Review standards: the gates every page must clear

Most review sites describe their standards in adjectives. We publish ours as a checklist, because a standard you can't audit is a slogan. Below are the rules every review on this site must clear before it goes live — the same list our own build system enforces. If a published page violates one of these, that is a correction-worthy defect: write to us and we will fix it publicly.

The sixteen rules

1. Every deliverability figure matches our test data exactly, and the test date is shown next to it. Our deploy pipeline recomputes every published score from the underlying measurements and refuses to ship a mismatch.
2. Nothing that identifies our test infrastructure is ever published — no seed addresses, sending domains, account names or send schedules. A service that could recognize our traffic could game the benchmark.
3. The two scores never mix. The Deliverability Score is measured; the Platform Score is editorial. Neither borrows points from the other, and neither is ever relabeled as the other.
4. Exactly one pricing table per reviewed service, so there is a single place a stale price could hide — and a single place to fix it.
5. Competitor prices quoted in a review are re-verified at the source within 30 days of publication, or softened to qualitative statements.
6. The working notes behind each review — trigger verdicts, source URLs, scoring arithmetic — are kept as an internal audit trail for every page.
7. Reader-facing language only. We don't pad reviews with internal jargon or tooling references.
8. Institutional byline. Reviews are published by InboxRatio Editorial, not by invented personal authors with stock-photo faces.
9. Every factual claim traces to a primary source checked within 90 days, and feature claims are verified inside the product, not on the vendor's marketing page — tier gating routinely differs between the two.
10. External opinions about a service's deliverability are quoted with source, URL and date, and framed as context. They never move our measured score.
11. Every number is either traced to a source or explicitly marked as an estimate. There is no third category.
12. Each review shows the per-dimension breakdown behind its Platform Score, so you can disagree with a specific input rather than a mystery total.
13. Hands-on sections report what we actually did: elapsed times, friction ratings, and a comparison against our anchor product. A test we did not run is labeled a projection with its basis stated — never written in past tense.
14. Dates in prose are written out; machine-readable dates live in metadata.
15. No production-stack credits. What we publish stands under our byline.
16. Any commercial relationship is disclosed on the page, next to the links it affects. Affiliate status never changes a score — scores are locked before commercial terms are ever discussed.

The pre-publish sequence

Before a review ships it passes, in order: deliverability snapshot verified against the data store → infrastructure-confidentiality scan → vendor source verification → multi-source cross-check → hands-on protocol documented → in-app feature audit → external-sources section dated → disclosure check → numeric audit → scoring breakdown → parity check against our anchor review → final editorial read. A failed gate blocks publication; there is no override switch for convenience.

What we don't promise

Honesty about limits is part of the method, so here are ours:

• Gates catch predictable errors, not all errors. They do not protect against failure modes we haven't seen yet. When a new one appears, we add it to the published list on our sources page.
• A measured score is a sample, not a guarantee. Inbox placement varies by sender history, content and audience. Our tests hold those constant to make services comparable — your results with your list will differ.
• Freshness has a horizon. Deliverability data older than nine months gets a visible staleness banner. Between test cycles, a provider can improve or regress before we re-measure.
• We can't test what we can't access. Services we hold no account on are listed as "Not yet tested" rather than scored from documentation.

For how the tests themselves work, read how we test email deliverability.