Transactional vs marketing email: why you must separate the streams

Gmail does not know which of your emails came from the marketing platform and which came from your application code. It sees a domain, a signing key, and an IP address. If those are the same for both streams, a bad campaign week can put your password resets in the spam folder — and this guide covers the architecture that prevents it.

Updated July 31, 202612 min read
  • Why providers see one sender
  • Subdomain architecture
  • DNS records per stream
  • A migration that does not reset reputation

Key takeaways

  • Mailbox providers build reputation against the sending domain, the DKIM signing domain (d=), and the sending IP — not against your internal org chart.
  • Marketing mail reliably attracts complaints; transactional mail reliably does not. Blended, the marketing complaints set the score for both.
  • The fix is structural: a distinct subdomain per stream, each with its own DKIM selector and key, ideally on separate sending infrastructure.
  • Separate subdomains also give you separate Google Postmaster Tools entries, so you can see a reputation problem before support tickets tell you about it.
  • Subdomains inherit from the parent domain and are not a way to escape an already-damaged reputation — they are a way to avoid creating one.

Inside your company, transactional and marketing email are obviously different things. One comes out of application code, owned by engineering, triggered by user events. The other comes out of a campaign tool, owned by marketing, scheduled against a segment. Different systems, different teams, different review processes.

None of that is visible to Gmail. A receiving mail server evaluates three identifiers: the domain in the From: header, the domain in the DKIM signature, and the IP address the connection came from. If both of your streams present the same three, they are the same sender, and they get one reputation between them.

That shared score is fine right up until it is not. Marketing email attracts complaints as a normal condition of doing business — some fraction of any list will hit "report spam" rather than unsubscribe. Transactional email attracts almost none. Blend them and the complaint rate that governs your password reset delivery is being set by your newsletter.

The real differences between the two streams

TransactionalMarketing
TriggerA specific user action or account eventYour campaign schedule
AudienceOne personA segment or list
ConsentImplied by the transaction or relationshipExplicit opt-in
Recipient expectationActively waiting for itTolerating it at best
Typical complaint rateNear zeroMeaningful and unavoidable
Typical engagementVery high — the message was requestedLow to moderate
Unsubscribe obligationNone for genuine transactional mailRFC 8058 one-click required for bulk senders
Cost of a delayed messageA blocked user and a support ticketA slightly worse campaign
Volume shapeSteady, follows product usageSpiky, follows the calendar

The last row matters more than it looks. Transactional volume is smooth because it tracks real usage. Marketing volume is bursty by design — a campaign sends a month of transactional volume in twenty minutes. Receiving providers treat sudden volume spikes from a given sending identity with suspicion, and when the two streams share an identity, your campaign spike is throttling your receipts.

How mailbox providers decide who you are

Reputation is not attached to your brand, your company, or your account with an email provider. It is attached to identifiers that appear in the message and the connection carrying it.

  • The DKIM signing domain. The d= tag in the DKIM-Signature header names the domain that cryptographically vouches for the message. This is the most durable identifier available to a receiver, because it survives forwarding and cannot be forged without the private key.
  • The From: header domain. What the recipient sees, and what DMARC requires the authenticated domain to align with.
  • The SPF domain. The envelope sender domain checked against the connecting IP — usually a return-path subdomain controlled by your provider.
  • The connecting IP address. Still weighted heavily by Microsoft; Gmail and Yahoo have shifted emphasis toward domain reputation, but IP reputation has not stopped mattering.

What blending actually costs

Work the numbers. Suppose you send 400,000 marketing messages and 100,000 transactional messages in a month, all from yourdomain.com with a single DKIM key. The marketing stream runs at a 0.15% complaint rate, which is unremarkable for opted-in promotional mail. The transactional stream runs at 0.005%.

Six hundred marketing complaints and five transactional complaints land on the same identity. The blended rate is 0.121% — comfortably past the 0.10% figure Google advises staying under, and creeping toward the 0.30% threshold where throttling begins. Your receipts did nothing wrong and are now being scored as if they did.

Now run one bad month. A poorly segmented campaign to a re-engagement list pushes the marketing complaint rate to 0.4%. The blended rate crosses 0.32%. Gmail begins throttling the identity, and the mail that gets throttled includes the password resets. Support volume rises, and nobody on the engineering side has changed a line of code.

Blended rate

0.121%

worked example above, one identity

Google target

< 0.10%

published guidance

Enforcement threshold

0.30%

published Gmail and Yahoo cap

Separated tx rate

0.005%

the same mail, own identity

The asymmetry is the point. Separation does not just protect transactional mail from marketing mail — it also lets marketing be evaluated honestly rather than being propped up by transactional engagement it did not earn. Both streams get judged on their own behaviour, which is what you want when you are trying to diagnose anything.

The subdomain architecture

The standard pattern is one subdomain per mail stream, each fully authenticated in its own right, all under the same organisational domain so the brand relationship stays visible to recipients.

StreamExample subdomainWhat sends from it
Transactionalmail.yourdomain.com or tx.yourdomain.comReceipts, password resets, verification, security alerts
Marketingnews.yourdomain.com or go.yourdomain.comCampaigns, newsletters, promotions, lifecycle nurture
Product notificationsnotify.yourdomain.comHigh-volume discretionary alerts, digests, mentions
Corporate mailyourdomain.com (root, via your mailbox provider)Human-to-human mail from employees
Keep the root domain for human mail. Machine-generated streams belong on subdomains where their reputation can be scored — and damaged — independently.

Where product notifications belong

Notification email is the awkward middle case. It is transactional in mechanism but discretionary in value, and at high volume it accumulates complaints in a way receipts never do. If your product sends a lot of it — comment notifications, digests, activity summaries — give it a third subdomain rather than letting it share with mail that genuinely cannot fail. The cost is one more DNS block; the benefit is that a noisy notification stream cannot take your password resets with it.

DNS setup for each stream

Each sending subdomain needs its own complete authentication set. They do not share records, and they should not share DKIM keys — separate keys are what make the identities separable in the first place.

Per-subdomain DNS (illustrative)

; --- Transactional stream ---
mail.yourdomain.com.            TXT   "v=spf1 include:spf.supersendtx.com -all"
ss1._domainkey.mail.yourdomain.com. TXT "v=DKIM1; k=rsa; p=MIIBIjANBg…"
rp.mail.yourdomain.com.         CNAME rp.supersendtx.com.

; --- Marketing stream (different provider, different key) ---
news.yourdomain.com.            TXT   "v=spf1 include:spf.your-esp.example -all"
sel1._domainkey.news.yourdomain.com. TXT "v=DKIM1; k=rsa; p=MIIBIjANBg…"

; --- One DMARC policy at the organisational domain ---
_dmarc.yourdomain.com.          TXT   "v=DMARC1; p=reject; sp=quarantine; rua=mailto:dmarc@yourdomain.com; adkim=s; aspf=s"
  • · Each subdomain has its own SPF record — SPF does not inherit from the parent.
  • · Distinct DKIM selectors mean distinct key pairs, which is what separates the identities.
  • · The return-path CNAME keeps SPF aligned with the visible From domain.

One DMARC record at the organisational domain covers every subdomain unless a subdomain publishes its own. The sp= tag sets the policy applied to subdomains specifically, which is useful during a rollout: you can hold the organisational domain at p=reject while a newly separated marketing subdomain runs at sp=quarantine until its alignment is confirmed clean.

The adkim=s and aspf=s tags request strict alignment, meaning the authenticated domain must match the From: domain exactly rather than merely sharing an organisational parent. Strict alignment is the right end state, but turn it on only after aggregate reports confirm every legitimate source is aligned — otherwise you will reject your own mail.

Separating the infrastructure layer

Subdomain separation partitions domain reputation. It does not partition IP reputation, and Microsoft in particular still weights IP heavily. If both streams leave from the same IP addresses, you have solved half the problem.

The complete version routes each stream through infrastructure that only carries traffic of that kind. In practice that means using a provider whose transactional network does not also carry marketing blasts, and — above a certain volume — moving to IPs that are yours alone.

Separation levelWhat it protects againstWhen it is enough
Same domain, same IPsNothingNever — this is the default you are trying to leave
Separate subdomains, same IPsDomain-level reputation bleed between your own streamsLow volume, or where the provider pool is high quality
Separate subdomains, separate poolsDomain bleed, plus IP bleed between your two streamsMost growing products
Separate subdomains, dedicated infrastructure for transactionalAll of the above, plus other tenants entirelyWhen transactional mail is business-critical, or a customer needs to allowlist your IPs

This is the design SuperSend TX is built around. The production pool carries transactional mail from paying customers only — no free-tier signups, no campaign traffic, no cold outbound — so the neighbours on the shared lane are senders with the same incentives you have. Above that, Dedicated gives you your own mail servers and IP addresses, provisioned, warmed through a full ramp, and operated by us, with the same POST /emails contract so nothing in your application changes.

Migrating to separated streams without a reputation dip

A new subdomain has no sending history, which means it needs the same gradual introduction any new sending identity does. Cutting 100% of transactional volume over on a Tuesday afternoon will produce exactly the throttling you were trying to avoid.

  1. 1

    Publish DNS and verify before sending anything

    Add the SPF, DKIM, and return-path records for the new subdomain and confirm the provider reports it verified. Leave the old sending path untouched.

  2. 2

    Move the highest-engagement stream first

    Password resets and verification mail have near-perfect engagement and near-zero complaints. Sending them first teaches receivers that the new identity is trustworthy, in the most favourable conditions available.

  3. 3

    Ramp over two to four weeks

    Shift roughly 10%, 25%, 50%, then 100% of transactional volume, holding at each step for several days. Watch bounce and complaint rates before each increase. The warmup guide has the full ramp mechanics.

  4. 4

    Register the new subdomain in Postmaster Tools

    Add it to Google Postmaster Tools on day one. Domain reputation data takes time to populate, and you want the history to start when the sending does.

  5. 5

    Move marketing last, and separately

    Once transactional is fully migrated and stable, move the marketing stream to its own subdomain on its own ramp. Never move both at once — if something degrades you will not know which change caused it.

  6. 6

    Tighten DMARC after alignment is confirmed

    With both streams aligned and reporting cleanly for 30 days, move the policy from p=none toward p=quarantine and then p=reject.

  • Each stream has its own subdomain and its own DKIM key pair
  • Each subdomain has its own SPF record, under ten DNS lookups
  • DMARC published at the organisational domain, with sp= set deliberately
  • Transactional migrated first, ramped over weeks not hours
  • Every sending subdomain registered in Google Postmaster Tools
  • Complaint rate alerting configured per stream, not blended

Give your critical mail its own lane

Verify a transactional subdomain in minutes and send from a network that carries paying transactional customers only — no free-tier noise, no marketing blasts, no cold outbound.

Related reading

FAQ

Frequently asked questions

Should transactional and marketing email use different subdomains?

Yes. Mailbox providers score reputation against the sending domain and DKIM signing domain, so a shared subdomain means a shared reputation. Sending transactional mail from something like mail.yourdomain.com and marketing from news.yourdomain.com, each with its own DKIM key pair, lets the two be judged independently — and gives you separate Google Postmaster Tools entries to monitor.

Can I send transactional and marketing email from the same provider?

You can, provided the provider supports separate sending subdomains and, ideally, separate IP pools per stream. The risk is not the provider account — it is shared identifiers. If both streams sign with the same DKIM key and leave from the same IPs, using one provider gives you no separation at all.

Does a subdomain protect my main domain reputation?

Partially. Subdomains are evaluated somewhat independently, but they are not isolated from the organisational domain — a persistently bad subdomain can affect the parent. Subdomain separation is a preventive architecture that stops one stream from contaminating another; it is not a way to escape an existing reputation problem.

What is the DMARC sp tag?

The sp tag sets the DMARC policy applied to subdomains, separately from the p tag that applies to the organisational domain. It is useful during migrations: you can keep p=reject on the main domain while running sp=quarantine on newly introduced sending subdomains until their alignment is confirmed clean.

Where should product notification emails go?

On their own subdomain, separate from both marketing and critical transactional mail. Notifications are transactional in mechanism but discretionary in value, and at volume they accumulate complaints that receipts and password resets never do. Giving them a third subdomain keeps a noisy notification stream from affecting mail that cannot fail.

How long does it take to migrate to a new sending subdomain?

Plan two to four weeks. A new subdomain has no sending history, so volume should ramp gradually — roughly 10%, 25%, 50%, then 100%, holding several days at each step and watching bounce and complaint rates before each increase. Move the highest-engagement mail first, and never migrate both streams simultaneously.

Critical email deserves infrastructure you can name

Start on Sandbox in minutes. Move to a transactional-only Pool for production, or to Dedicated servers and IPs that we build, warm, and manage for you.