Skip to content
$_ setuptracking

Methodology

How recipes on this site are tested, what gets included, and what gets cut. Voice rule: mechanism over recipe, real numbers, no marketing fluff. If a claim isn't verifiable from primary source code or first-party docs, it doesn't ship.

Test environment

Every install recipe is tested on a clean Hetzner CX22 VPS in Falkenstein (EU). Operating system: Ubuntu 24.04 LTS. Reverse proxy: Caddy 2.7 (or nginx where the upstream prefers it). Public DNS via Cloudflare proxy mode (orange cloud). All recipes are re-tested on each major-version bump of the underlying tool.

Pricing claims

Hetzner SKUs (€4.51/mo CX22, €8.21/mo CX32) are the published Hetzner Cloud prices in EUR including 19% German VAT. Cloud-plan pricing for Plausible / Matomo / Umami / PostHog / GoatCounter is taken from each vendor's public pricing page on the date of last revision (footer of each article). When pricing changes, the article shows the date the recipe was last verified.

Source-of-truth hierarchy

  • Tier 1 — tool source code on GitHub. When article and docs disagree, source wins. Hash composition, salt rotation, schema columns, default config values are all read from source, not from blog posts.
  • Tier 2 — first-party docs. developer.matomo.org, plausible.io/docs, umami.is/docs, posthog.com/docs, goatcounter.com/help.
  • Tier 3 — regulatory primary sources. CNIL, EDPB, DPC Ireland, Garante, Datatilsynet, Tietosuoja published rulings — cited by URL where claims involve legal posture.
  • Tier 4 — community / blog posts. Used to confirm independent reproduction, never as the primary source for a technical claim.

Voice rules

  • Real numbers or no numbers. Stat without source = stat without claim.
  • Mechanism before recipe — explain how it works before showing how to do it.
  • Every recipe has a “When NOT to use” section.
  • Never promise “fixed within X days” / “verified manually” / “hand-tested.” If we can't prove it, we don't claim it.
  • Active voice, contractions allowed.

Article review process

Tier 1 articles (pillars + revenue spokes) go through a published 11-phase procedure: 5-agent independent review (editor, SEO, domain-tech, marketer, keyword research) → concept sign-off → rewrite (factual fixes / structural / 6+ inline visuals / JSON-LD schema / link mesh / lead magnets / voice cleanup) → Indexing API submit → 3-agent fact-check against primary sources → critical+major fixes → visual review. Trust score target: 9/10 across all relevant tech domains. Articles that don't hit 9/10 don't ship.

Errata

Found a bug, broken command, dated price, or factual error? Email us — corrections published with attribution unless you ask not to be named. Articles show a “Last verified” date in the byline aside; major rewrites bump the version pin in the “Tested on” line so you can see exactly which tool versions the snippets were tested against.