To ensure our VPN speed test findings are as accurate as possible, we follow a strict scientific process refined over many years.
Our methodology is designed to extract just raw data, to remove any risk of bias or human error in our results.
To find the fastest VPNs, we measure:
- A VPN’s average download speed when connected to a nearby server.
- A VPN’s speed performance when connected to multiple far-away locations.
- How reliable and consistent a VPN’s speed results are.
Here’s a more in-depth overview of the measurements we take when finding the fastest VPN, along with how much of our overall rating they make up:
1. Local Download Speed: 45%
Minimum Requirement: Speed loss of 20% or less.
We Recommend: Speed loss of 10% or less.
As you’ll be connecting to a local VPN server most often, we prioritize short-distance speeds in our tests.
We’ve developed a test to measure the average local download speed of every VPN.
Here’s how we test for a VPN’s local speed performance:
- We perform this test every Friday of each month.
- We use a virtual machine with just the VPN we’re testing installed.
- Using Ookla’s Speedtest app, we record our download speed three times without the VPN running.
- We launch the VPN and connect to the nearest server location. We record our download speeds three times.
- We repeat the above steps for every VPN.
- At the end of the month, we work out the mean result without the VPN running and while connected to the nearest server location.
- We calculate the download speed loss for each VPN. The lower the speed loss, the better.
- The top-performing VPNs are ranked from 1-10. Depending on their position, each VPN service is given a rating from 5.5 to 10, with each position being a difference of 0.5.
- We update this page with our new findings.
2. International Download Speed: 45%
Minimum Requirement: Speed loss of 50% or less.
We Recommend: Speed loss of 20% or less.
VPNs can be used to watch geo-restricted Netflix libraries and play games on servers from the other side of the world. Because of this, it’s important that a VPN also has fast international speeds.
The extra distance traveled will have an impact on your download speeds. Generally speaking, the further a VPN server is from your physical location, the slower your speeds will be.
We’ve formulated a standardized and unbiased test to calculate the average international download speed of every VPN.
Here’s how we test for a VPN’s international speed performance:
- We perform this test every Friday of each month.
- We use a virtual machine with just the VPN we’re testing installed.
- Using Ookla’s Speedtest app, we record our download speed three times without the VPN running.
- We launch the VPN and connect to seven server locations from across all five inhabited continents of the world — Africa, Asia, Europe, North America, Oceania, and South America.
- We record our download speeds from each location three times.
- We repeat the above steps for every VPN.
- At the end of the month, we work out the mean result without the VPN running and while connected to all far-away locations.
- We calculate the download speed loss for each VPN. The lower the speed loss, the better.
- The top-performing VPNs are ranked from 1-10. Depending on their position, each VPN service is given a rating from 5.5 to 10, with each position being a difference of 0.5.
- We update this page with our new findings.
3. Consistency: 10%
Minimum Requirement: Difference of up to 10Mbps on local speed tests and 40Mbps on international results.
We Recommend: Difference of up to 3Mbps on local speed tests and 10Mbps on international results.
It’s not enough for a VPN just to have fast peak performance. It needs to deliver fast speeds reliably, across every server location.
We’ve developed a method to calculate every VPN’s consistency. The more fluctuation present in a VPN’s speed test results, the lower we rate it in this category.
Here’s how we test for a VPN’s speed reliability:
- We complete this calculation on the final Friday of every month after all speed test results have been completed.
- We work out the standard deviation of each VPN’s local download speeds. A lower standard deviation means more reliable speed performance.
- Each VPN is ranked from 1-10. Depending on their position, each VPN service is given a rating from 5.5 to 10, with each position being a difference of 0.5. This constitutes 5% of the overall consistency rating.
- We repeat the above steps for each VPN’s international speed results. This tells us how reliable a VPN’s speeds are across every global server location, not just the consistency per country. This constitutes 5% of the overall consistency rating.
- We update this page with our new findings.
Control Variables
Here are the steps we have in place to ensure our speed test results are the most reliable findings available:
Dedicated 100Mbps Broadband Line
We have a dedicated 100Mbps fiber-optic broadband line that is only used for conducting our VPN speed tests. This means that our speed is not slowed down by any external factors, like bandwidth-consuming activities from anyone else on the network. Any drops in speed are a direct result of the VPN running.
Consistent Infrastructure & Equipment
All of our speed tests are done using a Lenovo ThinkPad PC running Windows 10 and on an iMac running MacOS Big Sur. These devices are connected to the broadband line using a Cat7 ethernet cable.
By keeping all of our equipment consistent, we can always conclude that any speed changes are as a result of the VPN, and not any issues with our test devices.
Only Use the VPN’s Preferred Protocol
For our speed tests, we use the VPN with the fastest protocol available. This allows us to give an accurate representation of the speeds you can expect while using the VPN.
For the tests on this page, the protocols used were as follows:
- CyberGhost, Mullvad, NordVPN, Private Internet Access, Surfshark — WireGuard
- ExpressVPN — Lightway
- Hotspot Shield — Catapult Hydra
- HideMyAss, IPVanish, PrivateVPN — OpenVPN (UDP)
Same Speed Testing Tool
We use Ookla’s popular speed test tool for all of our tests. We elected to use this because a representative from Ookla assured us that it has “strong mitigations in place to prevent such behavior [VPNs prioritizing traffic to Ookla].”
None of the other speed test tools would provide this guarantee. In turn, this rules out the possibility of a VPN speeding up its service when it detects a speed test, which would give unreliable results.
Tested Over a Long-period
To minimize the risk of random and anomalous results, our final speed test result is the average from a series of tests, performed over a month. We tested all VPNs at regular six-hour intervals, equaling four different times per day. In total, we manually undertook thousands of VPN speed tests, ensuring our results most accurately represent the speeds of the fastest VPNs.
Research Limitations
Although we have taken steps at every stage to keep our speed test results as accurate as possible, there are still some unavoidable limitations to our research.
These limitations may result in a VPN performing differently when you test it yourself. These include:
Speed Fluctuations
To quantify speeds and account for speed fluctuations, we would have to test every single VPN server during every second of the day.
To counter this, we pace our testing out throughout the day. Tests are performed four times a day, with a six-hour break between each test. Although not perfect, this allows us to test each VPN during both peak and quiet hours of the day, giving us an accurate, but not perfect, representation of typical speeds.
Upload Speed & Ping
The fastest VPN is not necessarily the fastest VPN for all activities. For example, a VPN with fast download speeds but slow ping time will be great for streaming, but poor for online gaming.
For this page, we’ve chosen to focus solely on download speed. This is because the vast majority of online activities rely on a fast download speed, and fewer need a speedy upload speed or ping rate.
If download speed is less important for your use case, you can read our dedicated recommendations for the best VPNs for gaming and best torrenting VPNs.
Varying Device Performance
Different devices, with different network cards and performance capabilities, can yield dramatically different speed test results. For example, desktop PCs will often have larger and faster wireless cards than mobile devices. This means that you may not always experience the same results as us, depending on your device.
To address this, we use two different test devices — one Lenovo ThinkPad and one Apple iMac. These devices are made by different manufacturers, which gives us speed test data from the two major desktop operating systems.
Overall, we have tried our best to eliminate all possible factors that could hold back our investigation into the fastest VPNs. However, with the VPN industry constantly changing, just keep in mind that we can only ever provide a snapshot and not an absolute answer for the fastest VPN.
