Resources / Field guide
Your receipts are paying your newsletter's bills.
Mailbox providers score senders, not message types. If your password resets and your promotions leave from the same identity, they share one reputation, and the noisiest stream sets it. Here is how to know when that arrangement has started costing you, and what a correct split looks like.
The mechanism
One identity, one fate.
Receivers build reputation around what they can observe: the sending IP, the authenticated domain, the behavior over time. They do not grade on intent. A domain that sends 95 polite receipts and 5 complaint-generating promotions is, to Gmail, one sender with a complaint problem. The placement penalty lands on everything that identity sends, which means your most aggressive marketing day quietly sets the deliverability your order confirmations live with.
This is not a flaw in the receivers' logic. It is the only logic available to them, and it has a clean implication for senders: if two traffic classes deserve different fates, they need different identities. Separate subdomains, separate DKIM signatures, separate IPs where volume justifies it, separately warmed and monitored. Receivers will then literally see two senders, and score each on its own behavior.
The symptoms
Signs the shared identity has started costing you.
Password resets or receipts intermittently land in spam, and the incidents cluster in the days after large campaign sends.
Transactional latency varies with the marketing calendar: OTPs are instant on quiet weeks and sluggish during promotions.
A complaint spike on one campaign measurably moved placement metrics for unrelated operational mail.
Support tickets about missing critical email rise at month-end, quarter-end, or whenever marketing volume peaks.
You cannot answer "what is the complaint rate of our transactional mail alone?" because the data only exists blended.
Any one of these is suggestive. Two or more means the shared identity is already taxing you, and the only question left is how much of the tax shows up in revenue versus support load.
Doing it right
What a correct split involves.
The naive version (point marketing at a second tool and call it done) usually recreates the problem with extra steps. A correct split has structure. Identity separation: distinct subdomains per stream, each with its own aligned SPF, DKIM, and DMARC posture, so receivers can actually tell the streams apart. Reputation separation: separate sending capacity where volume justifies it, each warmed on its own ramp rather than borrowing trust the other stream earned. Policy separation: the transactional stream gets priority pacing and the strictest hygiene; the marketing stream gets receiver-aware shaping and one-click unsubscribe handled flawlessly. And monitoring separation: per-stream complaint, bounce, and deferral metrics, because the entire point is being able to see which stream is doing what.
Worth deciding deliberately: where lifecycle mail (onboarding, digests, re-engagement) belongs. It behaves like marketing to receivers more often than product teams want to believe. When in doubt, give it its own identity and let its own numbers settle the argument.
One warning for regulated and commerce senders: a split also cleans your evidence. When transactional mail has its own stream, "prove the notice was delivered" stops requiring you to explain why the delivery record sits in a marketing tool's dashboard.
What you get out of it
What you get if we run this.
The split as configuration, not a quarter of work
Identities, DNS, DKIM selectors, warmup ramps, and per-stream monitoring all provision through one platform. The multiplication that makes teams defer this project is exactly the overhead we absorb.
Streams that stay separated
Suppression, pacing, and reputation isolation are enforced at the delivery gate per identity. A misrouted campaign cannot quietly ride the transactional stream’s reputation.
Per-stream truth on a panel
Complaint, bounce, and deferral metrics per stream, with alerts against each stream’s own baseline. The "which mail is hurting us?" argument ends, because the data answers it.
Suspect your streams need separating?
Bring your traffic mix. We'll tell you the smallest split that isolates what matters, and run the migration without a cliff.