There is no single system that knows whether your email reached the inbox. Everything sold as a "deliverability dashboard" is a view onto one of five independent data sources, and each source is blind to things the others see clearly. If you know which source owns which fact, most deliverability triage becomes a twenty-minute job instead of a week of theories.
The five sources
| Source | What it actually measures | Limits |
|---|---|---|
| Google Postmaster Tools | Spam rate as marked by Gmail users, SPF/DKIM/DMARC pass rates, delivery errors, encryption — per sending domain | Gmail only. Needs meaningful volume before charts populate. The domain reputation chart was retired, so there is no longer a Bad/Low/Medium/High label to read |
| Microsoft SNDS | Complaint and trap data for IPs you own, plus filter verdicts | IP-level only, so it is useless on a shared ESP pool or on Microsoft 365 mailbox sending |
| Your ESP's logs and metrics | Accepted, deferred, bounced, complained, unsubscribed — the receiving server's SMTP answer | Stops at the SMTP boundary. A 250 OK says accepted, not inboxed |
| Seed inbox placement tests | Which folder a specific message landed in, per provider | Small sample, seed mailboxes have no human engagement history |
| Blocklists and DNS | Whether an IP or domain is listed; whether SPF, DKIM and DMARC are published and aligned | Binary and current-state only. No history, no placement |
Everything else — spam-word scorers, subject-line graders, "email health" scores — is a derived opinion built on top of these, not a sixth source.
The one confusion that causes most wrong conclusions
Your ESP's delivery rate is an acceptance rate. When Gmail answers 250 2.0.0 OK it has taken custody of the message; the folder decision happens afterwards, inside Gmail, and is never reported back to you. A campaign can show 99.4% "delivered" in SendGrid or Amazon SES and sit entirely in Spam.
The inverse trap: a seed placement test showing 95% inbox does not mean your real list sees 95% inbox. Seed mailboxes never reply, never star, never move you out of Spam. They measure filter verdicts on content, authentication and reputation — not your relationship with real recipients.
Where each number lives in each product
- Amazon SES — reputation metrics (bounce and complaint rate) live under *Account dashboard* in the SES console; per-message events need a configuration set with an event destination (SNS, Firehose or CloudWatch) attached before you send, or the data is simply never recorded.
- SendGrid — *Stats* for aggregates and the *Email Activity* feed for individual messages. Activity retention is limited by plan, so export before you need it.
- Mailgun — *Sending → Logs* carries the actual SMTP response string the receiver returned, which is the single most useful field in any ESP UI.
- Postmark — per-stream statistics; transactional and broadcast streams are scored separately, which is the point of streams.
- Google Workspace / Microsoft 365 mailboxes — nothing. Mailbox sending produces no provider-side analytics, which is why outbound teams on Gmail or 365 need seed testing to see anything at all.
Triage order when placement drops
- DNS and alignment first. Check that SPF still resolves inside the 10-lookup limit, that the DKIM selector your sender actually signs with is published, and that the visible From domain aligns with the SPF or DKIM domain. A DNS edit by someone else is the most common cause of a sudden drop.
- Read the bounce strings, not the bounce rate. One
550 5.7.1policy rejection from Google and one451 4.7.1rate-limit deferral from a small host mean opposite things. The rate hides that. - Then Postmaster Tools. If the Gmail user-reported spam rate is near or above 0.3%, nothing else you change will help until complaints fall. That threshold is Google's own published number for bulk senders.
- Then blocklists. Check the domain and the sending IP. Note which lists you are checking against — SORBS was shut down in June 2024 and some lookup pages still show stale results, while Validity launched the Heatwave blocklist on 3 September 2026, so a listing that did not exist last year may now.
- Then a seed test, to confirm whether the fix changed the folder.
What not to do
- Do not buy a tool for data a provider gives you free. Postmaster Tools, SNDS and your own ESP logs cost nothing and are closer to the source than any resold version.
- Do not compare "deliverability %" between two products. They are measuring different events, so the numbers are not commensurable.
- Do not act on a single seed test. Run one before a change and one after, with the same message and the same seed set, or you have measured noise.
- Do not treat a spam-word score as a reputation signal. Content scoring is the least predictive of the five sources and the most heavily marketed.
MailPilot sits across three of these sources at once: it checks SPF, DKIM and DMARC for each sending domain, measures real per-provider inbox placement through its seed network, and tracks per-mailbox reputation daily so a drop shows up as a trend line rather than as a surprise in next month's reply rate. It connects over OAuth for Gmail and Outlook, or over SMTP and IMAP for SendGrid, Amazon SES, Mailgun, Postmark and custom SMTP hosts.
Ready to reach the inbox every time?
Automated email warmup across real peer mailboxes. Live inbox placement monitoring. Free 7-day trial - no credit card required.
Start free trial