Inbox placement testing: what it measures and how to run one

A placement test sends a real message to real mailboxes you control at each major provider, then records where it landed: inbox, a tab, Junk, or nothing at all. That is the only measurement that answers the question you actually have. Your ESP's delivery rate cannot answer it, because a message accepted with 250 OK and filed in Junk is a successful delivery by that definition.

The five things a complete test covers

1. Per-provider placement. Gmail, Outlook and Microsoft 365, Yahoo, and whatever your recipients really use — iCloud, Zoho, Proton, regional hosts. An aggregate "92% inbox" number hides the case that matters, which is almost always one provider at zero.
2. Authentication. SPF alignment and lookup count, a valid DKIM signature that survives transit, and a DMARC policy that at least one of the two aligns with.
3. Reputation signals you can still see. Gmail spam rate and authentication panels in Postmaster Tools, Microsoft SNDS for IPs you own, your own bounce and complaint rates.
4. Blocklist state for both the sending IP and the sending domain, plus the domains of any links in the body — a listed tracking or redirect domain fails the message even when your own domain is clean.
5. Content and structure. HTML-to-text ratio, link count, link-domain mismatch, image-only bodies, attachment types.

How to run one so the result means something

Change one variable at a time. A test comparing a new template from a new domain through a new relay tells you nothing about any of the three.

1. Send the exact production message — same template, same links, same
   footer, same From display name.
2. Send from the mailbox that will actually send it, not a test account.
3. Record provider, folder, and the Message-ID for every seed.
4. Keep the send small and spread out; a burst to seeds at one provider
   is itself a filtering trigger.
5. Repeat on a schedule. One test is a snapshot; weekly tests are a signal.

Then read the headers of a message that landed badly. The useful lines are Authentication-Results (did SPF, DKIM and DMARC pass, and did they align?), Received chains for added hops, and any X- header the provider leaves behind explaining its verdict. Microsoft is relatively generous here; Gmail is not.

Reading the result

Two honest caveats about Gmail. First, the Promotions tab carries the INBOX label, so any tool reading labels through the API scores a Promotions landing as a clean inbox hit. If your numbers come from label inspection, treat them as an optimistic ceiling and check the tab by eye on at least one seed. Second, the domain and IP reputation charts in Postmaster Tools are retired, so you cannot close out a test with a reputation grade any more. Spam rate plus per-provider placement is the replacement, and it is a better measurement anyway because it is about your mail rather than a label.

What not to do

Do not trust a single composite "deliverability score". It bundles DNS checks, content heuristics and placement into one number, and the number moves for reasons that have nothing to do with where your mail lands. Do not use a seed list as a warm-up mechanism; seeds exist to be measured, and sending volume at them teaches the filters about mail nobody wanted. Do not test with a stripped-down "clean" version of your email and then ship the real one. Do not conclude anything from one provider's one message — filtering is probabilistic and per-recipient, and a sample of one is noise.

And do not read placement as a verdict on your copy when your bounce rate is above a couple of per cent. At that point you are measuring list quality with a very expensive instrument.

Testing is not a one-off

Placement moves when you change template, add a link domain, raise volume, add mailboxes, or when a provider adjusts filtering. The value is in the trend: a Friday where Outlook drops from 90% to 40% tells you exactly which change to look at, because you know what you shipped on Thursday. A one-off test before launch tells you only that launch day was fine.

MailPilot runs this continuously rather than on demand. It measures real per-provider inbox placement through its own seed-inbox network, checks SPF, DKIM and DMARC for every connected domain, and records per-mailbox reputation daily, so a placement drop arrives as a change in the chart instead of a surprise in next month's reply rate.