The most misleading number in email is 99% delivered.
It means the receiving server accepted the message. It does not mean a human saw it. A message filed straight into spam was delivered, successfully, and every report you read will agree that it went well.
If you are running cold outreach and judging it on delivery reports, you are looking at the wrong instrument. Here is how to find out where your mail actually goes.
What placement is, and why it is per-provider
Once a server accepts your message, it decides where to file it. Inbox, a tab, or spam. That decision is made independently by each provider using their own signals, which is why the single most common result of a first placement test surprises people:
Inbox at one provider, spam at another, on the same send.
That is not a fault in the test. It is the actual situation, and it is why a single test address tells you nearly nothing. Google and Microsoft weight different things. A regional provider may weight something else again. What you need is not a verdict, it is a pattern.
Running the test
1. Get seed mailboxes. One at each provider you actually send to. For most senders that is a Google account, a Microsoft account, and whichever regional provider shows up in your list. Free accounts are fine and are exactly what you want, because they have no relationship with you.
The important property: these mailboxes must have no history with your domain. Do not use a colleague's address you have emailed for two years. That address is the easiest possible case and it will pass while your real recipients do not.
2. Send the real message. Not a test message. Not "hi testing 123". The actual campaign email, with the actual subject, the actual signature, the actual link, the actual tracking configuration, sent from the actual mailbox through the actual platform.
This is where most tests fail before they start. A stripped-down probe tests a message you are never going to send, and the parts you stripped are frequently the parts causing the problem.
3. Record where each one landed. Inbox, promotions or other tab, or spam. Write it down. A screenshot costs nothing and is worth having in a month when you are trying to remember whether things got better.
4. Repeat at your real volume and cadence. Placement under a single test send is not placement under two hundred sends a day. If you can, seed a few addresses into a real campaign send rather than testing separately, which is the closest you will get to measuring the actual thing.
Reading the result
The pattern is the diagnosis. This table is most of the value of the whole exercise:
| What you see | Most likely cause | Where to look |
|---|---|---|
| Spam at every provider | Authentication or domain reputation | SPF, DKIM, DMARC, domain age |
| Spam at one provider only | Reputation with that provider | Volume ramp, bounces, complaints |
| Promotions tab, not spam | The message looks bulk | Images, tracking pixel, link count |
| Inbox on the test, spam in the campaign | Volume or list quality | Ramp rate, unverified addresses |
| Was fine, now is not | Something changed in your sending | Last 72 hours of volume and bounces |
Take them in order.
Failing everywhere is nearly always authentication or a domain with no reputation at all. Both are fixable in an afternoon and a fortnight respectively: the setup is in SPF, DKIM and DMARC: the 20-minute setup, and the fortnight is warm-up.
Failing at one provider is a reputation problem with that provider specifically, and reputation is built from your recent behaviour: how fast you ramped, how many addresses bounced, how many people marked you. A single campaign into a stale segment will do this and the effect outlasts the campaign by weeks.
Landing in promotions means your one-to-one email looks like a newsletter. The usual suspects are an open-tracking pixel, an image signature, more than one link, and any phrase that appears in a million marketing emails. Strip all four and retest; this one usually fixes in a single iteration.
Fine in test, poor in campaign is a volume or list problem rather than a message problem. Either you are ramping faster than your mailbox has earned, or you are sending to addresses you did not verify, and the bounces are doing the damage. The email bounce rate guide has the thresholds.
The discipline that makes it useful
Change one thing, then retest. This is boring advice and it is the difference between diagnosing and flailing.
The temptation after a bad result is to fix everything at once: add DMARC, strip the images, slow the ramp, rewrite the subject. If placement improves you have learned nothing, and you are now carrying four changes of which three might be unnecessary and one might be making a different thing worse.
One change. Retest. Record. Then the next.
Allow time between iterations, too. Reputation signals move over days, not minutes, so a retest twenty minutes after a change is measuring the old state.
What automated placement testing does for you
Doing this by hand is entirely possible and takes about twenty minutes per round. The reason platforms automate it is not the twenty minutes, it is the frequency: placement is not a state you check once, it drifts, and the drift is invisible until a campaign underperforms for reasons nobody can name.
Our warm-up system runs placement probes continuously as part of warming, which means two things happen without anybody deciding to look. Mail that lands in a spam folder gets rescued and marked as wanted, which is the signal that actually teaches a provider to trust you. And the health score per mailbox moves before your campaign does, so a mailbox degrading shows up as a number rather than as a quiet week.
That is the honest case for automation here: not that the manual test is hard, but that nobody remembers to run it on the Tuesday when it would have mattered.
A short checklist
- Seed mailboxes at every provider you send to, with no history with you.
- Send the real message, from the real mailbox, with the real settings.
- Record inbox, tab or spam per provider, with a screenshot.
- Read the pattern, not the individual result.
- Change one thing. Wait. Retest.
- Repeat monthly even when nothing appears wrong.
Delivered was never the number. This is.
Related: why your emails go to spam for the fourteen underlying causes, and the email deliverability checklist for the fixes in the order they matter.