dense consecutive timings to see its shape.

Probes travel over HTTPS, so this measures timing on the same path your calls would take. A WebRTC-based reflector measuring true media-path timing packet by packet is in development and will replace this engine when it ships.

Jitter questions

What is a good jitter score?

Under 30 ms for business calls, and under 10 ms is excellent. Jitter buffers absorb small variation invisibly, so anything in that range is genuinely fine. Between 30 and 50 ms the buffer compensates by holding audio longer, which adds delay; past 50 ms packets start arriving too late to use at all.

What causes high jitter?

Wi-Fi is the most common culprit by a wide margin — radio retransmissions make timing inherently uneven. After that: a connection saturated by backups, video, or large uploads, and routers that process traffic in whatever order it arrives instead of prioritizing voice. Jitter is nearly always a local problem you can fix, not something your internet provider needs to solve.

How do I fix jitter?

In order of value: run this test again on a wired connection — if jitter collapses, fix the Wi-Fi rather than the internet. Re-test with the network quiet to identify saturation. Turn on QoS in your router so voice traffic jumps ahead of downloads, which is the single highest-value setting for this specific problem. Persistent high jitter on a wired, quiet connection is your provider's to investigate.

Why does one big spike matter if my average is fine?

Because a jitter buffer is sized for typical variation, not exceptional ones. A single large spike arrives after the buffer has already played past that moment, so the packet is discarded and the caller hears a clipped word — even though the average looks healthy. That is why this test reports the worst swing alongside the average.

What's the difference between jitter and latency?

Latency is how long audio takes to arrive; jitter is how much that time varies from packet to packet. You can have low latency with terrible jitter, and that sounds worse than consistently moderate latency — human ears tolerate a steady delay far better than an unpredictable one.

Keep exploring