Most people judge their home connection from a single number on a speed test, then either shrug or phone the ISP. That number blends three separate limits: what the fibre or cable line delivers, what the router pushes through the air, and what your phone can pick up in the room where you actually sit. Separating those layers takes about twenty minutes and costs nothing.
Get a Wired Baseline Before You Blame the Wi-Fi
Run an Ethernet cable from the router to a laptop and repeat the same test. On a 100 Mbps fibre plan, a wired result between 90 and 96 Mbps is healthy — protocol overhead eats a few percent and no provider promises the headline figure. If the wired number is also poor, the fault sits upstream of your Wi-Fi and no router setting will recover it. The wired Ethernet setup guide covers the cable and port details if you have never used one.
Fix your units before you panic. Connections are sold in megabits; browsers and game launchers report megabytes. Divide by eight. A 100 Mbps line moves about 12.5 MB/s, so a 4 GB game patch takes roughly five and a half minutes at full speed.
With no cable to hand, check the link rate instead. Open Command Prompt on Windows and type netsh wlan show interfaces. The output lists “Receive rate” and “Transmit rate” in Mbps — the negotiated speed between laptop and router, not your internet speed. Seeing 866 Mbps there while a speed test reports 18 Mbps tells you the radio link is healthy and the bottleneck is somewhere else.
Remove the Fake Bottlenecks
Close whatever is quietly consuming capacity: cloud photo sync, a Steam download, a phone uploading video, a smart TV streaming in another room. Disable the VPN. Commercial VPNs typically cut throughput by 30–60% and add 20–80 ms of latency, which is fine for browsing and useless for a diagnostic run.
Stand within three metres of the router, connect on 5 GHz, and keep one device online. Run three tests back to back with Speedtest by Ookla and record the lowest figure, not the highest; the spread between runs tells you whether the link is stable or merely lucky. Choose a test server within about 50 km — one two countries away measures the backbone, not your house.
Then repeat the whole thing at 8 pm. Congestion in the building and on the provider’s uplink shows up in the evening, and a line reading 95 Mbps at 10 am against 30 Mbps at 9 pm is behaving normally for a shared segment rather than failing.
Latency and Jitter Describe the Connection Better
Scroll past the headline download figure. A healthy home line shows ping under 30 ms to a nearby server, jitter under 10 ms, and zero packet loss. Video calls break on 40 ms of jitter long before they break on a shortage of download capacity.
Upload deserves equal attention. A 10 Mbps uplink carries one 1080p call comfortably; two calls plus a phone backup will saturate it and everyone starts sounding robotic. Check the plan document — many “100 Mbps” packages are asymmetric, with 10–20 Mbps up, and no amount of router tuning changes that.
For bufferbloat, run the waveform test at waveform.com/tools/bufferbloat. If latency climbs from 15 ms idle to 200 ms under load, the queue in your router is uncontrolled. Enabling SQM or a smart queue on a supported model fixes that more effectively than paying for a faster plan.
Read Signal Strength in dBm, Not Bars
Bars are decorative. Signal is measured in dBm and the thresholds are unforgiving: −50 dBm is excellent, −60 is good, −67 is where calls begin to stutter, and below −75 the link drops packets without warning.
On macOS, hold Option and click the Wi-Fi icon in the menu bar to see RSSI, noise, channel and PHY mode. Windows hides dBm in the normal interface, so use the percentage from netsh — 70% or better is workable. Android builds differ; most do not expose the number, and a free analyser app is the quickest route. iPhones do not report Wi-Fi signal strength at all, so borrow a second device when comparing rooms.
Check the channel while you are in there. On 2.4 GHz only channels 1, 6 and 11 avoid overlap, so a router sitting on 4 beside three neighbours is losing to interference it cannot see. Channel width matters as well: 80 MHz on 5 GHz buys speed and gives up range, while 40 MHz is the better compromise in a dense block of flats.
Turn the Numbers Into Evidence
Log four values per test — time, room, download, ping — for a week. It takes two minutes a day. “It is slow in the evening” earns a script from the call centre; “wired 94 Mbps at 10 am, 31 Mbps at 9 pm, four nights running, ping rising from 12 to 90 ms” earns a line check.
Use the same log to decide where money should go. Wired fast and wireless weak in one room: reposition the router per the placement guide, or add an access point using an old router as an access point. Both slow at every hour: the line is the problem. Both fast but one laptop still stalls: the laptop is the problem, and its Wi-Fi driver is the first thing to update.