How to Read a Traceroute and Find Packet Loss
Use visual traceroute to identify network bottlenecks, latency spikes, and routing issues.
By the My IP Help team Last updated: Nov 8, 2026
Chapters
Chapters
To read a traceroute, go down the hops in order and watch two things: the latency, and whether replies go missing. A delay or loss that starts at one hop and carries on to every hop after it, the destination included, is real and starts there. A slow or silent hop followed by normal ones is just a router that does not bother answering probes. Everything else in the output, the addresses, the names and the places, tells you whose network each stretch belongs to.
This guide walks through a trace line by line, shows how to tell real packet loss from harmless gaps, and how to work out whether the problem sits on your network, your provider’s or the destination’s.
Why a Traceroute Is Easy to Misread
Every packet carries a Time to Live (TTL) value, and each router on the path lowers it by one. When it reaches zero, the router drops the packet and sends back an ICMP “Time Exceeded” message (RFC 792). Traceroute sends probes with a TTL of 1, then 2, then 3, and records which router answers each and how long the reply took.

The catch is in that reply. A router’s main job is forwarding traffic, which it does in fast hardware. Answering a probe addressed to the router itself is extra work, handled by its slower general-purpose processor, often with a rate limit. So the time and the loss you see at a hop describe how that router treats probes, not necessarily how it treats your traffic. Most misreadings of a traceroute come from forgetting this.
Step 1: Run the Trace From the Right Place
A trace shows the path from where it runs. To test your own connection, run it on your machine: tracert example.com in the Windows Command Prompt, traceroute example.com in a macOS or Linux terminal. To see how the rest of the internet reaches a host, or when you have no terminal at hand, run a traceroute online: our tool traces from our server to any public address and labels each hop with its network and place on a map.
Running both is the quickest way to split a problem in two. If the trace from our server reaches the host cleanly and yours does not, the trouble is on your side of the internet. If both struggle at the same network, it is that network’s.
Step 2: Read Each Line
Each line of the output is one hop. From left to right:
- Hop number: the position on the path, starting at 1, the first router after your machine (usually your home router).
- Name and address: the router that answered. Names often hint at the operator and city;
ae-5.r02.nycmny01.us.bb.gin.ntt.net, for example, reads as an NTT backbone router in New York. Add-n(or-don Windows) to skip the name lookups and speed up the trace. - Three times: the round-trip time of each of three probes, in milliseconds. Similar values mean a steady measurement; one outlier is usually a momentary blip. (Our online tool shows one averaged time per hop and no names.)
- Asterisks: a
*in place of a time means that probe got no reply within the wait time.* * *means none of the three did.
To find out who runs a hop, paste its address into the IP lookup or the ASN lookup. Treat the city it gives with care: geolocation databases often place a router where its network is registered, not where it stands. The latency is the honest signal, since light in fibre travels about 200 km per millisecond, so every 1 ms of round trip allows at most about 100 km of distance.
Step 3: Follow the Latency
Latency should rise gently along the path and jump only where the distance does. Read it as a trend, not hop by hop:
- A jump that stays: hop 6 goes from 20 ms to 150 ms and every later hop stays near 150 ms. Something between hops 5 and 6 adds the delay: a long link, a congested one, or a detour.
- A spike that recovers: one hop is slow and the next ones are fast again. That router was slow to answer probes; your traffic was not delayed there. In a trace from our server to 1.1.1.1 on 8 November 2026, hop 10 answered in 5.11 ms and hop 11, further along the same path, in 1.77 ms.
- A jump at an ocean: New York and London are about 5,600 km apart, so a round trip in fibre cannot take less than about 56 ms, and real cable routes are longer. A big step at a transatlantic hop is physics, not a fault. TeleGeography’s Submarine Cable Map shows where those cables land.
Step 4: Find Packet Loss
Packet loss in a traceroute follows the same rule as latency: only loss that carries on to the end counts.
- Loss at one hop, none after it: the router limits how often it answers probes. In the same trace to 1.1.1.1, hop 6 showed only asterisks and hops 7 to 12 answered normally. Ignore it.
- Loss that starts at a hop and continues to the destination: real loss, beginning at that hop or on the link into it. If hop 8 loses 20% and hops 9 to 14 all lose about 20% too, look at hop 8’s network.
- Loss only at the last hop: often the destination or its firewall dropping probes, not a fault. Check the service the way it is actually used, for example whether its port answers, with the port checker.
- Asterisks from one hop to the end: either the path is broken from there, or everything beyond it filters probes. If the service itself works, it is the filtering.
One plain traceroute cannot measure loss. With three probes per hop, a single lost probe reads as 33%, which means nothing. For loss, use a tool that sends 100 probes per hop. On macOS or Linux, mtr -r -c 100 example.com runs MTR for 100 rounds and prints a report with a Loss% column for each hop. On Windows, pathping example.com sends 100 queries to each router by default and prints the loss per hop and per link (Microsoft’s pathping reference); it takes a few minutes. Read their loss column with the same rule: what matters is the loss that persists to the final line.
Step 5: Work Out Whose Problem It Is
Once you have the hop where real latency or loss begins, its position tells you who can fix it.
First One or Two Hops: Your Local Network
The first one or two hops are your own network, and the next is often your ISP’s first router. Problems that start there sit between your device and your router or modem. Common causes include Wi-Fi interference, an overloaded home router, a faulty Ethernet cable or too many devices sharing the connection. Restart the router, switch from Wi-Fi to a cable, or check for firmware updates, then trace again.
Middle Hops: Your ISP and Transit Networks
Latency or loss that begins in the middle of the route is on your provider’s network or at the handoff between your provider and a transit network. It is most common at peak evening hours, when residential networks are busiest, and peering disputes can cause lasting trouble at a particular handoff. The hop’s owner, from the ASN lookup, tells you which company to contact.
Final Hops: The Destination Network
If latency or loss appears only at or near the destination, the problem is theirs: an overloaded server, a saturated uplink or a misconfigured load balancer. Report it to the service. If it sits behind a CDN, the problem may clear on its own as the CDN shifts traffic to another site.
Common Traceroute Patterns
- Private addresses at the start: 192.168.x.x, 10.x.x.x or 172.16-31.x.x on the first hops are your own network (RFC 1918). They have no public owner or location.
- Asymmetric routing: replies can come back by a different path, so a hop’s time includes a return trip you cannot see. If your trace looks clean but the problem persists, ask the other side for a trace back to you.
- Routing loops: the same routers repeat over and over until the hop limit. This is a routing misconfiguration, usually at a provider, and it tends to clear once someone notices it.
- Different paths on each run: load balancers spread probes across parallel links, so consecutive hops can show different routers. Paris traceroute keeps all probes on one path for a consistent picture.
Going Further
- TCP traces: when a firewall drops UDP and ICMP, trace with TCP to the service’s own port:
mtr -T -P 443 example.com, ortcptraceroute example.com 443on Linux. - Evidence for your ISP: an MTR report of 100 or more rounds, run while the problem is happening, with the date and time, from a wired connection. Add a trace from outside, such as ours, for the other direction.
- Check the basics too: a DNS lookup confirms the name points where you think it does, and the reverse DNS lookup gives a hop’s name when your trace ran with
-n. - Through a VPN: the trace runs inside the tunnel, so it shows the path from the VPN server onward. The VPN check guide shows how to confirm the tunnel works.

Frequently Asked Questions
How do you read a traceroute?
Go down the hops in order and follow the round-trip times and the asterisks. A rise in latency or a loss of replies that starts at one hop and continues to the destination is a real problem beginning there. A slow or silent hop followed by normal ones is a router that deprioritises probes, and can be ignored.
What does packet loss at one hop mean?
If the hops after it show no loss, nothing: the router limits how often it answers probes while forwarding traffic normally. Only loss that carries on through every later hop to the destination means packets are really being dropped at or after that hop.
Is packet loss in a traceroute always bad?
No. Most loss in a traceroute is routers declining to answer probes, which costs your traffic nothing. Real loss shows at the destination too, and you need many probes to measure it: use pathping, which sends 100 queries per hop by default, or MTR with -c 100 (its report mode sends 10 by default), rather than a single traceroute with three.
Why is latency lower at a later hop than at an earlier one?
Because the earlier router was slow to answer the probe itself, not slow to forward your packets. Each hop’s time is measured separately, so a busy router can report a high figure while the hops beyond it, reached through that same router, report lower ones.
Why does the last hop show packet loss or “Request timed out”?
Many servers and firewalls drop or rate-limit probes while serving real traffic normally. If the website or service works, that is the explanation. Test the service itself, for example whether its port is open, before treating it as a network fault.
How do I prove packet loss to my ISP?
Send them an MTR report of at least 100 rounds, taken while the problem is happening, on a wired connection, with the time and your public IP. Point to the first hop where loss starts and continues to the destination. A trace from outside, such as our online traceroute run to your IP, adds the other direction.
What is MTR and when should I use it instead of traceroute?
MTR combines traceroute with ping: it repeats the trace every second and keeps per-hop loss and latency statistics, jitter included. Use it for intermittent problems and for anything involving packet loss, which a single traceroute is too short to measure.
Can a VPN affect traceroute results?
Yes. With a VPN on, your packets travel inside the tunnel, so the trace shows the path from the VPN server to the destination, not from your location. Run it with and without the VPN and compare the two to see which part of the path each covers.
How many hops should a normal traceroute have?
There is no normal number: most traces on the same continent finish in roughly 10 to 20 hops, and some networks hide hops inside tunnels so a long path looks short. The count matters less than the latency: a 15-hop route over well-connected networks can be faster than a 10-hop route through congested ones.
What does a jump in latency at one hop mean?
Look at the hops after it. If they all stay high, the delay is real and starts on the link into that hop, often a long-distance link or a congested handoff between two networks. If the next hops drop back to normal, that router was only slow to answer probes and your traffic was not delayed there.
Create your free account
Get access to IP lookup tools, bulk reports, and more. Free forever.