How We Evaluate Proxy Services
A score is only useful if readers can see what went into it. When a provider is reviewed, the page should identify the test date, location, access network, client and the limits of the test. The framework below is a checklist for future service reviews, not a claim that every listed provider has already been measured.
Evaluation areas
| Area | What to record | Suggested weight |
|---|---|---|
| Speed | Repeated download/upload results and sustained throughput | 30% |
| Stability | Failed connections, unexpected disconnects and recovery time | 25% |
| Latency | RTT, jitter and packet loss | 15% |
| Region coverage | Availability of the locations actually needed | 10% |
| Value | Traffic allowance, simultaneous devices and effective monthly cost | 10% |
| Client compatibility | Whether common desktop and mobile clients can use the profile | 5% |
| Privacy and transparency | Published policy, operator information and data handling | 5% |
These weights are an editorial framework, not an industry standard. A review should explain any adjustment rather than presenting the final score as objective truth.
Test conditions matter
Record the date, local time, city/region, ISP or access network, device, client version, profile and node. Test at least one ordinary period and one busy period. Repeat measurements and report the typical result rather than selecting only the highest number.
Public data is context, not a provider score
Cloudflare Radar and M-Lab NDT can help describe the broader network conditions of a region. They do not measure the performance of a specific proxy provider and should not be presented as if they did.
Advertising and editorial judgment
Paid placements must be labelled and kept separate from test results. An ad does not improve a service's score, and a review should not imply hands-on testing unless the test was actually run.