Why Is Smartlead Deliverability Dropping? What to Check

When placement falls, the instinct is to rewrite the email. That is almost always the wrong first move. Copy affects placement at the margin; infrastructure and reputation decide it. Work through this in order, because each step rules out a category of cause that would otherwise waste a week of copy testing.

1. Authentication, first and always

Verify the boring things before anything else, because they break silently and they break hard.

SPF must be a single record. A domain is allowed exactly one SPF TXT record. If someone added a second when connecting a new tool, that is a permanent error and receivers treat it as a failure rather than combining them. This is the most common self-inflicted authentication fault, and it typically appears right after a stack change. Merge every include: into one record.

Watch the lookup limit too. SPF permits ten DNS lookups; each include consumes at least one, and a stack that has accumulated several senders can quietly exceed it, which also fails.

DKIM should sign with a 2048-bit key, and the selector must actually resolve in DNS. A rotated or partially-published key produces intermittent failures that are maddening to trace because some mail passes.

DMARC needs alignment, not just presence. A published DMARC record does nothing if neither SPF nor DKIM aligns with the visible From domain. Check alignment specifically rather than assuming a record equals a pass. The full mechanics are in SPF, DKIM and DMARC explained.

2. Whose reputation are you sending on?

If authentication is clean, look at the infrastructure.

Smartlead, like most sequencers, routes through shared sending infrastructure by default. Shared IP pools blend the reputation of everyone using them. When another tenant mails a stale list and generates bounces and complaints, the penalty attaches to the pool, and it reaches your mail even though your own list is clean and your records are correct.

The tell for this is a sudden drop with no corresponding change on your side. Check whether your sending IP, not just your domain, appears on any blocklist, and check them separately because they can be listed independently. Blacklists and how to get removed covers the process.

Smartlead's own answer to this is SmartServers, dedicated IP isolation at roughly $39 per server per month. That genuinely fixes inherited reputation, with the caveat that a fresh dedicated IP has no history at all, so it needs proper warm-up before carrying real volume.

3. Velocity and volume

Compare your per-mailbox daily volume against two weeks ago. Providers enforce hourly velocity thresholds independently of daily limits, and crossing them produces temporary blocks that look like a deliverability problem because in effect they are one.

If mail is logging as delivered while opens collapse toward zero, treat that as quarantine rather than a content issue. Stop sending from the affected mailboxes and let them rest. Pushing more volume through a throttled account converts a temporary problem into a durable one.

Keep per-mailbox daily volume conservative and scale by adding mailboxes rather than by increasing per-mailbox output.

4. Tracking clutter

Open tracking inserts a remote image into every message. Link tracking rewrites every URL to route through a redirect host. Both are normal in marketing mail and both are unusual in the kind of plain, personal message cold outreach is trying to imitate.

Two specific problems. If your tracking hostname is generic rather than a subdomain of your sending domain, the mismatch between the visible From domain and the link destination is exactly the pattern phishing detection looks for. And a message that is mostly HTML scaffolding with several rewritten links and a tracking pixel has a poor text-to-HTML ratio, which pushes it toward Promotions even when it is not classified as spam.

Try one campaign with open tracking disabled and compare. Losing the open-rate metric is a real cost, but open rates have been unreliable since Apple Mail Privacy Protection anyway, and reply rate is the number that matters. Why emails land in Promotions goes deeper on this specific failure mode.

5. Content fingerprinting

This is where spintax comes in, and where a lot of established practice has quietly stopped working.

Spintax rotates words. Modern filters do not compare messages by exact string; they compare them semantically, which means thousands of messages that differ only in synonym choice still resolve to nearly the same underlying vector. The campaign clusters as one campaign regardless of how many variants you generated.

Defeating that requires genuine variation in structure and substance, not vocabulary. Different opening angles, different sentence structures, different specifics per recipient. That is difficult to do by hand at volume, which is precisely why per-recipient generation rather than token substitution has become the practical approach. RepMail uses GPT-4o to rewrite each message rather than rotating spintax tokens, which changes the actual content rather than the surface string.

Also check your spam trigger words and current spam score, but do this last. Content is a real factor and rarely the cause of a sudden drop.

6. The numbers that decide

Finally, look at your bounce and complaint rates directly. Complaint rate above roughly 0.3% is the threshold at which Google's postmaster signals turn hostile, and it is very difficult to recover from without pausing entirely. Rising hard bounces mean list decay, and list decay compounds: every stale address you keep mailing costs you reputation you then spend months rebuilding.

Real-time suppression matters more than most teams assume here. If bounce data arrives through a delayed log sync, you keep sending to addresses you already know are dead. RepMail's AWS SNS integration suppresses on the bounce event itself, which is the difference between removing a bad address immediately and removing it tomorrow.

When to repair and when to rebuild

If the cause is a fixable configuration error, a velocity spike, or a temporary block, repair it: correct the records, pause, and warm back up over a few weeks. Domains recover from mechanical faults.

If the domain has accumulated months of high complaint rates or sits on multiple authoritative blocklists, repair is usually the more expensive path. Isolate it, move active campaigns to clean outbound domains configured correctly from the start, and treat the old domain as a sunk cost rather than a project.