The inbox placement test
we actually run
Your open rate does not tell you whether you reach the inbox.
This test does. For us, it reversed the obvious answer.
From our own operation: 50 sending mailboxes, one week of measurement, and a result that talked us out of what we were about to do.
Open rate is not proof of delivery
Almost every outbound team measures deliverability through open rate. It was always a proxy, and since mail clients started pre-loading images it is broken: opens happen without a human ever seeing the message, and a missing open does not mean the message failed to arrive.
More importantly, there is something open rate cannot answer in principle: where did the message land? Inbox, promotions, spam, or never accepted at all? A message in spam does not get opened. Neither does one that never arrived. Both look identical in your stats and need completely different responses.
What you want to know
Does a real campaign email from this sender land in the inbox at the recipient's provider?
What open rate tells you
That a pixel loaded somewhere at some point. Or did not.
What a seed test is
An inbox placement test answers that question by asking it directly. You create your own mailboxes at the providers your recipients actually use. Then you send the same email through every one of your sending mailboxes to those seed addresses, and you look at where it landed.
No model, no estimate, no vendor reputation score. You see the folder.
A real campaign email
Not "test 1, 2, 3". Content is part of the verdict. Use the email you actually send.
Every sender separately
Reputation sits with the mailbox, not the tool. An average across senders hides exactly the mailbox that is burning.
Split by recipient provider
This is the step most people skip, and it is the one that made the difference for us.
What we measured
Measured in our own operation, seed results from 22 to 29 September 2026, 50 sending mailboxes:
land in the inbox between 0 and 14 % of the time at Microsoft recipients.
reach considerably more of the same recipients.
One week, one workspace, seed mailboxes. Not an industry benchmark: your own numbers will look different. The point is the pattern, not the level.
The mistake we made
Our first instinct was to pull the Microsoft mailboxes out of rotation. Microsoft has a reputation for being difficult, the numbers there looked noisy, so out they go.
Measuring reversed that. The Microsoft mailboxes are the good ones. The problem is not a sender, it is a route: Google to Microsoft. Had we followed the instinct, we would have switched off precisely the mailboxes that work for that group of recipients, and made the problem worse.
Do not rank senders. Rank routes. The output of a seed test is a matrix, not a leaderboard.
How to run it
Build the seed list
At least two mailboxes per recipient provider that actually appears in your target group. In B2B that is usually Microsoft 365 and Google Workspace, plus one or two self-hosted providers depending on the industry. Look at your lead list instead of guessing: the email domains tell you who you need.
Use a real email
Same subject line, same body, same links as the live campaign. A friendly test message with no link passes every filter and proves nothing.
Send once from every mailbox
Individually, not as one blast from a single mailbox. You are measuring mailboxes, not the tool.
Check after 30 to 60 minutes
Four possible states per seed mailbox: inbox, promotions, spam, never arrived. Checking immediately is too early, checking the next day is too late: filters reclassify.
Write it down as a matrix
Rows are your senders, columns are recipient providers, each cell holds the inbox share. Read the cells, not the row averages. A mailbox averaging 50 % can sit at zero with one provider, and that provider may be half your target group.
What to do with the result
The limits, honestly
Seed mailboxes are not real recipients: they have no history with you, they never reply, and they never rescue anything from spam. A seed test therefore understates deliverability to people who know you, and it is a snapshot. It answers one question very well, where a message from this sender sits at this provider right now, and nothing beyond it.
Measure once before you switch anything off
Build the seed list from your own lead domains and run the test once. The matrix tells you more in an hour than three months of open rates.
Back to the topDeliverability as a standing watch
We set the seed test up as a recurring measurement, with a route matrix and an alert when a cell turns, before a campaign runs through it.
Book an intro call