Methodology
Inbox placement testing methodology: how the seed panel works and where it stops
Deliverability tooling has an accuracy-theater problem: dashboards that imply certainty the underlying method cannot deliver. This page explains exactly how our testing works, what it measures well, and what no seed test can tell you.
The seed panel
A placement test sends your real email, from your real ESP, to a panel of live mailboxes we operate at the providers your audience actually uses. The standard panel is around 30 seeds; extended panels (Growth and up) run around 80, adding B2B-hosted mailboxes and deeper Gmail Promotions detection. Approximate distribution:
| Provider group | Standard panel | Extended panel | What we read |
|---|---|---|---|
| Gmail (consumer + Workspace) | 8 seeds | 22 seeds | Inbox vs Promotions vs spam, per-seed headers |
| Outlook / Hotmail / Microsoft 365 | 8 seeds | 22 seeds | Inbox vs junk, SmartScreen verdict headers |
| Yahoo (incl. AOL infrastructure) | 5 seeds | 12 seeds | Inbox vs bulk folder |
| iCloud Mail | 3 seeds | 8 seeds | Inbox vs junk |
| GMX / European providers | 3 seeds | 8 seeds | Inbox vs spam |
| Zoho + B2B hosts | 3 seeds | 8 seeds | Inbox vs quarantine behavior |
Seed counts are approximate and rotate; panels are maintained continuously as providers change behavior.
What every test measures
- Placement per seed: inbox, Promotions (Gmail), or spam, read directly from the mailbox, not inferred from opens.
- Authentication as received: SPF, DKIM, DMARC and alignment parsed from the actual
Authentication-Resultsheaders each provider wrote, the same verdict its filter used. Details on the SPF, DKIM, and DMARC checker pages. - Infrastructure: the sending IP against 130+ public blacklists, reverse DNS, TLS on the delivery path.
- Content and structure flags: the template signals correlated with filtering and Promotions routing (image-to-text ratio, link density, list-unsubscribe headers), reported as flags with reasons, not a mystery score.
Every finding lands in one fix list ranked by impact (HIGH / MED / LOW), phrased as the exact record or setting to change. When nothing is wrong, the report says "nothing to fix" and recommends monitoring, not an upsell.
Accuracy: what a seed test is, and is not
Seed results are directional estimates, not replicas of your real recipients. Gmail and Microsoft personalize filtering per recipient using each user's history; a seed mailbox has no relationship with you, so it approximates a cold or typical recipient. What a seed test detects reliably: systematic failures (broken authentication, blacklist hits, wholesale spam routing at a provider, template flags). What it cannot detect: how your most engaged subscriber's inbox will treat you, or per-user variance. This is true of every seed-based tool on the market, including ours; we are just the ones saying it on a page you can link to.
Corroborate trends with provider-native data where it exists: we ingest Google Postmaster Tools (Growth and up) and parse your DMARC aggregate reports, which cover your real traffic, not the panel. Where seed data and provider data disagree, provider data wins and the report says so.
The no-warmup policy
Inboxes operates no warmup network. No bot mailboxes opening your mail, no automated replies, no dragging messages out of spam folders to simulate engagement. Those tactics violate the provider policies that now govern bulk sending (Google and Yahoo requirements from February 2024, Microsoft enforcement from May 2025, Gmail's hard SMTP rejections from November 2025), and providers actively detect and punish them. Our seed mailboxes are passive measurement instruments: they receive and report, and they never generate engagement signals. Long version: the honest truth about email warmup.
Standards we align with
Recommendations follow the published rules, not folklore: RFC 7208 (SPF), RFC 6376 (DKIM), RFC 7489 (DMARC), RFC 8058 (one-click unsubscribe), the Google and Yahoo bulk-sender requirements, and M3AAWG sender best practices. Where a recommendation is judgment rather than standard (engagement segmentation, volume ramps), the fix says so and explains the reasoning.
Who this is for
Legitimate senders: newsletters people signed up for, transactional mail, marketing to your own opted-in list. We do not help unsolicited bulk email get delivered, and signups that are about that get declined. Some problems tooling cannot fix, a bought list, content people never wanted, and when that is the diagnosis, the report says it plainly.