When the internet feels slow, almost everyone opens a speed test site and presses start. The number that appears then becomes the basis for blaming the provider. Unfortunately, a test run carelessly often produces a figure that does not reflect the actual situation.
Below is how to test a connection with a method whose results can be trusted, followed by help in reading each figure it produces.
Setting up a valid test
Use a LAN cable where possible
This separates the performance of the internet service from the performance of Wi-Fi. If the cabled result is already low, there is no point fiddling with the wireless router.
Stop other network activity
Switch off streaming, downloads, cloud sync, and automatic updates on every device. One photo backup uploading on another phone can halve the result.
Disable VPNs and blockers
A VPN routes traffic through an extra server; the result then measures that path, not your connection.
Choose the nearest server
Testing against a server on another continent measures the quality of the international route, not your subscription's capacity. For users in Indonesia, choose a server in Jakarta or the nearest city.
Repeat at least three times at different hours
Test morning, afternoon, and evening across two days. The pattern is what is useful, not a single figure.
Understanding each figure
Download speed
Shows how fast data flows from the internet to your device. This is the figure most relevant to streaming, browsing, and downloading files. Rough requirements:
| Activity | Requirement per stream |
|---|---|
| Browsing and social media | 3–5 Mbps |
| HD video calls | 3–4 Mbps |
| 1080p video streaming | 5–8 Mbps |
| 4K video streaming | 20–25 Mbps |
| Online gaming | 3–6 Mbps (latency matters more) |
Upload speed
Shows how fast data leaves your device. Often ignored, yet it determines video call quality, file uploads, and cloud backups. On asymmetric connections such as coaxial cable, this figure can be far smaller than the download.
Ping, or latency
The round-trip time of a signal to the server, in milliseconds. A low figure means a quick response. Latency determines the "feeling of speed" when opening a page far more than a large download figure does.
Jitter
The variation between pings. A high value means unstable latency, which shows up as choppy audio on a video call even when the download speed looks fine. Below 10 ms is considered excellent.
Packet loss
The percentage of data packets lost in transit. Ideally zero. Loss above 1 per cent is already enough to disrupt real-time calls and games. The most common cause on Wi-Fi is interference or a weak signal.
A frequent mistake: many people conclude their connection is bad from a download figure lower than their plan. Yet everyday complaints such as stuttering calls or pages loading slowly come more often from jitter and packet loss than from bandwidth.
Mbps versus MB/s: the classic confusion
Providers sell in megabits per second (Mbps). Browsers and download managers display megabytes per second (MB/s). One byte is eight bits, so:
- A 100 Mbps plan → a maximum download of about 12.5 MB/s
- A 50 Mbps plan → about 6.25 MB/s
- A 20 Mbps plan → about 2.5 MB/s
Add protocol overhead of around 5–10 per cent, and the real figure usually sits slightly below that calculation. This is normal and not a sign of trouble.
When to report to your provider
Gather evidence before contacting customer service. An effective report contains:
- At least six test results at different times across two days.
- All tests carried out over a LAN cable.
- The name of the test server used, ideally the same one each time.
- A note of any other activity running during the tests (ideally none).
- Screenshots of the results, complete with the time.
If the cabled result matches your plan but Wi-Fi is far lower, the problem is in your wireless network and the provider cannot help much. For that, see the guide to fixing slow Wi-Fi.
Three kinds of test with different purposes
An ordinary speed test
Measures momentary maximum capacity by sending as much data as possible over a few seconds. Useful for verifying that the subscribed plan is being delivered, but it does not describe everyday experience.
A latency-under-load test
Measures latency while the network is busy. This is often the best explanation for the complaint "the internet is fast but feels slow". If latency spikes sharply while a download runs, your network is suffering queue build-up, which can be improved with queue management settings on the router.
A long-duration stability test
Running a continuous ping for several hours to catch faults that appear occasionally. For problems that are hard to reproduce, this kind of test is the most useful.
Testing at home versus at a business
At a business, the average figure matters far less than the worst experience. Ten customers having a poor connection at the busy hour has more impact than a high peak speed at a quiet one.
For venues, we therefore suggest testing under three conditions: the quietest hour, the busiest hour, and while operational equipment such as the till or the cameras are active. The gap between the three shows whether the problem is subscription capacity or bandwidth sharing.
The role of DNS, often overlooked
Every time you open a page, your device has to translate a domain name into an IP address. If the DNS server in use is slow, pages feel slow to open even when a speed test shows a large figure.
You can measure it with the nslookup or dig command and watch the
response time. Under 50 milliseconds is considered good. If it is far slower, trying an
alternative DNS server is a cheap fix that is easy to undo.
A suggested recording format
| Time | Method | Down | Up | Ping | Jitter | Notes |
|---|---|---|---|---|---|---|
| Monday 08:00 | Cable | - | - | - | - | No other activity |
| Monday 20:00 | Cable | - | - | - | - | Peak hour |
| Tuesday 08:00 | Wi-Fi, main room | - | - | - | - | 3 metres away |
| Tuesday 20:00 | Wi-Fi, back bedroom | - | - | - | - | Two concrete walls |
A simple table like this turns a subjective complaint into data that can be acted on, by you and by the provider alike.
Closing
A speed test is a diagnostic tool, not a verdict. Its value comes from the pattern, not from a single figure. Testing methodically over two days gives you a far more useful picture than ten random tests in one afternoon.
Frequently asked questions
Why do speed test results differ every time?
Because the result is affected by the destination server, network congestion at that moment, how the device is connected, and background processes. That is why one test is never enough to conclude anything.
What is the difference between Mbps and MB/s?
Mbps is megabits per second, the unit of network speed. MB/s is megabytes per second, the unit used when downloading files. One byte equals eight bits, so 80 Mbps is roughly 10 MB/s.
What counts as a good ping?
Below 30 ms is excellent for almost any use, 30–60 ms is good, 60–100 ms is still comfortable for browsing and streaming, above 150 ms starts to be noticeable on video calls and games.
Is it normal for speed to be lower than the provider promised?
Some gap is normal because of protocol overhead and network conditions. But if results are consistently far below the plan at various times and tested over cable, that is worth reporting to the provider.
