Is It Your ISP or Your Wi-Fi? A 10-Minute Test

The stream stalls, the speed test comes back at a third of what you pay for, and you are one hold-music session away from losing an evening. Before you dial, spend ten minutes proving which side of the wall the fault is on.

One test settles it, and it needs a cable rather than an app. The rest is reading the number correctly and opening the call so nobody recites the reboot script at you.

The wire is the boundary#

Your connection has two halves that fail independently. Everything from the internet to the LAN port on your router belongs to your ISP: the line into the building, their modem or ONT, the shared segment your street sits on. Everything past that port is yours — radios, walls, channel, devices.

An Ethernet cable from the router to a laptop bypasses the second half, so whatever you measure over it is the best your house can do.

• Wired is slow too. The fault is upstream of every radio in the building, and no placement change, extender or mesh kit alters that number by a single Mbps. This is when calling is right.
• Wired is fine, Wi-Fi is not. The call will waste an afternoon. The problem is inside your house, and it is yours to fix.

One caveat: the router straddles the boundary, so a router that cannot route at your plan speed looks like an ISP fault over the wire too. Where the speed goes on a gigabit plan walks the rest of the chain.

The 10-minute test#

Bring a laptop that takes an Ethernet cable, directly or through a USB-C adapter, plus the cable itself.

1. Write down the plan you actually pay for. The exact advertised numbers, up and down — 500/50, 300/10, 1000/1000. Guessing here wrecks everything that follows.
2. Test over Wi-Fi where it hurts. Run a speed test where you normally sit, and record download, upload and idle latency.
3. Test over Wi-Fi beside the router. Same device, same test, within two minutes. That separates a coverage problem from a house that is slow everywhere.
4. Test wired at the router. Plug the laptop into a LAN port and turn its Wi-Fi off — this is the step people get wrong. Run the test twice, thirty seconds apart, and keep the better result.
5. Watch the latency while the wired test runs. Note the peak, not the average. A figure that climbs from 20 ms to 300 ms while data moves is a queueing problem, and it ruins calls long before the Mbps number looks bad.
6. Repeat the wired test with the house quiet. Pause backups, stop the 4K stream, take the console offline. If the number jumps, you were competing with yourself.

Reading a speed test without fooling yourself#

Plenty of "my ISP is throttling me" conclusions are measurement errors. Six traps, in the order they catch people.

The port is narrower than the plan. A 100 Mbps Ethernet port — common on older laptops, cheap adapters and forgotten switches — caps near 94 Mbps in practice. A gigabit port tops out around 940 Mbps once framing overhead is subtracted, so 940 is a full gigabit, not a shortfall.

The client radio is the ceiling, not the line. A phone cannot validate a fast plan. A two-stream Wi-Fi 5 client on an 80 MHz channel negotiates up to roughly 866 Mbps and delivers 300 to 400 Mbps in the same room, so a 300 Mbps Wi-Fi result on a gigabit plan says your Wi-Fi is healthy, not that your plan is short.

The server matters. Run your ISP's own test server and a distant one. Fast locally but slow to everything else is a peering complaint, not a bad line.

One stream is not the plan. Browser speed tests open several parallel connections; a single download, or anything through a VPN, gets a fraction of that. Judge the plan on the multi-stream number.

Upload has its own budget. Cable plans are deliberately lopsided — 500/20 and 300/10 are normal — and upload is what calls, backups and screen sharing live on. Compare it against the advertised upload figure, never against download.

Time of day changes the answer. A wired result that is fine at 10 a.m. and terrible at 9 p.m. is contention, not a broken line, and it needs its own log — the repeatable version is in why Wi-Fi gets worse in the evening.

What your two numbers mean#

Find the row matching the two results you recorded.

Two rows do most of the work. If wired is close to the plan, put the phone down — the one-slow-room walkthrough or why video calls keep dropping beats any technician. If wired is well under the plan at every hour with the house quiet, you have a case.

Making the call count#

What to write down first#

First-line support runs a script that assumes you measured nothing. Numbers skip it.

• Your advertised download and upload speeds, and your account number.
• The make and model of the modem or ONT, and of the router if separate.
• Each result with date, time, server, and whether it was wired or wireless.
• How long it has been happening, and whether it is constant or clock-shaped.
• The modem uptime, which shows whether it has been resetting itself.
• The signal readings from the modem status page.

That last one turns a complaint into a case: most modems expose a status page in a browser, and its numbers are the ones a technician will look at anyway.

Treat these as rules of thumb; a value outside them is still a concrete thing to read aloud.

The script#

Open with the conclusion, not the story:

"I have a 500 down, 50 up plan. With a laptop plugged into the router by Ethernet and Wi-Fi turned off, I get 180 down and 45 up. I tested at 10 a.m. and 9 p.m. on three days, on two servers, with nothing else running in the house. My modem shows a downstream SNR of 29 dB and upstream transmit at 52 dBmV, and I have already power-cycled both boxes. I would like a line test, and an engineer visit if it comes back clean."

Four habits make that work.

1. Say "wired" in the first sentence. It deletes the whole Wi-Fi branch of the script, where most calls go to die.
2. List what you already tried so the agent cannot loop you through it.
3. Ask for specific things: a line test while you hold, the readings they see at their end, any known outage on your segment, and an appointment if the test comes back clean.
4. Get the ticket number, the agent name and the time. Ask what result would count as fixed, so the follow-up call has a standard.

When the answer is not the ISP#

If the wired number is close to your plan, the phone call is the one thing that cannot help. Nor can an extender: it raises signal level where signal level is missing, and nothing else. Put one in a room that already reads better than about -60 dBm and you have added a radio to a busy channel.

A good wired number leaves a short list. One slow room is a placement problem, often a free one. Every room slow points at the router, the channel width or one old device eating airtime — why full bars can still crawl covers that. And fine bandwidth with bad timing is what breaks calls while the speed test still looks excellent.

Settling the coverage half on a floor plan#

The cable settles the ISP half for free. The coverage half normally costs an afternoon of carrying a phone from room to room, and Range Up replaces that walk with a simulation you could run while you are on hold. Trace the place you live in, set the material preset — wood frame, brick or concrete — to whatever the building is, mark the router's real position, and every room comes back with a predicted signal level. Both verdicts earn their keep. A failing room the model shows as well covered is not short of signal, which leaves the router, the channel, one greedy device or the timing. A room the model puts down in the -70s is a real hole, and you can name it as the room that matters, have candidate spots for a second unit weighed against it, and flip between the coverage you have now and the coverage that spot would give you. It predicts from the drawing rather than reading your live network, so the wired test stays the thing that answers the ISP question. Draw your place and see the map, or read how a placement decision gets made first.