VPN Scoring Methodology
Every VPN on top10vpns is scored across six categories that count equally, from inputs the provider publishes about itself. No pay-for-placement, and no test results we did not produce.
The mean of the six category scores below, each counting equally. It is calculated at build time from those six numbers rather than typed in, so it cannot drift away from the categories it summarises, and the rank order on the site is simply the sorted result.
What the provider says it keeps, and which country's law it says that under. We record whether an independently audited no-logs policy appears on a provider's record, not a firm-by-firm audit register. Where a specific audit is documented in a provider's own materials, it is named in that provider's review rather than assumed here.
The protocols and protective features each provider publishes. These are recorded from provider documentation rather than tested here, so this category measures what a provider offers and commits to, not what we have verified on a bench.
We record one figure per provider: an average speed drop against an unprotected connection, running from 2 percent at the fastest we rank to 15 percent at the slowest. It is a single recorded average rather than a test log, and we hold no per-server, per-protocol or per-country breakdown behind it.
Which services a provider names among the ones it unblocks, and how many. These are provider claims rather than per-title results we produced, and a platform block lands on individual server addresses, so naming a service is a statement of intent rather than a guarantee on a given evening.
The size of the network and what a subscription actually includes. Everything here is a figure or a list item the provider publishes about itself.
What the rest of the record costs. Value means quality relative to price rather than the lowest price, which is why the cheapest provider we list does not hold the highest value score.
VPN companies cannot pay to change their position. Rank is derived in code from the six category scores, so moving a provider up the page would mean changing a score that is published and inspectable on its own review.
Scores are built from specs, features, protocols, prices and jurisdictions that each provider publishes about itself. Anyone can open the provider's own page and check the same inputs we used.
All six categories count the same. That is a choice rather than a claim about what matters most to you, so if privacy or speed is your whole reason for buying, read that category score rather than the overall.
Specs and pricing carry the date they were last checked against provider sources, and a provider re-checked on its own carries its own date. We do not claim an update schedule we cannot evidence.
Where a provider has a documented logging incident, its review says so and keeps saying so after reforms. PureVPN's 2017 disclosure to the FBI and IPVanish's 2016 incident under previous ownership are both recorded on their pages, alongside the audits that followed.
Reviews carry an explicit warning banner where the thing being reviewed is not what a reader expects: browser proxies sold as VPNs, peer-to-peer exit node networks, closed protocols that log by policy, and ownership that is worth knowing about before you install.
Overall = (privacy + security + speed + streaming + features + value) / 6
Each category runs from 0.0 to 10.0. The average is worked out in tenths and rounded half up to one decimal place, so a boundary case such as 8.85 lands on 8.9 rather than depending on how a computer happens to store the number. The overall is derived from the six categories every time the site is built and is never stored separately, which is what stops it disagreeing with them. The colour on a score follows four bands: 9.0 and above, 8.0 to 8.9, 7.0 to 7.9, and below 7.0.
The ten ranked providers each carry six category scores assigned individually. The wider A to Z catalog covers many more services, and for those we hold an overall, a speed, a privacy and a streaming score; the remaining categories are derived from the figures we do hold rather than assigned separately. A provider's review page and its catalog entry always show the same numbers.
Providers that have shut down keep their review page and their stored score, because the brand still gets searched for, but they are dropped from the ranked list and labelled rather than presented as something you could buy.