TL;DR
When Speedtest was a Flash-based speed test, it ignored the fastest 10% of the measurements. Today, Ookla's Speedtest still does that. The consequences of this are:
- Almost all speed tests "measure" 90% of the true speed
- All ISPs deliver a 12% higher speed than advertised
- To ensure your home network is not the bottleneck, you should have at least 12% oversized hardware
Contents
- Advertised speed
- Measuring your internet speed
- Speedtest's measurement method
- HTML5, the alternative for Flash
- Most modern speed tests mimic Speedtest
- First consequence
- The customer measures 90%
- Second consequence
- What about the hardware?
- Third consequence
Advertised speed
The advertised speed is the internet speed that the ISP once advertised. This is the download speed that is stated in the contract with your internet service provider (ISP).
Some examples of advertised speeds are:
Measuring your internet speed
Almost everyone who has internet will check their internet speed at some point. The most used speed test for such a check is undoubtedly Ookla Speedtest.
Most ISPs recommend using Ookla Speedtest (or a clone of it), although in our opinion Ookla's Speedtest is not the best. That's why we recommend other speed tests.
Speedtest's measurement method
The first occurrence of Ookla's Speedtest® in the Wayback Machine is from November 7, 2006. At the same Wayback Machine we discoverd the page wiki.ookla.com/test_flow.
Here it is stated that:
(...) The fastest 10% and slowest 30% of the slices are (...) discarded (...)
Speedtest ignores the fastest 10% of measurements
The consequence of ignoring the the fastest 10% of measurements is that Speedtest measures 90% of the true speed.
Why measuring 90% instead of 100%
The big question is of course why Speedtest measures 90% instead of 100% of the download speed. The answer is given at the same page of the Wayback Machine: Since we are measuring data transported over HTTP via Flash there is potential protocol overhead, buffering due to the many layers between our application and the raw data transfer and throughput bursting due primarily to CPU usage. This accounts largely for dropping the top 10% and bottom 10% of the samples. We also keep our default test length short for the user experience, and compared to this duration the ramp-up period is fairly significant driving us to eliminate another 20% of the bottom result samples.
HTML5, the alternative for Flash
As could be read at our page regarding the Ookla Speedtest, an alternative for Flash was available in 2011.
At our page regarding SpeedOf.Me we cited: In December 2011, we created the first HTML5 broadband speed test. SpeedOf.Me (Speed of me), is a pure HTML5/JavasScript speed test which relies only on browser core technologies.
The biggest difference with Ookla's Speedtest, however, is not the use of modern technology instead of the outdated Flash technology, but the measurement of the actual speed without ignoring the fastest 10%.
Most modern speed tests mimic Speedtest
Speedtest is such an influential speed test that most new speed tests make sure their results match Ookla's Speedtest.
We concluded earlier at the page Your true internet speed that a modern speed test like Cloudflare also takes (...) the 90th percentile speed (...).
Mainstream speed tests
We consider speed tests that follow the Ookla Speedtest methodology as mainstream speed tests. Notable examples of modern mainstream speed tests including Ookla Speedtest (!!!) are:
First consequence
Almost all speed tests try to be as accurete as Ookla's Speedtest. All these speed tests "measures" 90% of the true speed.
At the page Accurate Speed tests we have marked these speed tests as Mainstream.
The customer measures 90%
Since almost all speed tests give 90% of the actual speed, ISPs have a problem.
If an ISP advertises 100 Mbps and the consumer consistently measures 90 Mbps, the ISP is inundated with complaints.
To prevent this, ISPs deliver a higher speed than they advertise. Purely mathematically, ISPs should then deliver 1/90% = 111.111%. However, we have the idea (based on tests One hundred point eight and Two hunderd and one point six, among others) that ISPs deliver 112%. We call this your true internet speed.
Second consequence
All ISPs deliver a 12% higher speed than advertised. This is to prevent customers from complaining about a structurally too low speed test.
What about the hardware?
Suppliers of network equipment such as routers indicate which speeds their equipment can handle.
Common routers are 10/100/1000 Mbps routers. For consumers who have a lower speed than 1000 Mbps, such a router is fine.
However, if you have a 1 Gbps subscription, the router used can easily be the bottleneck. Although it is suitable for processing 1 Gbps, the consumer measures 90%.
We experienced this ourselves when we tested the RealSpeed speed test. At the time, we had a 100 Mbps router in our network. The measurements then remained somewhere around 92 Mbps. Only with a realistic speed test such as SpeedOf.Me or True Speed at a Glance did we see measurements of 100 Mbps.
Now that more and more consumers are switching to 1 Gbps, the hardware manufacturers are in the same position as the ISPs. A 1 Gbps router must be able to process 1.12 Gbps.
It comes as no surprise that when we were offered to upgrade to 1 Gbps -which we didn't- we were told that a 1120 Mbps modem would be included. We are now confused. On the one hand, it seems that hardware manufacturers stick to the actual internet speed and do not secretly deliver a higher speed than advertised, on the other hand, we do not find any advertisements that recommend a 1120 Mbps router.
For now, we assume that hardware manufacturers advertise the actual speed (although we are not 100% sure about this).
Third consequence
To make sure that your home network is not the bottleneck, make sure that you buy hardware that can achieve at least 12% faster speeds than the speed that the ISP advertises. Only then will you be able to reach the maximum internet speed.