Packet loss test
Packet loss is the #1 killer of VoIP call quality — it's what turns conversations into “you're breaking up.” This test probes your connection and grades it against the thresholds business calls actually require.
Test your connection now
60 probes in about 10 seconds. No signup, nothing installed — runs entirely in your browser.
Reading your results
| Metric | Good | Borderline | Problem |
|---|---|---|---|
| Packet loss | < 1% | 1–3% | > 3% |
| Response time (delay) | < 150 ms | 150–300 ms | > 300 ms |
| Jitter | < 30 ms | 30–50 ms | > 50 ms |
These cutoffs come from international telecom standards and what we see in the field. A call can survive one bad number; two or more and callers will hear it.
Seeing loss? Fix it in this order
- 1
Go wired for the test
Re-run on Ethernet. If the numbers clean up, your internet is fine — the problem is Wi-Fi interference or coverage, which is fixable without calling your ISP.
- 2
Look for saturation
Backups, video streams, and large uploads fill your link and force routers to drop packets. Re-test with the network quiet to isolate it.
- 3
Turn on QoS for voice
Business routers can prioritize voice traffic so a big download never starves a phone call. This is the single highest-value router setting for VoIP.
- 4
Check equipment and cabling
Aging routers, failing cable modems, and damaged Ethernet runs all manifest as intermittent loss. A modem's signal-level page often shows it directly.
- 5
Escalate to your ISP with data
Persistent loss on a wired, quiet connection is upstream. Run this test several times at different hours and bring the numbers — reports with data get escalated, vague complaints get scripted answers.
How this test works
The test sends 60 timed probes to VoIPMetrics servers over your active connection, discarding the first three (connection setup skews them). It reports the failed-probe rate as loss, the average round-trip time, and jitter as the average variation between consecutive probes — the same statistic VoIP jitter buffers care about.
Probes run over HTTPS, so this measures request-level loss and timing on the same path your calls would take — a congested or lossy link fails probes and voice packets alike. A WebRTC-based reflector that measures true UDP media-path loss, packet by packet, is in development and will replace this engine when it ships.
Packet loss questions
What is an acceptable packet loss for VoIP?
Under 1% for business-quality calls. Voice codecs conceal a rare lost packet, but at 1–3% callers start hearing artifacts and clipped words, and above 3% conversations become genuinely hard to follow. Zero measurable loss is the right target on a wired business connection.
What's the difference between packet loss and jitter?
Packet loss is audio that never arrives — packets dropped somewhere on the path, heard as gaps and robotic stretches. Jitter is audio that arrives at uneven intervals; buffers absorb small amounts, but past roughly 30 ms of variation the buffer either delays the call or discards late packets — which becomes loss. High jitter is usually the early warning for the congestion that causes loss.
Why do my results change between runs?
Because your network conditions change minute to minute — a cloud backup kicks off, someone starts streaming, Wi-Fi interference spikes. One clean run doesn't prove much; run the test a few times, including during your busiest hours, since that's when your phones need the connection most.
Does Wi-Fi cause packet loss?
It's the most common culprit we see. Radio interference, distance from the access point, and channel congestion all drop or delay packets before your traffic ever reaches the internet. If your Wi-Fi results are poor, re-run the test on a wired connection — if the numbers clean up, the fix is your wireless setup, not your internet provider.