Wifi slows down in the evening because two queues fill at once: your neighbourhood’s share of the last mile during the peak window, and your own home’s airtime as every device wakes up. A wired speed test at 2pm and again at 9pm tells you which of the two is yours to fix.
Nothing changed on the laptop between 2pm and 9pm. The router is the same router. Yet the video buffers, the game sits at 80 ping, and the call in the other room freezes and catches up. Time of day is the only variable that moved.
People call this the ISP throttling them at night. That does happen on some capped plans. Most weeknights, two quieter things pile up at once. Everyone in your house is home. Everyone in the building is home. Wi-Fi is a shared radio. The last mile on cable is a shared pipe. After 8pm both of them fill up.
Why is my wifi slow at night?
Because two queues fill at the same time. Outside the house, the shared last mile is busiest on weeknight evenings. Inside it, every phone, TV and console that wakes up after dinner competes for the same airtime on the same channels.
The FCC’s Measuring Broadband America program treated weeknights from 7pm to 11pm local time as the peak window. That definition is in the Thirteenth Fixed Broadband Report, which used tests from September and October 2022. Ofcom in the UK used a tighter window, 8pm to 10pm, and found speeds lowest around 9pm.
Those hours match the household, not the sunset. People get home, TVs wake up, consoles sign in, phones start cloud backups. HighSpeedInternet.com puts the busy stretch a little earlier, 6pm to 11pm.
There are two places that traffic can jam.
Inside the apartment, every phone, TV, laptop, and cheap smart plug is talking over the same air. Wi-Fi is one conversation at a time on a channel. Neighbors’ routers leak through plaster. XDA’s August 2026 piece on evening slowdowns is blunt: if Ethernet is fast and Wi-Fi is not, the ISP is not the one you should be yelling at.
Outside the apartment, cable last-mile is shared to a node. Fiber to the home is much closer to a private pipe. Ofcom’s March 2023 measurements put average 8-10pm download at 67.7 Mbit/s, which was 95 percent of the 71.0 Mbit/s average maximum. Full-fibre lines took the smallest hit. ADSL2+ lines, the old copper ones, averaged 92 percent of their maximum at peak.
The FCC’s own peak-hour wired tests were kinder to US ISPs than the living-room argument usually is. In the Thirteenth Report, eight of twelve ISP and technology setups delivered 100 percent or better of advertised download during 7-11pm. The other four landed between 86 and 90 percent. Those boxes sat on the gateway. They did not measure the hop from the router to the couch.
I keep that in my head when a speed test on a phone looks tragic at 9pm. The phone is sitting on the worst hop in the chain.
How do I tell whether it is my wifi or my internet?
Run the same test twice, once on an Ethernet cable and once on wifi, at about 2pm and again at about 9pm. If the wired test holds up at night and only wifi collapses, the problem is inside your walls.
Do this once at about 2pm and once at about 9pm. Same laptop if you can.
Plug Ethernet into the router. Run a test. Unplug. Stand where you actually sit. Run it again on Wi-Fi.
Broadband Switch’s UK guide uses a table I wish more people would copy:
| What you see | What it means |
|---|---|
| Wired is fast, Wi-Fi is slow | The air in the room, not the line |
| Wired dips only at peak | Neighborhood or ISP congestion |
| Wired is slow all day | Line, modem, or the plan itself |
Optimum, which is an ISP and has every reason to blame your house, still tells customers to do the wired-versus-Wi-Fi check first. Advertised speeds are for a wired connection. Wi-Fi will undershoot that even on a good night.
If you only test on a phone in the bedroom, you will “prove” the internet is dead. You have proved the bedroom is a bad radio.
A modest evening dip on the wired test is normal. Ofcom’s nationwide average was about 5 percent down at peak. A wired drop that cuts the speed in half, every weeknight, is a different complaint. That one you can take to the provider.
I have not pulled a 2026 national US chart that replaces Measuring Broadband America. The FCC says that program stopped collecting in-home data after 1 August 2023. Your own 2pm-versus-9pm pair is more honest than a blog table with round numbers.
Should I use 2.4 GHz or 5 GHz at night?
Move what you can to 5 GHz. 2.4 GHz has only three non-overlapping channels, and after dinner every router in the building is crowding them. The network people pick because it “has better range” is often the one that fails first.
Open the router app. If the TV and the laptop are on the 2.4 GHz network because it “has better range,” that is often why evenings feel cursed.
The 2.4 GHz band is short on spectrum. In the US, channels 1 through 11 are the usual list. Only 1, 6, and 11 sit far enough apart to avoid stepping on each other. Tektronix’s 802.11 primer and the Wikipedia WLAN channel list both land on that same three-channel rule. Europe can use 1, 5, 9, and 13. Everyone else in the building is still picking from the same tiny menu.
5 GHz is faster and usually quieter. It dies sooner through walls. 6 GHz, on Wi-Fi 6E and Wi-Fi 7 radios, is the least crowded of the three, with the shortest reach. I have not measured wall loss in meters for this piece, so I will not invent a range chart.
XDA’s warning about auto channel is the one I actually follow. The router picks a “best” channel at 10am, then clings to it while the building fills up. If your firmware lets you, pick 1, 6, or 11 on 2.4 GHz after a Wi-Fi analyzer pass, and park the TV and the laptop on 5 GHz.
Ethernet still wins for the box that cannot stutter. I wire the main TV when I can. Paying for a gigabit plan does not reserve a radio channel.
A $20 “extender” that repeats 2.4 GHz into more noise often makes the shouting match louder. Mesh can help a long apartment. I am not ranking brands here. Mesh versus extender is a different article.
Is 4K streaming what is slowing everyone down?
Rarely on its own. A single 4K stream is a small share of a modern connection. What fills the pipe is the total: several streams plus a game, a video call and a phone backing itself up to the cloud at the same time.
People assume 4K is the villain. Netflix’s own help page, checked for this draft, recommends 3 Mbps or higher for 720p HD, 5 Mbps or higher for 1080p, and 15 Mbps or higher for 4K. One stream is small. Three 4K streams plus a game plus a video call plus a phone backup is not.
| Source | Number | What it is |
|---|---|---|
| Netflix Help | 3 Mbps or higher | Recommended for 720p HD |
| Netflix Help | 5 Mbps or higher | Recommended for 1080p |
| Netflix Help | 15 Mbps or higher | Recommended for 4K |
| FCC MBA 13th Report | Weeknights 7pm–11pm | Peak window; tests sat on the gateway |
| Ofcom, March 2023 | 67.7 vs 71.0 Mbit/s | Average 8–10pm download vs average maximum |
| Ofcom, March 2023 | 92% of maximum | ADSL2+ peak download vs that tech’s maximum |
I have not pulled official YouTube or Disney+ bitrate tables for this piece. Treat those as unverified here.
If the wired 9pm test still has 80 Mbps and the TV buffers, look at the Wi-Fi band the TV joined, not the plan. If the wired 9pm test is 12 Mbps on a 300 Mbps plan, that is congestion or a fault on the line. Raising the TV’s quality setting will not fix a shared node.
Throttling, the thing people mean when they say the ISP turned them down, is a real policy on some usage-based plans. It is also the explanation everyone reaches for first. Optimum’s own explainer says most nighttime slowdowns are congestion and Wi-Fi, not a switch flipped on your account. I cannot see your ISP’s traffic-management page from here. A VPN that magically doubles speed is a clue, not a diagnosis. I have not verified that test on a named US cable plan.
Cloud backups are the silent hog. Laptops and phones often wait until they are on home Wi-Fi, then dump photos while you try to watch something. Pause that job for an hour and run the test again before you assume the street is full.
When should I call my internet provider?
After the wired test also fails at night. If the cable is slow at 9pm and fine at 2pm, the congestion is outside your walls and the provider is the right call. If only wifi drops, calling gets you nowhere.
The useful order is boring.
Move the router off the floor and out of the TV cabinet. Put the heavy devices on 5 GHz or Ethernet. Pause the backup. Re-test at 9pm, wired first.
If Wi-Fi is the gap, fix the radio before you buy a faster plan. A gigabit down to a phone on a crowded 2.4 GHz channel still looks like a bad night.
If Ethernet is the gap, and it only appears after 7pm, you have a peak-hour case. Write down the times. Keep the wired results. Then call. Fiber to the home, where you can get it, is the access type Ofcom found least sensitive to the evening rush. Cable and copper can be fine. They just share more.
Openreach’s 2023 year-end audit, as reported by Computer Weekly, put the UK network’s typical busiest hour between 8pm and 9pm. That lines up with Ofcom’s 8-10pm window. It does not prove your block is congested. It does tell you when to measure.
Why your wifi is slow at night is usually the building waking up, not a dead modem. Measure both queues before you buy a new router or a new plan.
Frequently asked questions
What counts as peak internet hours?
The FCC’s Measuring Broadband America program treats the peak usage period as weeknights from 7pm to 11pm local time. Ofcom in the UK uses a tighter 8pm to 10pm window and found speeds lowest around 9pm.
Is my provider throttling me at night?
Usually not. In the FCC’s Thirteenth Report, eight of twelve ISP and technology setups delivered 100 percent or better of advertised download speed at peak, and four delivered 86 to 90 percent.
How much speed does 4K streaming actually need?
Netflix’s help page recommends 3 Mbps or higher for 720p, 5 Mbps or higher for 1080p and 15 Mbps or higher for 4K.
