Measurement tool

Packet Loss Test

Forty probes, sent one after another, with every failure counted. This page is explicit about the difference between request loss and the packet loss a game or call would see.

Request loss test

40 probes. Counts requests that fail or time out — see the note below about UDP.

idle

Probes sent

—

Failed

—

Request loss

—

Median latency

—ms

Successful probes (ms)

A browser cannot measure UDP packet loss. This tool measures request loss, which is related but not the same thing. Why?

Ready. Nothing is measured until you start.

What a browser can and cannot measure

Packet loss is the percentage of individual packets that never arrive. Games and calls use UDP, where loss is not retransmitted — the packet is simply gone, and the receiver has to guess. Browsers have no access to UDP and cannot count those losses.

What a browser can do is send requests over a real path and count how many fail or time out. TCP retransmits lost packets, so a request usually still succeeds — but the round-trip time grows and, occasionally, the request fails outright. The figure here is therefore called request loss.

Why this is still useful

  • A path with zero failures over 40 probes is very unlikely to be dropping packets routinely.
  • Repeatable failures point at a real problem: interference, a saturated uplink, failing cabling or an unstable modem.
  • The latency chart shows the retransmission cost: successful requests that took far longer than the median are the visible symptom of lost packets being resent.

For a direct measurement of UDP behaviour you need a native tool — mtr, Pathping on Windows, or the ping utility against a host that answers ICMP. Those are worth running alongside this page if you suspect loss.

Common causes and what to try first

For home connections, wireless interference and over-subscribed uplinks cause most loss. Try, in order: connect over Ethernet, power-cycle modem and router, check cable condition, then test again at a different time of day. If loss persists wired, contact your ISP with the timestamps and the failure counts from this page.

Questions about this tool

Is this real packet loss?
It is request loss: how many HTTP probes failed. Browsers cannot count UDP packets, so this is the closest honest proxy available on the web.
What is an acceptable request loss rate?
Zero failures over 40 probes is the normal case on a healthy line. Isolated single failures can be transient timeouts; a rate above 1% is worth investigating.
Can packet loss be fixed?
Local causes — Wi-Fi interference, bad cables, a saturated uplink — are fixable. Loss inside the ISP network is not something you can fix yourself, but recording the evidence helps support staff.
How is this different from the bufferbloat test?
Bufferbloat measures delay under load; this page measures failures. A connection can have perfect latency and still lose packets, and vice versa.