Research and commercial separation
Research pages contain no provider affiliate links, deal cards, provider recommendations, or purchase calls to action. Their navigation and footer point to research, data, methods, authorship, and contact information. Affiliate click tracking does not run on these pages.
DoVPN's wider site is commercially funded. That funding does not decide which records enter a dataset, how a field is classified, whether an inconvenient result is published, or how a correction is handled. Provider claims remain labelled as claims unless independent evidence supports a stronger description.
Scope is fixed before collection
Each project starts with a written question, inclusion rules, exclusion rules, field definitions, collection dates, and a list of known limitations. The scope is narrow enough that another person can tell why a record was included and why a similar record was not.
A dataset does not quietly expand to fit a headline. If the scope changes, the release notes record the change and the version number changes with it.
Collection keeps the evidence attached
Structured observations retain their source URL, collection or verification date, and the evidence state used by the project. Unknown, unavailable, not applicable, and not located are separate values. Missing data is not inferred from a provider's reputation or from a similar product.
Time-sensitive research, including VPN price tracking, also records the market, currency, displayed plan length, upfront total, renewal terms, capture time, and offer conditions. This prevents a monthly equivalent from being mistaken for the amount charged at checkout.
Checks run before a release is frozen
Automated checks validate required fields, allowed values, dates, counts, duplicates, and generated statistics. Manual review then checks the largest changes, ambiguous evidence, and a sample of unchanged records. Headline numbers must be generated from the released data, not copied into an article by hand.
The published files receive checksums. A site build uses a pinned local snapshot so a later source edit or outage cannot silently change an existing analysis.
Releases, repositories, and identifiers
A public release includes the structured data, a plain-language methodology, computed statistics, and a change log. GitHub holds the versioned source release. A data repository supplies the preferred DOI and preserves the files independently of the website.
Repository mirrors are linked only after publication. Their metadata names the same author, release date, license, version, source release, and preferred DOI. The research landing page exposes Dataset structured data with the author ORCID, DOI identifiers, license, variables, and download location.
Corrections and later versions
Small wording corrections can update the analysis page without replacing a dataset release. A change to records, field definitions, or computed statistics requires a new version. The change log says what changed and whether previous conclusions still hold.
Send corrections, replication notes, and media questions to steve@dovpn.com. Please include the dataset version, record or passage, supporting source, and the date you checked it.