You cannot see where your cold emails land. Opens are unreliable, replies are slow, and by the time a campaign's numbers look wrong you have already burned through a chunk of your list. A seed test is the closest thing you get to looking over the recipient's shoulder before you commit real prospects.
This guide covers how to set one up so the result means something, how to read it, and, just as important, what a seed test cannot tell you.
What a seed test actually is
A seed test, also called an inbox placement test, sends your email to a list of test mailboxes spread across the major providers. Those mailboxes are monitored, so after delivery the tool checks each one and reports where the message ended up: the primary inbox, a secondary tab such as Gmail's Promotions, the spam folder, or nowhere at all.
The result is a snapshot of how filters at each provider treat this specific message, from this specific mailbox, at this moment. That is useful, as long as you remember all three qualifiers. You can run one with the inbox placement test tool.
Setting up the test so it means something
The most common mistake is testing something other than what you actually send. A test from a different mailbox, with simplified copy and tracking turned off, tells you about an email you are never going to send. Mirror production exactly:
- Send from the real mailbox. Reputation is attached to the sending domain, the IP, and the mailbox. Test from the account that will run the campaign.
- Use the real copy. Same subject, same body, same signature, same links. If you use spintax or personalization variables, fill them in with realistic values rather than leaving raw tokens.
- Keep the real tracking settings. If production uses open tracking or a click-tracking domain, the test should too. Tracking pixels and rewritten links are part of what filters evaluate.
- Send through the same path. Same platform, same SMTP route, same headers. A test sent from your mail client instead of your sending tool can look quite different.
- Send at a normal time. Run it within your usual sending window, not right after a burst of campaign traffic from the same mailbox.
Choosing a provider mix
Your seed list should reflect where your prospects actually read email. For B2B outreach that usually means business mailboxes more than free ones.
- Google Workspace and Gmail. Separate them if you can. Workspace filtering is influenced by each organization's admin settings and can behave differently from consumer Gmail.
- Microsoft 365 and Outlook.com. Same idea. Corporate Microsoft 365 tenants often add their own filtering on top of Microsoft's defaults.
- Yahoo and AOL. Less common in B2B, but worth including if you target small businesses or consumers.
- Others. If a meaningful slice of your list sits behind specific security gateways or regional providers, include seeds there if your tool supports it.
If most of your list is on Microsoft 365, a great Gmail result is not much comfort. Weight your attention by where your audience lives. The breakdown in how Gmail and Outlook filter cold email explains why the two often disagree.
Reading the results
Each seed lands in one of a few buckets. Here is how to interpret each one for cold email specifically.
| Result | What it means | What to do |
|---|---|---|
| Primary inbox | The filter trusted this message enough to show it normally. | Good. Keep volume steady and retest periodically. |
| Promotions or Other tab | Delivered, but classified as bulk or marketing. | Strip HTML, images, and extra links. Make it look more like one-to-one mail. |
| Spam or Junk | The filter distrusts the sender, the content, or both. | Check authentication, blacklists, and mailbox health before touching copy. |
| Missing | Not found in any folder. Possibly rejected, silently dropped, or delayed. | Check bounces and logs, wait and recheck, then investigate reputation. |
Look at patterns, not single seeds. One Gmail seed in spam out of several is noise. Every Microsoft seed in junk is a signal. If you want to know why a specific message was filtered, open it in the seed mailbox, copy the full headers, and run them through the email header analyzer. SPF, DKIM, and DMARC results, plus any spam scoring headers the provider adds, usually point at the cause.
Why results vary from run to run
Run the same test twice and you may get different answers. That is normal. Filters are not static rule lists; they combine many signals that shift constantly, including recent sending volume from your IP and domain, how other recipients have reacted to your mail in the last few days, and the provider's own model updates. Seed mailboxes themselves also differ in age and history.
So treat a single run as a reading, not a verdict. Run two or three tests before drawing a conclusion, and compare trends over weeks rather than reacting to one bad afternoon.
The limits of seed testing
Seed mailboxes have no relationship with you. They have never replied to you, never moved your email out of spam, never added you to contacts. Real recipients do those things, and engagement is one of the strongest signals providers use. That cuts both ways: an engaged real audience may see better placement than your seeds, while a cold, disengaged list may see worse.
Seeds also cannot reproduce every corporate filter. A prospect's company may run a security gateway with its own rules that no seed list covers. Seed tests are excellent at catching broken setups and obviously problematic content. They are weaker at predicting exact placement for a specific list.
What to change when results are bad
Work from the foundation up. Changing copy first is tempting because it is easy, but it is rarely the root cause.
- Confirm SPF, DKIM, and DMARC pass and align, using the SPF and DKIM checker.
- Check the domain and sending IP against major blocklists. If listed, see the blacklist delisting guide.
- Review recent volume and bounce rate for that mailbox. A spike in either is a common cause.
- If the mailbox or domain is new, slow down and let warmup do its job longer.
- Only then simplify content: plain text, one link or none, no images, no tracking domain you do not control.
- Retest after each change so you know which one helped.
How often to retest
Test before any new campaign, after any change to DNS, copy templates, or tracking settings, and whenever reply rates drop without an obvious reason. For ongoing campaigns, a weekly test per active sending domain is a sensible baseline, more often if you are scaling volume. Sendbox includes unlimited placement tests for this reason: a test you have to ration is a test you skip. Pair the results with per-mailbox health data, and you will usually spot a problem days before it shows up in your reply rate.
