- What determines your internet speed
- How to measure your speed: the test done properly
- 6 ways to improve your internet speed
- DNS: it won't add Mbps, but it changes how it feels
- Cabling and powerline: taking the network where Wi-Fi can't reach
- Diagnostic table: what to try, depending on the symptom
- When the problem is outside your home
- Three typical «slow internet» cases
- Frequently asked questions
- The bottom line
Somebody in your house says «the internet is slow» and inside a minute three people are blaming three different things. Your partner blames the provider. Your kid blames the Wi-Fi. You blame the router, because it is the box with the little lights and the lights always look guilty. Nobody has measured a thing.
That sentence, «the internet is slow», can mean at least five different things. Your provider is not delivering the speed you pay for. The Wi-Fi does not reach that room. The router has been running for three years without a proper restart. Someone is uploading a 4 GB video while you try to hold a video call. Or the line is not slow at all, it simply drops packets and stutters. Each one has its own fix, and none of them is «switch to a bigger plan».
So here is the deal. Before you touch a single setting, measure. Half of the «it is slow» complaints are solved with one test and one number in your hand. The other half are solved the moment you accept that speed and stability are two different animals. Let me walk you through it step by step.
What determines your internet speed
When we say internet speed we are lumping together four things that behave differently. They are not measured the same way, and they are not fixed the same way. Keep them apart in your head and half the mystery disappears.
- Bandwidth (Mbps). How much data fits through the pipe each second. It is the number your provider prints in giant letters: 600 Mb, 1 Gb. Watch the unit, because this is where a great many arguments are born. It is megabits. A megabyte is eight times bigger, and 1 MB/s equals 8 Mbps. A download that shows 12 MB/s is pulling 96 Mbps, not 12.
- Latency or ping (ms). How long one piece of data takes to leave your device and come back. Low latency is what makes a connection feel quick, even when the raw bandwidth is modest. The ITU places the point where voice starts to suffer at around 150 ms one-way. Past that line, people on a call begin talking over each other without noticing.
- Jitter or stability (ms). How much that latency moves around from one packet to the next. A steady ping of 60 ms feels perfectly fine. A ping of 20 ms that spikes to 300 ms whenever it feels like it feels broken. Jitter is the difference between a smooth call and one where every fourth word falls off.
- Packet loss (%). The share of data that never arrives. Even 1% is enough to force retransmissions and to wreck anything happening in real time. A file transfer survives it, because the system quietly resends what went missing. A live conversation does not have that luxury.
That is why two connections with the same Mbps figure can feel completely different in daily use. One lands a video call without a blink, the other chops it into pieces. And it is why «it is slow» almost never gets fixed by paying for a bigger plan.
How to measure your speed: the test done properly
An internet speed test measures how long data takes to travel from your device to a server and back. Think of it as a photo of a single moment. A photo is only useful when you take it under conditions you can compare. Follow these five rules and the number starts to mean something.
- Connect by cable if you possibly can. Wi-Fi adds its own noise and its own latency to the reading, so a test over Wi-Fi measures your wireless link as much as your line.
- Close everything that is downloading (updates, cloud backups, streaming, torrents). One background upload can flatten the result and leave you certain the fault is somewhere it is not.
- Repeat the test at different hours. Your provider’s network has rush hours, just like a motorway, and the same line can hold up at noon and crawl at nine in the evening.
- Look at the ping, the jitter and the packet loss, not only the download figure. The big number is the one everybody reads and the one that explains the fewest problems.
- Try more than one tool. An independent platform like M-Lab and a commercial tester give different results because they use different servers and different methods. When the two disagree, that disagreement is the useful part.
A practical rule I keep coming back to. If a cable test lands close to what you pay for and a Wi-Fi test does not, your internet is healthy and your local network is the suspect. That single comparison has saved people more support calls than any other trick in this article.
What each number actually tells you
The big number only rules in some situations. For almost everything else, latency is in charge.
- Large downloads and streaming: bandwidth. A 4K film asks for roughly 15 Mbps, a Full HD one about 5.
- Browsing and opening pages: latency and DNS. You can have megabits to spare and still spend your day waiting, if the ping is high.
- Video calls and online gaming: latency, and above all jitter and packet loss. A Full HD video call runs at about 4 Mbps, so it rarely fails for lack of bandwidth. When a call breaks, the shape of the connection matters far more than its size.
- Several devices at once: aggregate bandwidth and sensible prioritisation. Four people streaming 4K together will notice a thin pipe before anyone else does.
One detail almost nobody checks: the idle test. A low ping means nothing if it explodes the moment something starts uploading. That failure has a name, bufferbloat, and you hunt it down with tests that deliberately create load. The bufferbloat community keeps the explanation and the tools, and the remedy is standardised in RFC 8290: smart queue management, or SQM. If your router supports SQM, switching it on is often the single biggest win available for call quality, and it costs nothing.
6 ways to improve your internet speed
1. Restart the router (properly)
A router is a small computer that has been running for months, juggling connections, warming up, filling its memory with half-finished tasks. Unplug it for thirty seconds and you clear that memory and force a fresh conversation with your provider. That is plain housekeeping. These boxes were never meant to run for years without a break. And if you have to do it every day, the router is telling you it wants to be replaced. While you are in there, update the firmware. A surprising share of drops and dead spots are cured by a version the manufacturer shipped months ago and you never installed.
2. Keep Wi-Fi away from interference (and pick the right band)
Microwaves, cordless phones, wireless speakers and thick walls all chew on the signal. The closer you sit to the router and the fewer obstacles sit in between, the better it gets. The ideal place is the middle of the home, high up, out in the open. Not inside a cupboard, not behind the television, not next to the aquarium with its pump running all night.
Then there is the band. 2.4 GHz reaches further and pushes through walls, but it is crowded and slow. 5 GHz is faster, offers more channels and suffers less interference, at the cost of range. 6 GHz, which arrived with Wi-Fi 6E and Wi-Fi 7, adds about 1,200 MHz of new spectrum (the Wi-Fi Alliance documents it) and is the cleanest of the three, and also the worst at crossing a wall. The sensible move is to have all of them and let the router steer each device, or join the band that fits where you are sitting. Radio is not the only way to carry a signal either: Li-Fi sends data on light instead of radio waves, immune to the interference that clogs 2.4 GHz, in exchange for needing a line of sight.
3. Change the Wi-Fi channel
On the 2.4 GHz band, everyone around you broadcasts on the same handful of channels, and only three of them do not overlap: 1, 6 and 11. Open your router settings and pick the least crowded of the three, or let a modern router do the choosing for you. Few fixes cost less or return more. On 5 and 6 GHz there is far more room and the problem mostly fades, though some channels are shared with radar and certain devices skip them by default. A Wi-Fi analysis app tells you in seconds which channels are free around your own flat.
4. Set a decent password (and keep the neighbours out)
An open Wi-Fi or one with a weak password is an invitation stuck to your front door. Change the key the router came with, use WPA2 or, better, WPA3, and now and then open the list of connected devices. If something shows up that you do not recognise, you have a guest you never invited. To put a name to the intruder, it helps to know what a MAC address is and how to find it. A modern router also gives you a separate guest network, so you never have to hand the main key to everyone who walks past.
5. Watch the apps that «nibble» bandwidth
Cloud backups, system updates, games running in the background, the television uploading who knows what. All of it runs behind your back and eats the pipe while you blame the provider. On Windows, Task Manager shows you who is consuming. On a phone, look at data usage per app. Shutting down what you do not need frees more bandwidth than you expect. If your router allows it, switch on QoS so a video call does not fight on equal terms with an 80 GB backup, and schedule the heavy jobs for the small hours, when the network is asleep.
6. Upgrade your equipment before your plan
If you have 1 Gb fibre and a router from eight years ago, you are paying for speed you will never touch. A decent router, or a mesh system if your home is large, does more for your real speed than any upgrade in tariff. Before you sign a bigger contract, check three classic bottlenecks: 100 Mbps Ethernet ports on routers sold as «gigabit», old Wi-Fi cards that only speak 2.4 GHz, and cables or wall sockets that have seen better days. And if the machine itself is short on memory, here is how much RAM you actually need. A computer that is drowning makes everything look slow, the network included.
DNS: it won’t add Mbps, but it changes how it feels
Changing your DNS does not give you a single extra megabit. What it does is resolve domain names sooner, that invisible first step that happens before a page begins to load. With a fast public resolver (Cloudflare 1.1.1.1, Google 8.8.8.8) or with your provider’s DNS, the difference shows up while browsing, not in a download.
Two things to keep straight. DNS is cached, so the gain is felt most on the sites you visit for the first time. And encrypting it with DoH or DoT improves your privacy, because your provider stops seeing which domains you ask for, and does nothing for speed. If yours is slow or filters more than you like, switching is free and takes two minutes, on the router or on each device.
Cabling and powerline: taking the network where Wi-Fi can’t reach
If you want real speed and stability, the cable still wins. A well-crimped Cat5e Ethernet cable carries 1 Gb up to 100 metres, Cat6 and Cat6a go further, and they make sense when you are thinking years ahead. Wi-Fi is a shared, variable medium. Cable is neither. Whenever you can, wire what does not move: the television, the console, the work computer.
When running cable is not an option, two shortcuts remain. A powerline adapter sends the network over your electrical wiring. It works, but it depends on the age and quality of that wiring and usually lands well below the number on the box, so do not plug it into a power strip or a charger. A mesh system spreads several nodes around the home and does the job, and does it far better when you link the nodes to each other by cable. A lone repeater is the weakest option of the three, because it amplifies the bad signal you already had. Keep it as a last resort.
Diagnostic table: what to try, depending on the symptom
The fastest way to find the culprit is to match the symptom you feel with one concrete test.
| Symptom | Likely cause | What to try |
|---|---|---|
| Fine over cable, poor over Wi-Fi | Local network or coverage | Channel, 5/6 GHz band, placement, mesh |
| Over cable, far below your plan | 100 Mbps port, cable or wall socket, device | Check the negotiated link speed, swap the cable, try another device |
| Fine by day, slow from 9 p.m. | Provider congestion | Repeat the test for several days and complain with data |
| Low ping at idle, high while downloading | Bufferbloat | Enable SQM or QoS on the router |
| Video calls break up but the test shows 600 Mb | Jitter or packet loss | Measure jitter and loss, check Wi-Fi and congestion |
| Only one website is slow | The server or its CDN | Try other sites and another network (mobile data) |
| The phone is slow, the PC is fine | The phone’s own Wi-Fi | Band and channel, move closer to the router |
When the problem is outside your home
If you have ruled out the local network, the problem lives with your provider or out on the internet itself. The usual suspects: peak-hour congestion, because the fibre is shared upstream and everybody streams at the same hour, an outage in your area, long routes toward one particular server, and the gap between the «up to» figure in your contract and what actually arrives.
No provider guarantees 100% at every moment. What helps is having your own set of measurements, taken at different times and over cable, before you pick up the phone. With that data the conversation changes. You stop saying «it is slow» and start saying «here are my numbers». If the gap is always there and no outage explains it, you have grounds to ask for a review or to move to another provider.
Three typical «slow internet» cases
You have 1 Gb and download at 12 MB/s. Twelve megabytes per second is about 96 Mbps, dead on for a 100 Mbps link. Nobody is cheating you. Something is negotiating at a hundred, and it is almost always the cable, the wall socket or an old port. Replace that link and you fix what no contract can fix.
Fine by day, terrible at night. This is the textbook shape of peak-hour congestion. Your router is innocent. The provider’s network is full. Measure for several days at the same hours, keep the results, and complain with them in hand. Switching providers may help and it may not, because congestion belongs to the network, wherever you buy your service.
You work from home and the calls keep dropping. The test reads 600 Mb and the meeting still breaks apart. Here the size of the pipe is irrelevant. You have high jitter or packet loss, usually from a crowded Wi-Fi band or from bufferbloat. A cable to the computer and SQM on the router tend to cure it without touching the plan.
Frequently asked questions
Why am I getting less speed than I pay for?
Almost always for several reasons at once: you measured over Wi-Fi, the router is tired, the network is saturated at peak time, or the server on the other end has its own limit. Measure over cable and at different hours before you decide your provider is cheating you.
Is a speed test reliable?
It is indicative. It is good for comparing and for spotting a problem, but the figure depends on the distance to the test server, the device and the moment. Forget the exact number and watch the comparison: cable against Wi-Fi, morning against night.
Is it worth going from 600 Mb to 1 Gb?
Only if the rest of your setup keeps up (router, cabling, network card) and you genuinely move a lot of data at once. For browsing, remote work and streaming, 300–600 Mb well used beat 1 Gb badly shared.
Is 5 GHz Wi-Fi always better?
It is faster and it suffers less interference, but it travels less far and crosses walls worse. The ideal is to have both bands and let each device use the one that fits where it sits.
Does changing DNS make the internet faster?
It does not speed up downloads, because DNS does not carry your data. It only translates domain names into addresses. What it can do is cut the wait before a page starts loading. You notice it while browsing, and you will never see it in a speed test.
Why do my video calls break up when the test shows 600 Mbps?
Because the test mostly measures bandwidth and leaves out what actually destroys calls: jitter and packet loss. A link can be very fast and very uneven at the same time, and real-time traffic hates uneven. Measure jitter and loss, and suspect the Wi-Fi or bufferbloat before you blame the plan.
The bottom line
Improving your internet speed is not about paying more. It is about measuring first and deciding after. Cable before Wi-Fi, router before plan, stability before the big number. That order solves around 90% of the «it is slow» complaints without you opening your wallet.
So do this before you call anyone. Run a test over cable. Run another over Wi-Fi. Run both again tonight. Write the four numbers down and compare them. If the cable result sits close to your plan and the Wi-Fi result does not, you already know where to look, and you can fix it this afternoon.
One last thing. When the connection breaks without losing bandwidth, the answer almost always hides in latency and jitter. Measure both, at idle and under load, before you touch your plan. Those two numbers tell you more about a dying call than the big one ever will.