Why Your Wi-Fi Gets Worse in the Evening

Everything works at ten in the morning. By half past eight in the evening the same video call breaks up, the same show buffers, and the same speed test comes back at a third of what you pay for.

Nothing in your home changed between those two moments — which is exactly what makes this diagnosable. Three different failures produce that identical evening curve, and the fix for each one is useless against the other two.

Three suspects wearing the same costume#

From the couch these look the same. Bars are full, the router lights are normal, and the problem clears itself by midnight. The only thing that separates them is where you measure, and when.

Notice what is missing from that table: none of the three is a coverage problem. If one room fails all day long rather than only after dark, you are reading the wrong article — why one room is slow covers that case.

The evening test that tells them apart#

The whole test rests on one idea: compare the same measurement to itself at a different hour. A single speed test at 9 pm proves nothing, because you do not know what that number looks like at 10 am on the same device in the same spot.

You need a computer with an Ethernet port or a USB-C adapter, a network cable, and ten minutes on two occasions.

1. Quiet hour, wired. Around 10 or 11 am on a weekday, plug the computer into a LAN port on the router. Run a speed test and record download, upload and idle ping. Then start a second speed test and watch the ping while it runs — latency under load matters more than the Mbps figure.
2. Quiet hour, Wi-Fi beside the router. Unplug, stand next to the router, repeat. This is your Wi-Fi tax with distance removed from the equation.
3. Quiet hour, in the problem room. Repeat once more where you actually sit.
4. Repeat all three at 9 pm on a weekday. Same device, same speed test server, same spots.
5. One extra reading at 9 pm, wired, with the house paused. Turn off the TVs, pause phone photo sync, stop any console or computer downloads. If the wired number was bad and now it is good, the call is already made.

Keep it honest: same device, same server, same spot, same activity. Comparing a phone at 9 pm against a laptop at 11 am tells you about the phone.

Reading your log#

Two of these often run together — a marginal room and a busy channel is the usual apartment combination. Work from the top of the table down, because the causes above outrank the ones below: there is no point tuning a channel while the wire itself is slow.

If it is ISP contention#

Your connection is shared somewhere upstream. A cable line puts your home on a coax segment with dozens to a few hundred other homes; fiber typically splits one feed across 32 or 64 homes; fixed wireless and satellite share a cell or a beam. When those households all sit down at 8 pm, the shared part fills.

Two patterns give it away. On cable, upload usually collapses harder than download, because the upstream channel is the narrower one — and upload is what video calls and cloud backups need. Second, contention keeps to the same clock every day and eases by around 11 pm.

What to do about it:

• Do not buy hardware. A faster router, a mesh kit and an extender all sit on your side of the bottleneck. So does everything in the gigabit plan, 50 Mbps reality chain of caps.
• Do not automatically upgrade the plan. A bigger number on a saturated segment still queues behind the same neighbors. Upgrading helps when your plan is genuinely too small for your household, which the paused-house reading in step 5 already tested.
• Collect two weeks of evidence using the same wired test at the same two times each day, then call. A log beats a complaint, and it is the difference between a scripted reboot suggestion and a line card being checked. How to prepare that call has the details.

If it is your neighbors#

If the wire holds up but Wi-Fi beside the router collapses at 9 pm, you are queuing for airtime with the building, not for bandwidth with your ISP. Same evening curve, completely different reason: everyone else's Wi-Fi wakes up when yours does.

That is a settings problem, not a hardware problem, and it is covered end to end in channel congestion, channel width and DFS. The short version: channel width is usually a bigger lever than channel number, and adding a repeater makes congestion worse.

If it is your own household#

This is the cause people dismiss first and confirm most often. Your evening is when every device in the house wakes up at once.

Two 4K streams, a console update and a backup will bury a 100 Mbps connection, and the backup alone can ruin a call on almost any consumer upload. But raw Mbps is only half of it. Airtime matters again inside your own home: an old 2.4 GHz-only tablet that links at 20 Mbps takes roughly ten times as long to move the same file as a modern device at 200 Mbps, and while it is talking nobody else can.

The fixes that are genuinely in your control, in rough order of payoff:

1. Wire the biggest consumer. The TV never moves. A cheap Ethernet run takes 20 Mbps of streaming off the air completely and removes the household's heaviest airtime user.
2. Turn on QoS or SQM in the router and set it to about 85 to 95 percent of your measured wired line rate. This is what stops one upload from flattening every call in the house.
3. Schedule the greedy jobs for 3 am. Console auto-update windows, computer backups and photo sync are all configurable, and none of them need to run at dinner time.
4. Push everything capable onto 5 or 6 GHz and leave 2.4 GHz to the sensors and plugs that cannot do anything else.
5. Retire or replace the one ancient device that still links at legacy rates. It costs the whole network more airtime than it costs itself.

What an extender will and will not do here#

Be blunt about this, because it is where most of the money gets wasted.

An extender adds no bandwidth. If your wired reading at 9 pm is bad, packets are already queuing before they reach the air, and no amount of rebroadcasting inside the house changes that. If the building is the cause, a basic repeater actively hurts: it takes two turns on the air to deliver what the router delivered in one, so the far room ends up with roughly half the throughput while the building's shared channel carries twice the airtime.

There is exactly one evening scenario where an extender earns its place: the third row of the decision table, where a specific room has so little signal margin that it is the first thing to fold when the air gets busy. Lift that room from -72 dBm to -58 dBm and it stops being the canary. If that is your case, the position matters more than the model — where an extender actually belongs explains the target window, and what to use when a repeater is the wrong tool covers the alternatives.

Planning the part you can control#

Range Up cannot see your ISP, does not measure your live connection and does not run speed tests — your wired log owns that half. What it plans is the margin question, which is the half you can actually change: sketch your floor plan, set the material preset to match the build, mark where the router sits, and read the predicted signal in the room that keeps failing at 9 pm.

A seat the heatmap puts in the -50s has margin to spare, so its 9 pm collapse is contention rather than coverage and your answer is in the ISP or household sections above. A seat it puts at -70 dBm or worse is the canary from the callout: hand the app that room, let it weigh positions for a second radio against it, and compare the two coverage views before you decide the gain is worth buying. Sketch your layout and find out before you spend anything.