InboxRatio
Check your domain

Glossary term

Inbox placement test: measuring where mail actually lands

What an inbox placement test is

An inbox placement test is a controlled experiment that answers the question sending metrics cannot: of the mail that was accepted, how much reached the inbox? The tester sends a defined message to a seed list, meaning mailboxes they control across Gmail, Outlook, Yahoo and other providers, then opens every mailbox and records the outcome per seed: inbox, a categorized tab, spam, missing, or bounced at send time. Counts become placement rates per provider.

The reason the experiment exists is a structural blind spot. SMTP ends at acceptance — the receiving server takes the message and the transaction closes. Spam-folder filtering happens after that, inside the provider, and no provider reports per-message folder outcomes to senders. Your platform's "delivered: 99.2%" is a statement about accepted connections, fully compatible with half the campaign sitting in spam. Inbox placement is the metric behind the metric, and a placement test is the standard way to observe it.

It is also the method behind this site's numbers. InboxRatio's Deliverability Scores come from placement test cycles run under one protocol, summarized on how we test, and cycles are announced as they complete. Until a service has been through a cycle, it shows "Not yet tested" rather than an estimate.

How a placement test differs from a delivery rate

The two numbers describe different stages of one journey, and confusing them is the most consequential misreading in email analytics.

Delivery rate is send-side arithmetic: accepted messages divided by attempted, with hard bounces and soft bounces as the misses. It is measured by your platform, available instantly, and covers your entire audience. What it measures stops at the receiving server's front door.

Inbox placement rate is receiving-side observation: of accepted messages, the share that landed in the inbox rather than spam, a tab, or nowhere. It requires access to mailboxes, arrives only for the seeds you control, and is a sample rather than a census. What it measures is the thing your program actually depends on.

The failure modes of each are instructive. Delivery rate fails by flattery: it stays high while placement collapses, because filters prefer accept-then-junk over rejection. Placement testing fails by sampling — seeds are not your audience, and their neutrality is both the method's power and its limit. Mature programs read both, plus the receiver's own view via Postmaster Tools, and treat disagreement between them as the diagnostic signal it is.

How an inbox placement test works

A defensible test is a chain of controlled steps; weaken one and the rates stop meaning anything.

Fixed message, fixed conditions. The test defines a campaign archetype (realistic content of a stated kind) and holds it constant. In comparative testing, our use case, the same archetype goes through every service under test with the same authenticated sending domain setup, so the service is the isolated variable. Authentication must be clean before the test means anything; a failed SPF or DKIM pass would be measured as the service's placement problem when it is the tester's configuration problem.

A managed seed set. Coverage across the providers that matter, counts per provider large enough for rates, mailbox state kept neutral. The seed list entry covers the instrument in detail.

A settling window, then reading. Placement is read after enough time for greylisting retries and deferrals to resolve — reading too early counts delayed mail as missing. Every seed's outcome lands in one of five states: inbox, categorized tab (see Promotions tab), spam, missing, bounced.

Scoring. Rates per provider aggregate into a headline number by an explicit formula. Ours is published and versioned on our methodology page: inbox placement anchors the score, tab placement earns partial credit as delivered-but-categorized, and spam, silent loss and bounces subtract. When the formula version changes, historical snapshots are recomputed and the change is logged.

Disclosure discipline. The protocol is public; the infrastructure is not. Seed addresses, sending domains, account identifiers and timing patterns stay confidential, because a tested service that can recognize test traffic can special-case it, and the test would then measure the special-casing.

Placement tests and your deliverability

Three uses justify the effort for a working sender.

Detection. A placement slide is invisible in delivery metrics and slow to surface in engagement data — open rates decay over weeks as spam-folder mail goes unseen. A standing seed row on major sends converts that lag into same-day detection, with the provider breakdown pointing at where to dig: a Gmail-only shift reads differently from a collapse everywhere, and a Postmaster Tools check tells you whether domain reputation moved with it.

Attribution. Because seeds are engagement-neutral, before/after placement around a change (new template, new dedicated IP, new list source, a warm-up milestone) isolates sender-level cause from audience-level noise better than any dashboard trend.

Verification. Deliverability claims, your vendor's or anyone's, are checkable against controlled placement data. That is the premise of our rankings: identical protocol per service per cycle, published rates, and an empty state rather than provisional numbers until cycle data exists.

Limitations and failure modes

Extrapolating to your audience. Seed rates isolate the sender-and-content component; your subscribers add engagement history on top, for better or worse. A placement test is a controlled baseline, not a forecast of your next campaign's opens.

Sample sizes that cannot carry the claim. A provider rate built on a handful of mailboxes has error bars wider than the differences being reported. Any published test that hides its seed counts and protocol (ours gates publication on both) is offering an impression, not a measurement.

Reading before the dust settles. Greylisting, throttled deferrals and slow queues mean early reads count in-flight mail as lost. The reading window is part of the protocol, not an inconvenience to skip.

One-off tests treated as verdicts. Filters are adaptive and reputations move; a single test is one draw. Cycles (repeated, identical, dated) are what produce trend-quality data, which is why every deliverability figure on this site carries the month it was measured.

Unrepresentative test mail. Testing with a stripped-down "test message" measures the placement of test messages. The archetype must resemble the mail whose placement you care about — real structure, real links, real unsubscribe furniture per the Gmail requirements.

Ignoring the bounced state. Mail refused at the gateway never had a placement to measure. Recording bounces separately keeps a send-side failure from masquerading as a filtering result — and points you at the SMTP reply codes that explain the refusal.

Related terms

Seed list, inbox placement, Promotions tab, Postmaster Tools, sender reputation, hard bounce, soft bounce, deferral.

Frequently asked questions

What does an inbox placement test measure? The post-acceptance outcome of a send: what share of accepted mail reached the inbox versus a categorized tab, the spam folder, or nowhere, observed through controlled mailboxes across providers.

Why is my delivery rate high but my placement bad? Because delivery rate stops at acceptance and filters prefer accepting then junking over rejecting. Placement into spam happens inside the provider, invisible to send-side metrics — which is the entire reason placement testing exists.

How often should I run placement tests? For senders: a standing seed row on every significant send makes placement a continuous metric instead of an occasional audit. For comparative purposes: on a fixed cycle with an unchanged protocol, which is how our test cycles run.

Can a placement test predict my campaign's open rate? No — placement is a gate, not a driver. It bounds what your audience can see; engagement history, subject lines and timing decide what they do see. Seed neutrality deliberately excludes those factors.

Does tab placement count as inboxed? It is delivered mail, categorized — not spam, not Primary. Our formula scores it as its own state at partial credit, and the Promotions tab entry covers why collapsing it either way misleads.

Where are InboxRatio's placement test results? Cycles are announced as they complete, and scores appear in the rankings with the measurement month attached. Until then, listings show "Not yet tested" — the methodology publishes before the numbers do.

Before your next major campaign, add a cross-provider seed row and schedule thirty minutes the same day to read it. One habit, one spreadsheet column per provider — and the gap between "delivered" and "seen" stops being a matter of faith. The protocol details worth copying are on how we test.

Sources