How to test your internet speed without skewing the result

A speed test does not measure your subscription. It measures the slowest link in the chain running from the test server to your screen, and that link is almost always inside your home. Working out which one changes everything: the same flat can display 80 Mbps or 930 Mbps depending on how you test, with the line playing no part in the difference.

Your test measures the weakest link, not your line

Between the server sending the data and the browser displaying it sit a dozen stages: the provider's line, the optical network terminal, the modem, possibly an extender, the radio or cable hop to your machine, that machine's network card, the operating system, the browser. Each imposes its own ceiling. The number you see corresponds to the lowest of them all, and to nothing else.

That is why "I have fibre but it's slow" is nearly always a badly framed sentence. The line itself is fine. The bottleneck sits downstream, inside the home, on a link the test never names.

Three figures to frame the problem. An FTTH (fibre to the home) line at 1 Gbps delivers to the modem everything it promises, give or take a few per cent. A five-year-old laptop on Wi-Fi from the upstairs bedroom will often pull back under 200 Mbps. And if that same laptop has an old 100 Mbps network port, it will cap at 94 Mbps even when plugged straight into the modem with the finest cable in the world.

Why does Wi-Fi almost always skew the measurement?

Because Wi-Fi is a shared, half-duplex network whose throughput collapses with distance and obstacles.

A radio link always advertises a theoretical rate that adds together transmission and reception across every antenna at once. In real life, expect half that value at best, often less. Add a solid brick load-bearing wall, very common in older Belgian housing, and you lose another half to three quarters of the signal. Add the neighbours' networks sitting on the same channels, the television streaming in 4K downstairs and the phone syncing its photos: each one takes airtime away from your measurement.

The result is not wrong. It is simply unrelated to the question asked.

The objection deserves airing, because it is a strong one: ultimately the only speed that matters is the one reaching the device, in the room where it is used. Why measure a line you never actually reach?

It is a strong objection, and it misses a distinction. To judge your comfort, yes, the sofa measurement is the right one. To judge your subscription, to complain about it, to compare it with a rival offer or to have it fixed, that measurement is worthless, because your provider never made any commitment about the quality of your Wi-Fi. It committed to what it delivers to its own equipment. The two measurements complement each other rather than replacing one another, and our comparison of extenders, powerline adapters and mesh systems starts precisely from the gap between them.

What does each link actually let through?

The table below gives the orders of magnitude observed on each link, set against the tiers genuinely sold in Belgium. On its own it explains most of the disappointment.

Orders of magnitude recorded by Nicolas on setups common in Belgium, August 2026. The theoretical Wi-Fi figures add up every antenna and both directions of transmission: no single device ever reaches them. The Ethernet values, by contrast, are stable and stem from network protocol overhead.

One consequence stands out, and it is a Belgian one. Providers here sell plans at 2 Gbps, 8.5 Gbps and up to 10 Gbps. None of those figures can be measured with an ordinary laptop, whose network port caps at 1 Gbps. Verifying a multi-gigabit line requires a 2.5 Gbps adapter or better, storage fast enough to keep up, and often several simultaneous transfers. Worth knowing before you sign: on those tiers the contractual figure will stay unverifiable at home for a long while.

Six conditions for a measurement that counts

BIPT, the Belgian telecoms regulator, publishes its own method, and it is more demanding than what you read elsewhere. Here it is, condition by condition.

• An Ethernet cable between the computer and the modem. Not the TV set-top box, not an extender: the modem itself, on a free network port.
• The machine's Wi-Fi switched off. Otherwise the operating system may keep routing part of the traffic over the radio, and you end up measuring a blend of both.
• Nobody else online in the home. Every active stream consumes capacity the test will lack, and the measurement reflects it immediately.
• Every other program closed on the measuring device. A background backup or a game update is enough to halve the result.
• Repeated measurements, over several days, between 7 pm and 9 pm. That is peak hour: it serves to show the problem is constant or recurring rather than one-off.
• A screenshot of every measurement, stored with its date and time. Without traces there is no case file.

Those six points do not serve the same purpose. The first four make the measurement accurate. The last two make it enforceable, and that is a difference in kind.

Which speed test should you use in Belgium?

No tool is neutral, and the choice of server matters more than the choice of tool.

A test hosted by your own provider, such as the one from Orange Belgium or the one from VOO, measures the route as far as its own network. That is the best possible case, and it is useful: if even that test returns a poor figure, the problem is unarguably at their end. A general-purpose test such as Ookla's Speedtest or nPerf picks a server among hundreds, not always the nearest, and a badly chosen server can cost several hundred megabits per second on a perfectly healthy line. BIPT for its part offers a quality-of-service measurement app, aimed more at mobile coverage than at home fibre.

The method that holds up: cross-check two tools, note which server was used, run the test again. Three consistent figures beat one spectacular figure.

The speed in your contract is not the one in the advert

Here is the point almost no article covers, and it decides everything that follows.

Since EU Regulation 2015/2120 on open internet access, the contract for a fixed connection must state four distinct speeds rather than one. Its Article 4 lists them: the minimum speed, the normally available speed, the maximum speed and the advertised speed. The marketing figure, the one printed large on the poster with an "up to", is the last of the list, and the least binding of the four.

Categories set by Article 4(1)(d) of Regulation (EU) 2015/2120, applicable to contracts concluded or renewed since 29 November 2015.

At Proximus these values can be looked up address by address, and they vary enormously from street to street on copper lines: the distance to the exchange and the state of the cable set the guaranteed minimum, which can fall to a few tens of megabits on a long line. On optical fibre the gap between the minimum and the advertised figure narrows sharply, which is one of the least promoted arguments in favour of a fibre plan. Our article on upload speed provider by provider covers the other half of the subject, the half adverts mention even less.

What if the speed stays below the guaranteed minimum?

In order, and the order matters.

Build the file first: the six conditions above, over at least a week, with timestamped screenshots. Then send a written complaint to the provider, citing the contractual value you are not reaching rather than the marketing figure. That wording is what separates a support ticket from a non-conformity case.

If no answer comes, or none that satisfies you, the Ombudsman Service for Telecommunications takes over. Its intervention is free. BIPT itself does not handle individual disputes: it supervises the application of the regulations and steps in on that basis, which is not the same thing. And Article 4 of the Regulation sets a condition worth knowing in advance: the facts must be established by a monitoring mechanism approved by the national regulator. In other words, a Speedtest screenshot alone does not automatically trigger the sanction. It remains, nonetheless, the piece that opens the conversation, and mediation in practice usually settles for it when the measurements are numerous and consistent.

That leaves the fastest route. After six months of contract, Belgian law allows termination without reason and without charge, as we set out in our guide to cancelling and switching provider. On a persistently mediocre line, leaving often costs less than being right.

Brussels, Flanders, Wallonia: same method, different results

How you test does not change from one region to the next. What you are entitled to expect does.

Fibre rollout remains deeply uneven: on the order of 78 % of households passed in Brussels, 71 % in Flanders and 44 % in Wallonia according to mid-2026 figures, with notable gaps between sources depending on whether homes passed or homes actually connectable are counted. An address still served by VDSL will show a speed dictated by the length of its copper line, dropping sharply beyond a few hundred metres, with a guaranteed minimum set accordingly. An address on coaxial cable shares its capacity with the neighbourhood, which makes the 8 pm measurement particularly revealing. An address on fibre will return roughly the same figure at any hour. Check what actually serves your street in the BIPT coverage atlas before drawing any conclusion from a disappointing result.

And if the measurement confirms the line is at fault, the problem is no longer testing but choosing. The speed actually delivered depends first on the technology reaching your door, and only then on the brand printed on the bill. That is the order in which we compare offers in our ranking of the best fibre internet deals in Belgium, with the physical network behind each brand and the price after the promo. To turn the figures you obtain into concrete download times, our guide to fibre download speed does the rest of the work.