How we test email deliverability

Every deliverability number on this site comes from our own controlled inbox placement tests. This page is the short version; the full methodology — the complete protocol, the calibration plan and the formula changelog — ships with our first published test cycle (cycle 0).

The test, in one paragraph

We open a real sending account on each service, authenticate a dedicated domain with SPF, DKIM and DMARC, and send the same campaign archetype to a controlled seed list spanning Gmail, Outlook and Yahoo mailboxes. For every seed we record where the message landed: inbox, a tab like Promotions, spam, nowhere at all, or bounced at send time. Identical protocol for every service in a cycle — same content, same seed list, same day window.

The score

The Deliverability Score is computed from those measured rates by a published formula. Inbox placement is the anchor; tab placement earns half credit; spam, silent loss and bounces subtract. The formula is versioned — currently v0.1-pre-calibration — and when a version changes, every historical snapshot is recomputed and the change is logged publicly.

Two things the score is not:

It is not an editorial opinion. No feature list, interface or price affects it.
It is not a guess. A service we have not tested shows “Not yet tested” — never an estimate, never a zero.

What we publish and what we do not

We publish the rates, the score, the per-provider breakdown and the test date — every deliverability figure carries the month it was measured. We do not publish seed addresses, sending domains, account identifiers or send-timing patterns: the protocol is public, the infrastructure is confidential, because a service that could recognize our test traffic could game it.

Two scores, kept apart

Reviews also carry a Platform Score — an editorial rubric over 17 weighted dimensions covering features, pricing and support. The two scores never mix. A service with brilliant automation and a 40% spam rate keeps its 40% spam rate at the top of its review.

The full methodology page will document the rubric weights, the evidence ladder and the trigger system in detail. Until cycle 0 publishes, the current rankings show an empty state rather than provisional numbers.