Your speed test says 400 Mbps and your Zoom call still turns you into a robot. Speed tells you how much fits through the pipe. Packet loss tells you how much falls out of it, and a speed test barely notices.
Our view: one packet loss number from one browser test tells you almost nothing. You don't know where the packets went missing, so you don't know who has to fix it. Do two runs instead, one on Wi-Fi and one wired at the router, and the number becomes a diagnosis. It's also evidence an ISP support agent can't brush off with "have you tried restarting your router?"
How do I test for packet loss in my browser, and is the result accurate?
The test takes about a minute. The setup is what makes the number worth anything.
- Close video calls, cloud sync and game launchers, and pause any downloads on every device in the house.
- Open a browser test that reports loss. As of 2026, Cloudflare's speed test has shown a packet loss figure, and some WebRTC-based testers are built only for loss.
- Let it run to the end without touching anything.
- Write down the loss percentage, plus the latency if the test shows it.
How it works: the test sends a stream of small UDP packets over a WebRTC data channel to a single server and counts how many never arrive. It has to use WebRTC because ordinary web traffic runs over TCP, and TCP hides loss by quietly sending missing packets again.
Is it accurate? It's good at catching steady loss between you and one nearby server. But it covers a few seconds, on one route, at one moment. Think of it as a snapshot, not a ruling on your line. Treat a single result as a clue.
What percentage of packet loss is bad?
Anything above 0% on a wired connection is something to fix. A healthy wired line delivers every packet it sends.
The vendors set the bar low. As of 2026, Microsoft's Teams network guidance has asked for packet loss under about 1% in a 15-second window. As of 2026, Zoom has recommended packet loss of about 2% or less. These are the targets each vendor designs for. They aren't cliffs: at 1.1% a Teams call won't suddenly drop.
How the loss arrives matters as much as the average. A burst where 5% of packets vanish for one second wrecks a sentence or a gunfight. A steady 0.3% spread evenly is usually smoothed over by the app. A test that averages over 30 seconds can make a bursty line look fine, which is one more reason to run it several times.
The packet loss levels that matter
- 0%What a healthy wired line delivers
- Under 1%Microsoft Teams target in any 15-second window
- 2%Zoom's recommended maximum
- Over 5%Calls drop and games become unplayable
What each packet loss level feels like
| Loss | Zoom/Teams calls | Competitive gaming | Streaming |
|---|---|---|---|
| 0% | Clean audio and video | Smooth | Smooth |
| Under 1% | Within vendor targets, rare glitch | Occasional hitch | No visible effect |
| 1-2.5% | Robotic moments, brief freezes | Rubber-banding, missed hits | Rare buffering |
| 2.5-5% | Frequent freezes, choppy audio | Hard to play competitively | Quality drops |
| Over 5% | Calls break up or drop | Unplayable | Frequent buffering |
Microsoft Teams and Zoom network requirement pages, as of 2026; gaming and streaming effects are editorial estimates.
How much packet loss makes Zoom, Teams or a game fall apart?
Calls fail gradually. As of 2026, our estimate is that around 1-2% loss, voices start to turn metallic and words drop out. As of 2026, our estimate is that past about 5% loss, video freezes and calls often drop. To watch it happen live in Zoom, open the Statistics panel in the settings during a meeting. It shows packet loss and jitter for the audio and video streams.
Games are harsher. As of 2026, fast online shooters such as Fortnite and Valorant generally send game data over UDP, which does not resend lost packets. As of 2026, our estimate is that even 1% loss can show up as rubber-banding, missed shots or the game's packet loss warning icon. Most of these games have an on-screen network stats option in their settings, and it measures to the actual game server, not a test server.
Loss is only one of three numbers to read:
- Latency (ping): as an editorial estimate as of 2026, under about 50 ms feels sharp in shooters, and over about 100 ms puts you behind. As of 2026, Teams guidance has asked for round-trip latency under about 100 ms.
- Jitter: the variation in latency. As of 2026, Teams guidance has asked for jitter under about 30 ms, and Zoom has recommended about 40 ms or less. High jitter feels just like loss, because late packets get thrown away.
- Loss: the hard failure. A packet that never arrives can't be smoothed over.
Streaming is the forgiving one.
Why does a speed test show fast speeds but I still get packet loss?
A speed test downloads over many TCP connections at once. When a packet is lost, TCP notices and sends it again. The retry costs a few milliseconds, the transfer finishes, and the result looks great. What the speed test measures is throughput. It doesn't measure loss.
A call or a game works the other way. It sends its packets over UDP, because a voice packet that arrives 300 ms late is useless. When a UDP packet is lost, nobody sends it again. It's just gone, and you notice the gap: a clipped syllable, a frozen face, your character snapping back three metres.
So a fast line can still drop packets. The usual causes:
- A weak Wi-Fi signal, two rooms and a wall away from the router
- Interference from neighbouring networks on the same channel
- A full upload, for example someone backing up photos or streaming on Twitch
- A failing Ethernet cable, a loose coax connector or a noisy line into the house
How do I tell if packet loss is from my Wi-Fi, my router or my ISP?
Each run cuts out part of the path. That's the whole trick.
- Run A: your laptop or console in your usual spot, on Wi-Fi, as you normally use it.
- Run B: the same device on an Ethernet cable plugged straight into the router, with other devices paused.
Do each run three times and write down the worst and the typical result. Then read the verdict:
- A shows loss, B is clean (0%): your Wi-Fi. Range, walls, interference or a crowded channel.
- A and B both show loss: the router or the line from your ISP. Wi-Fi is off the hook.
- A and B both clean, but the app still lags: the remote server or the route to it. Try another test server, then run MTR to the game or Zoom server.
To split router from ISP: reboot the router and modem, swap the Ethernet cable, and if you can, plug the computer straight into the modem or ONT with the router taken out. Many cable modems also have a status page that shows signal levels and error counts. If loss is still there with the computer plugged into the modem, the problem is on the line, and the line belongs to your ISP.
The interpreter below follows the same logic. Enter your Wi-Fi loss and your wired loss and say whether you mostly make calls or play games. It names the likely fault, rates the severity against the thresholds above, and gives you the next step.
Say you get 3.2% on Wi-Fi from the sofa and 0% wired at the router. That's a Wi-Fi problem, and at 3.2% it's bad enough to wreck competitive games. Calling your ISP won't help. Now say you get 2.8% on Wi-Fi and 2.5% wired. The Wi-Fi is fine, and the line is losing packets before they ever reach your network.
The two-run packet loss diagnosis
- Run A: Wi-FiTest three times from your usual spot
- Run B: wiredTest three times on Ethernet at the router
- A bad, B cleanFix your Wi-Fi
- Both badCheck router and modem, then call the ISP
- Both clean, app lagsTest the route to the game or Zoom server
What can't a browser packet loss test detect?
Picture the path your packets take, one stretch at a time: your device, the Wi-Fi link, the router, the modem, the ISP line out of your home, the wider internet, and finally the Zoom or game server. Loss can start on any of those stretches.
Each tool sees a different part of that path:
- Browser test: device to one nearby test server. It doesn't see the stretch to your actual game or Zoom server.
- Ping: device to any single address you pick, over as long as you let it run.
- Traceroute: lists every hop to a server, though its per-hop loss figures aren't reliable.
- MTR: every hop to a server, with loss and latency per hop over minutes.
- In-app stats (Zoom Statistics, in-game netgraph): the full path to the real server, during real use.
So a browser test can't tell you which hop drops packets. It also misses loss that only happens at certain hours, and loss in one direction only, such as uploads failing while downloads stay clean.
Two tests disagree because they use different servers and routes, run for different lengths of time, and count late packets differently. Ping also uses ICMP, which busy routers deliberately put at the back of the queue. That's why a traceroute hop can show 40% "loss" while the hops after it show none. That hop is ignoring pings, not dropping your traffic.
Is 1% normal on Wi-Fi? It's common on a crowded or distant connection, but it isn't healthy. On a wired connection it should be 0%.
How long should I run a packet loss test, and when should I run it?
Three short browser tests are the minimum. A longer run tells you far more: 5 to 10 minutes of ping or MTR to the server you actually care about. On Windows, ping -n 600 sends one ping a second for ten minutes. On macOS or Linux, use ping -c 600.
Test when it hurts. If the line is clean at 10 am and losing packets at 9 pm, the cause is likely evening congestion, which in our estimate as of 2026 typically peaks between about 7 and 11 pm. Test at both times so you can show the pattern.
Keep a log with one line per run: date and time, location, Wi-Fi or wired, loss %, latency and jitter. Two evenings of that log is worth more than any single screenshot.
What should I tell my ISP, and how can I fix packet loss on Wi-Fi?
Bring numbers from the wired run. A Wi-Fi result just gets you the "move closer to the router" script.
Checklist before you call:
- Wired results taken at the router, at two or more times of day, with three runs at each time
- The servers you tested against (at least two different ones)
- The time pattern, for example "clean mornings, 2-4% loss after 8 pm"
- Your router and modem make and model, plus any error counts from the modem status page
- What you've already tried: rebooting, swapping the cable, testing with one device only
- Screenshots of Zoom Statistics or in-game stats during the problem
Open with one sentence: "I'm seeing wired packet loss of 3% to several servers, every evening, for the past week." That skips the first page of the support script.
If the loss is only on Wi-Fi, fix it in this order:
- Move closer, or move the router into the open, up off the floor and away from the TV.
- Use the 5 GHz or 6 GHz band rather than 2.4 GHz.
- Change the Wi-Fi channel to get away from your neighbours.
- Update the router firmware.
- Run a cable to the gaming PC or console.
- Add a mesh system, with wired backhaul if you can.
If the ISP doesn't fix it, escalate. As of 2026, in the UK, after about eight weeks, or sooner with a deadlock letter, Ofcom's rules let you take a complaint to an approved ADR scheme for free. As of 2026, in the US, you can file an informal complaint with the FCC.
It's your ISP's problem to fix, and your log proves it.
| Loss | Zoom/Teams calls | Competitive gaming | Streaming |
|---|---|---|---|
| 0% | Clean audio and video | Smooth | Smooth |
| Under 1% | Within vendor targets, rare glitch | Occasional hitch | No visible effect (editorial estimate as of 2026) |
| 1-2.5% | Robotic moments, brief freezes | Rubber-banding, missed hits | Rare buffering (editorial estimate as of 2026) |
| 2.5-5% | Frequent freezes, choppy audio | Hard to play competitively | Quality drops (editorial estimate as of 2026) |
| Over 5% | Calls break up or drop | Unplayable | Frequent buffering (editorial estimate as of 2026) |
| Tool | Can measure | Cannot measure |
|---|---|---|
| Browser packet loss test | Loss to one test server over UDP, in seconds | Path to your game or call server, which hop drops, time-of-day patterns |
| Ping | Loss and latency to one address over time | Where on the path loss starts |
| Traceroute | The hops on the route | Reliable loss per hop |
| MTR | Loss and latency per hop, over minutes | Loss in app-specific UDP traffic |
| Zoom Statistics | Loss, latency, jitter during a real call | Cause of the loss |
| In-game netgraph | Loss and ping to the game server | Cause of the loss |
Frequently asked questions
Can a VPN fix packet loss?
Sometimes, and only in one case: when the loss sits on a congested route between your ISP and the game or call server, a VPN can send your traffic around it. It can't fix loss on your Wi-Fi, your router or the line into your home, and it adds a little latency.
Can a bad Ethernet cable cause packet loss?
Yes. A damaged or loose cable can drop packets or fall back to a slower link speed. If your wired test shows loss, swap in a short cable you know is good before you blame the ISP.
Does packet loss on a mobile hotspot mean the same thing?
The thresholds are the same, but the two-run method doesn't apply, because there's no wired path to compare against. Test in two locations instead, since mobile loss usually comes down to signal and how busy the cell is.