Editorial Policy

Testing Window

Every service reviewed here gets at least 30 days of active use before anything goes to draft. For services included in comparisons the window is usually longer, because testing overlaps with other active subscriptions running in parallel. Speed tests run daily against the same server endpoints, at the same times, from the same Beacon Hill connection, so numbers being compared come from a controlled baseline rather than scattered spot checks.

What Gets Measured

Throughput is the starting point, but variance across sessions matters more than peak numbers. I track how results shift over a multi-week window, note when latency to the nearest server climbs without an obvious ISP cause, and record which protocol was active for each test. When a provider advertises a kill switch, I test whether it fires on an unexpected disconnect and whether that protection survives a reboot on both Windows and macOS clients. DNS leak behavior under different protocols, and split-tunnel configurations at the edges of their documentation, get tested specifically because they surface in actual daily use.

Billing and Renewal

Renewal experiences are part of the review. A price increase at renewal without prior notice gets noted. A trial period that ends in a way requiring a calendar alert to catch goes in the writeup. Refund requests are documented for any service where one was sought, with a note on how long the process took and whether it needed escalation. A billing dispute that takes four days and three ticket threads to close is a product experience, and it gets reported that way.

Affiliate Relationships

Some links on this site are affiliate links. A click and a purchase earns a commission at no extra cost to the reader. That arrangement has no bearing on which services get recommended or how they are scored. Providers with affiliate deals who underperform get reviewed accordingly. Services that were previously recommended have been removed when product quality dropped, regardless of whether an affiliate deal was in place. Sponsored content would be labeled explicitly; ungated editorial is not sponsored.

Corrections

If a technical claim turns out to be wrong, or if a provider changes something material after a review publishes, the review gets updated with a datestamp at the top noting what changed and why. Corrections are not buried in a footer note. For factual errors or outdated information, the contact page goes directly to me.

Scope

Everything here reflects testing from one location in Seattle on CenturyLink fiber, on consumer and developer hardware. Results will vary on different ISPs or in different geographic regions. When something looks like a routing quirk specific to this connection rather than a general product issue, the review says so. Enterprise procurement, corporate IT policy, and provider-side infrastructure auditing are outside the scope of this site.