First, confirm it is placement, not delivery
"Our emails are going to spam" arrives as three different complaints, and only one of them is about spam. If the platform shows bounces, the mail was refused, which is a delivery problem with its own reasons in the bounce messages. If opens fell but clicks held, the measurement may have changed rather than the placement, as the email metrics and testing guide explains. Placement in the spam folder looks different: delivery stays high, clicks fall, replies stop, and customers say they found the message in junk.
Read placement from the provider's side. Split engagement by mailbox provider; a fall at one provider while the others hold steady is a placement problem at that provider, not a content problem. Open that provider's postmaster view, where Google's sender guidelines point senders for spam rate, reputation and authentication results. Then date the change against the send log: a new list source, a platform migration, a seasonal whole-list send or a DNS edit in the same week is usually the answer. Use the table in order; the top rows are the quickest to rule out.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Spam placement across providers, authentication failures in headers or postmaster data | SPF, DKIM and DMARC results for a real message; whether the passing domain matches the From domain | Domain not authenticated, or From domain not aligned | Publish and correct the records; sign with your own domain; align the From |
| Sudden move to spam at one provider soon after a send | Complaint rate by list source and send; the provider's spam rate | Complaint spike from an import or a sign-up source without clear consent | Pause the source, suppress complainers, re-permission or remove the source |
| Gradual slide into spam over months | Share of the list that has clicked recently; whole-list sends in the calendar | Inactive addresses diluting engagement | Introduce engagement tiers and stop mailing the inactive tier |
| Spam placement starts with a seasonal push, a migration or a new subdomain | Daily volume against the previous months; age of the sending domain | Volume spike or unwarmed domain | Drop to engaged segments and step volume up gradually |
| Complaints high although consent is sound | Message headers for one-click unsubscribe; visibility of the body link | Unsubscribe missing or hard to use | Add the one-click headers and a plain, visible unsubscribe link |
| Receipts and password resets also land in spam | Whether marketing and transactional mail share a domain, subdomain or system | One sending identity for both streams | Separate the streams by subdomain and, where possible, system |
Fix, verify and prevent each cause
Identity the provider cannot verify
Providers verify a sender through SPF, which lists the servers allowed to send for a domain, DKIM, which signs each message, and DMARC, which asks that one of them passes for a domain aligned with the visible From address. A platform that signs with its own domain, a From address on a domain with no records, or a new tool added without updating SPF all leave the message unverifiable.
Fix: read the authentication results in the headers of a real message, then correct the records and turn on signing with your own domain as the deliverability guide describes; the alignment rule is set out by DMARC.org.
Verify: a fresh test message shows SPF or DKIM passing for the From domain and DMARC passing, and the postmaster view shows authentication recovering over the following days.
Prevent: keep an inventory of every system that sends as the domain, and make "added to SPF and signing with DKIM" part of switching on any new one.
A complaint spike from one source
Recipients marking mail as spam is the heaviest signal a provider reads. A spike almost always traces to people who did not expect the message: an imported spreadsheet from an event or a sales team, a co-registration offer, or a form whose wording did not say what would follow.
Fix: find the source by comparing complaint rates per list source and per send, stop mailing that source, suppress everyone who complained, and either run a single re-permission message or remove the source, as the list building guide recommends for imports.
Verify: the complaint rate on the next sends returns to its usual level and the provider's spam rate stops rising.
Prevent: stage every import in a holding segment and mail it in small batches before it joins the main list.
Inactive addresses dragging engagement down
A list that is mailed in full every time accumulates people who stopped reading long ago. Each ignored message weakens the engagement the provider sees, and long-abandoned addresses can be turned into traps. The slide is slow, so it is often misread as content fatigue.
Fix: define engagement tiers by recent clicks, send a short re-permission sequence to the inactive tier and suppress those who do not respond, following the segmentation guide.
Verify: engagement by provider improves over the next few sends while the delivered volume falls, and placement at the worst-affected provider recovers.
Prevent: refresh the tiers automatically and make "who receives this" a decision for every send, not a default of everyone.
A volume jump without warming
Providers trust consistency. A domain that has sent steadily to engaged segments and suddenly mails the whole list, a new subdomain that starts at full volume, or a migration to a new platform in the same week as a big campaign all look like a change of sender.
Fix: fall back to the most engaged segment and step the volume up in stages, holding whenever complaints or bounces rise; split very large sends across days. Google's sender guidelines advise against sudden spikes, and M3AAWG's sender practices cover the operator side.
Verify: each step holds its complaint and bounce rates before the next one, and reputation in the postmaster view stays level.
Prevent: put big sends and migrations in the calendar early, and never combine a platform move, a list import and a new cadence in one week.
An unsubscribe that is missing or hard to use
When leaving is difficult, the spam button becomes the unsubscribe. For bulk senders Google documents one-click unsubscribe through the headers that RFC 8058 specifies, alongside a visible link in the body, and prompt processing of requests.
Fix: confirm the platform adds both headers to marketing mail, put a plain unsubscribe link where readers look for it, and check that requests are honoured within the documented window.
Verify: the headers appear on a received test message, the mail client offers its own unsubscribe control, and complaints fall while unsubscribes rise slightly, which is the trade you want.
Prevent: include the unsubscribe headers and link in the pre-send checklist, and test the whole exit path after any template or platform change.
Two streams sharing one identity
If receipts and campaigns leave from the same domain and system, a complaint spike on a sale day drags order confirmations into spam with it, and a provider sees one sender behaving erratically.
Fix: move marketing and transactional mail onto separate subdomains with their own authentication, and onto separate systems where possible, as the promotional versus transactional comparison explains.
Verify: postmaster data shows the two subdomains separately, and transactional placement holds steady through the next promotional send.
Prevent: decide the stream for every new message type before it is built.
A prevention routine before every large send
Most spam problems are created on an ordinary day and noticed a week later. A short routine catches them while they are still small. Before a large send, check that the authentication results on a test message pass and align, that the audience excludes the inactive tier and any unstaged import, that the volume is in line with recent weeks, and that the unsubscribe headers and link are present. After the send, read complaints, bounces and clicks by provider and by list source the next morning, not only the campaign total. Weekly, open the postmaster view for spam rate and reputation and skim the DMARC aggregate reports for unfamiliar sources. Monthly, put the handful of measures that matter, complaint rate, active share of the list and clicks by provider, into the report as deliverability indicators (KPIA metric tied directly to an objective.), so the trend is visible before placement fails. When an alert does fire, start with the send log for the days before it; the cause is nearly always the newest change. When the newest change is a new domain or platform, the email sending domain setup checklist covers the authentication, warm-up and postmaster registration it needs before the first send.
A garden retailer's spring newsletter lands in junk
A garden retailer sends a weekly newsletter to about 40,000 subscribers. In March, clicks from Gmail addresses fall by roughly half over two sends while other providers hold steady, and two customers mention finding the newsletter in spam. Postmaster Tools shows the spam rate rising from well under 0.1% to 0.4% across the same fortnight. The send log shows 6,500 addresses imported from a garden-show competition two weeks earlier, mailed with the main list from the first send. Complaints from that source run at about ten times the rest. The team suppresses the complainers, moves the import into a holding segment, sends it one re-permission message that 900 people confirm, and drops the rest. Within three weeks the spam rate is back below 0.1% and Gmail clicks recover to their February level. The import procedure now stages every event list before it joins the newsletter.
Common mistakes
Reading the delivered rate as inbox placement. Rewriting subject lines when the cause is a list source. Moving to a new domain instead of fixing the behaviour that damaged the old one. Mailing an import with the main list on its first send. Hiding the unsubscribe link to protect list size, which turns exits into complaints. Treating the promotions tab as spam. Checking only the campaign total when the problem sits at one provider.
