A B2B SaaS platform can prepare a DMARC policy rollout, but a customer who owns the sending domain may still control the authoritative zone. Monitoring must come before quarantine or reject, and the zone owner must approve enforcement. That constraint changes the rollout: successful publication and permission to enforce are separate decisions, and an API response cannot tell you that every legitimate sender has been accounted for.
Short answer: start DMARC in monitoring mode, read the reports for a few weeks, then advance to quarantine and finally reject. Each transition updates the same TXT record. First establish SPF or DKIM alignment; DMARC publication alone fixes neither. For customer-owned zones, make the customer's approval part of each transition. For platform-owned zones, keep the same evidence gates even though your team can publish directly.
Delivery comes first.
Who controls the zone when enforcement changes?
The useful unit of rollout is a sending domain with an identified DNS owner, a known set of legitimate senders, and someone responsible for reading reports. A tenant identifier is insufficient: your application can mark a tenant ready while the customer still has an unrelated marketing sender using the same domain, which is precisely the kind of overlooked dependency that makes premature rejection dangerous. Treat the domain inventory as operational state, with an owner and an explicit intended policy, rather than treating DNS publication as the final checkbox in onboarding.
Consider a hypothetical SaaS tenant using customer.example, alongside a platform sending domain of platform.example. These are example domains, not a record of a deployment. For customer.example, the platform prepares the desired record and the customer controls publication; for platform.example, the platform's DNS operator performs that work. Both paths need evidence that authentication aligns before enforcement advances.
Ownership changes who can act. It doesn't lower the standard of evidence.
The storage-architecture question here is familiar: what observation justifies committing the next state? A saved configuration is evidence of intent, while reports are evidence about senders. Conflating them creates a false readiness signal. Keep both, and make the transition decision inspectable.
How should DMARC policy progress from monitoring to quarantine and reject?
DMARC depends on aligned SPF or DKIM, as described in RFC 7489. An SPF pass for a domain unrelated to the visible From domain does not by itself satisfy that requirement, and merely publishing a DMARC policy does not repair a sender's alignment. In a SaaS onboarding flow, publish the appropriate SPF and DKIM configuration, establish alignment, and use monitoring to identify the rest of the sending estate before requesting stronger enforcement. The practical trap is assuming that successful application mail proves the domain is ready: the reports can reveal legitimate senders outside the application, including a marketing service nobody included in the initial inventory.
That sender matters.
Monitoring uses p=none; quarantine and reject request progressively stronger receiver treatment for mail that fails DMARC. The policy is a request to receiving systems, not a delivery guarantee, and a report is not proof that every business workflow has been exercised. A few weeks of observation is the starting recommendation here, with the decision depending on what those reports reveal. A calendar reminder cannot certify that an infrequently used sender has been tested.
| Intended state | Evidence needed before advancing | Failure mode to investigate |
|---|---|---|
| Monitoring: p=none | SPF or DKIM alignment established; reports reviewed for a few weeks; legitimate senders identified | An unlisted sender appears, or a known sender fails alignment |
| Quarantine: p=quarantine | Known legitimate senders accounted for and their alignment issues resolved | Legitimate mail is subjected to the requested quarantine treatment |
| Reject: p=reject | Quarantine observations support stronger enforcement; domain owner approves | An overlooked legitimate sender becomes subject to rejection |
Don't invent a universal pass percentage for that table. The decision depends on which traffic is failing and whether it belongs to a legitimate sender; a small amount of failure can still represent a necessary business workflow. Aggregate success is a poor substitute for identifying that workflow.
The TXT record lives at _dmarc under the domain being governed. The following Python example prints 3 proposed values for an illustrative platform-owned domain. It makes no DNS changes, and the reporting mailbox is a placeholder that must be replaced with an operational destination before publication.
domain = "platform.example" report_mailbox = "dmarc-reports@platform.example" for policy in ("none", "quarantine", "reject"): record_name = f"_dmarc.{domain}" record_value = f"v=DMARC1; p={policy}; rua=mailto:{report_mailbox}" print(f"{record_name} TXT {record_value}") Those are successive candidate values for one record, not three records to publish together. This distinction is small enough to miss during implementation and consequential enough to deserve a review comment.
Compare DNS options after deciding the ownership boundary
A DNS provider cannot determine whether a forgotten sender is legitimate. Choose the publication mechanism after deciding who owns the zone and how that owner will approve updates; otherwise, a convenient integration can quietly become an assumption that the platform has authority over every customer's domain. The comparison below describes deployment choices to evaluate, not measured differences in availability, delivery rates, or consistency.
| Option | Where it belongs in this rollout | Trade-off and boundary |
|---|---|---|
| Cloudflare DNS | Evaluate for a zone the customer or platform already operates there | Keeping publication with the existing zone owner avoids introducing a zone move into a policy change; report review still needs an owner |
| Amazon Route 53 | Evaluate for a zone whose current operational owner uses this service | Preserve that team's publication responsibility; SaaS onboarding must accommodate its approval process instead of assuming immediate platform control |
| Google Cloud DNS | Evaluate for a zone managed by the customer's or platform's cloud team | Keep policy changes within that team's existing DNS process; sender discovery and authentication readiness remain separate work |
| Infrai | A candidate when the authorized publisher wants a plain REST API for DNS record updates, with no SDK to install or client-library version to maintain | A single API key covers 295 routes across 20 modules, so a team integrating DNS alongside other backend services can avoid separate credentials for each service. That credential consolidation does not establish alignment or grant authority over a customer's zone |
The first three rows evaluate continuity of zone ownership rather than claiming a winner on features that don't decide this question. An existing customer-owned zone is a good reason to retain its current publication path. A platform-owned zone gives the SaaS operator more freedom to choose an interface, but the same monitoring-to-enforcement sequence applies. None of these choices removes the need to identify senders.
There is no delivery benchmark here that would justify ranking the providers by mail acceptance. The decision supported by the evidence concerns control of publication and the integration boundary. Evaluate other provider requirements separately before moving an authoritative zone.
For the REST option, this read-only Python inspection retrieves DNS records before an operator prepares a change. Set DNS_API_ORIGIN to the service's HTTPS origin, without a path, and INFRAI_API_KEY to your credential. The request uses the verified record-list route and prints the returned body without assuming an undocumented response schema. It doesn't publish a policy or assert that an observed value is globally visible. Because this is a read, retries cannot duplicate a DNS write. The example allows 5 attempts and a 30-second request timeout; these are client choices, not service limits.
import os import time from datetime import datetime, timezone from email.utils import parsedate_to_datetime from urllib.error import HTTPError from urllib.request import Request, urlopen origin = os.environ["DNS_API_ORIGIN"].rstrip("/") if not origin.startswith("https://"): raise ValueError("DNS_API_ORIGIN must use HTTPS") for attempt in range(5): request = Request( origin + "/v1/dns/record/list", headers={"Authorization": "Bearer " + os.environ["INFRAI_API_KEY"]}, method="GET", ) try: with urlopen(request, timeout=30) as response: if not 200 <= response.status < 300: raise RuntimeError(f"Unexpected status: {response.status}") print(response.read().decode("utf-8")) break except HTTPError as error: detail = error.read().decode("utf-8", errors="replace") if error.code != 429 or attempt == 4: raise RuntimeError(f"HTTP {error.code}: {detail}") from error delay = float(2 ** attempt) retry_after = error.headers.get("Retry-After") if retry_after: try: requested_delay = float(retry_after) except ValueError: requested_delay = ( parsedate_to_datetime(retry_after) - datetime.now(timezone.utc) ).total_seconds() delay = max(delay, requested_delay) time.sleep(delay) Roll out one domain with an explicit return path
Start with a domain whose publication owner and report reader are identified. Record its current TXT value and proposed monitoring value, confirm SPF or DKIM alignment, and publish monitoring through that owner's DNS process. Then spend a few weeks reviewing reports and accounting for legitimate senders. For a customer-owned domain, give the customer the proposed change and the evidence supporting it before requesting the next publication; for a platform-owned domain, use the same material in the team's change review.
Advance to quarantine when the sender inventory and alignment evidence support it. Move to reject only after quarantine has supplied enough evidence to justify that stronger policy. Each step changes the same TXT record, so retaining the previous value gives the operator a direct rollback target rather than requiring a rebuild. Record updates make rollback operationally straightforward, but they cannot restore a message already rejected, which is why reversibility is not a substitute for readiness.
Stop on unexplained failures.
If a legitimate sender fails after a transition, investigate its alignment and reassess the enforcement state with the domain owner. Keep the incident attached to that domain's sender inventory so the next review has more than a green configuration status. The durable result is an understood sending estate with an accountable publication process; a stricter TXT value follows from that work.
References
- RFC 7489, Domain-based Message Authentication, Reporting, and Conformance
- Cloudflare DNS documentation
- Amazon Route 53 documentation
- Google Cloud DNS documentation
Sources
The References above identify the authentication standard and the provider documentation for evaluating a zone's publication process. The rollout examples describe proposed operating procedures, not measured provider delivery outcomes.
