Email Deliverability Strategy: Owners, Streams and Rules

A framework for an email deliverability strategy: ownership, separate sending streams, DMARC enforcement, complaint thresholds and a quarterly review.

Published on

•

Email Deliverability Strategy: Owners, Streams and Rules
Do not index
Do not index
The symptoms rarely arrive as one problem. A rep forwards a customer complaint that a password reset never arrived, lifecycle notices Gmail opens slipping, and finance asks why promotional revenue is flat while send volume doubled. Those look like three unrelated fires, and they usually share one cause: a company that added sending streams faster than it added owners and rules.
An email deliverability strategy is the operating model that connects those streams. It covers who is accountable for sender reputation, which domain carries which kind of mail, how authentication gets enforced, who may add an address or a sending tool, what gets reviewed weekly, and what happens in the first hour of an incident. Teams that build it stop firefighting; teams that skip it rebuild the same fixes every quarter.
Table of Contents

Map Each Sending Stream to an Owner and a Domain

The most expensive version of this problem is a single domain carrying everything. Password resets, the monthly newsletter and a brand-new outbound sequence all leave from the primary company domain. A rented list goes into the sales tool, complaints climb at one provider, and within a week the resets customers need in order to log in are filtered alongside the promotions. No code changed. The blast radius did.
Architecture comes first because every later decision depends on it. Inventory each stream the company sends, then record four facts per stream: the domain or subdomain it uses, the platform that sends it, the person who approves changes, and the volume rule that governs it.
Stream
Sending domain pattern
Owner
Main risk if mixed
Transactional (password resets, receipts, invoices)
Dedicated subdomain on its own provider
Product or engineering
Complaints from campaigns lower trust in every reset and receipt
Marketing (lifecycle, newsletters, promotions)
Separate subdomain from transactional
Marketing or CRM lead
Stale lists and high frequency drive complaints across the domain
Outbound (cold outreach)
Separate domains, never the primary company domain
Sales or growth ops with deliverability sign-off
Volume and list quality can blacklist the main domain
Corporate mail (Microsoft 365 or Google Workspace)
Primary company domain
IT administrator
Bulk tools sending as the company hurt the mail staff rely on
The logic behind the table is separation. A complaint spike in promotional mail must never be able to lower the reputation of password resets and invoices. Outbound is where teams get careless, so cold sequences belong on separate domains that never appear in the register as the company's primary domain, with volume capped per mailbox and sourcing checked before anything sends (cold email infrastructure).
Ownership is the other half of the register, and the half that gets skipped. One accountable executive signs off on deliverability risk, one operator watches the numbers every week, and a written list names who may add a new sending tool.
Without that list, the failure is predictable. Marketing ops owns the platform, engineering owns DNS, and nobody owns the outcome, so a new vendor sends from the main domain for three weeks. Write the one-page register before changing a single setting.

What We Check First and in What Order

Diagnosis follows the register. These four checks run in this order, because the first two can take down every stream at once, and no subject line test survives that.
notion image
  1. Authentication per stream. Healthy: SPF and DKIM pass for every sending platform, alignment with the visible From domain holds, and DMARC aggregate reports arrive for the domain. Red flag: a report showing an unknown IP sending as the company, or a stream whose DKIM selector nobody on the team recognizes. That finding blocks any enforcement change until it's explained.
  1. List sources. Healthy: every address has a recorded origin and a last-engagement date. Red flag: a cohort nobody can trace back to a form, an event or a signed contract.
  1. Reputation signals. Healthy: complaint rate flat or falling, hard bounces low, deferrals confined to one provider for a day. Red flag: complaints trending up week over week, or a new blocklist listing on a sending IP.
  1. Placement by provider. Healthy: Gmail, Microsoft 365, Yahoo and the corporate domains holding within a few points of each other. Red flag: one provider dropping while the rest hold, which points at volume, a stream or infrastructure rather than at copy.
The full ten-point version sits in the deliverability audit checklist, which is worth working through once as a team rather than once per person. After the four checks, the decision is sequencing: whichever stream carries the highest revenue risk gets fixed first, and it's usually the one nobody was monitoring.

Build the Authentication Baseline and a DMARC Roadmap

Authentication is a policy system that runs continuously. The records get published once, then senders change, vendors get added, and alignment drifts. Across the wider population of sending domains, coverage remains uneven enough that aggregate reports keep surfacing sources nobody on the team remembers authorizing.
A message authenticates through alignment: SPF and DKIM have to line up with the visible From domain, because a pass on a domain the recipient never sees satisfies nothing. The RFC 7489 DMARC specification states that the published policy applies only to messages failing both aligned SPF and aligned DKIM, and that the pct tag lets a domain owner expose a share of failing traffic instead of all of it. That mechanism is what makes a staged rollout possible instead of a single switch.
notion image

The three stages

Stage
Published policy
Move on when
Watch for
1. Inventory and align
none, with reports sent to the team
Every legitimate sender passes SPF or DKIM with alignment
Unknown sources in aggregate reports
2. Partial enforcement
quarantine, applied to a percentage of traffic
Remaining failures come only from senders being fixed
Legitimate platforms still failing alignment
3. Full enforcement
reject or quarantine across all traffic
Two review cycles pass with no legitimate failures
New vendors added later that send from the domain
The percentage in stage two is where the pct tag earns its place. Rather than guessing at 10 or 50, set it so the exposed share stays small enough that a wrongly failing platform loses a handful of messages rather than a day's worth, then raise it as the failure list shrinks.
Do not tighten policy while unknown sources still fail. Aggregate reports answer three questions worth asking every week: which IP addresses can't be placed, which third-party platforms were never added to the inventory, and whether a forwarding service is breaking SPF on the way through. Before each stage change, verify alignment and report delivery with the free DMARC checker from mailX, MailAdept's sister brand. The check confirms which senders align, because a published record alone proves nothing about the traffic behind it.
The cost of skipping the sequence is quiet. A stream that still fails alignment when policy moves to quarantine or reject starts losing messages with no bounce to point at, so password resets and invoices vanish and the support queue becomes the monitoring system. Re-run the sender inventory after every new vendor, because an unlisted sender is the most common reason a stage stalls.

Set List Sourcing, Sunset and Volume Rules

Who receives mail decides long-term placement more than any subject line. Four rules carry most of the weight.
Acquisition. Addresses come from explicit sign-up or a documented business relationship, with the source stored on the record. Purchased and scraped lists never enter the marketing stream, and a partner list arrives as its own cohort, reviewed before its first send. On the outbound side the sourcing question is just as decisive, and how to build signal-based prospect lists is a useful starting point for qualifying an address before it reaches a sending tool.
Sunset. Each stream gets an engagement window, agreed with the team rather than copied from a template. A cohort with no opens or clicks inside that window moves into a re-permission sequence, and if it stays silent, it gets suppressed. Removing people feels like shrinking the list. Keeping them feeds complaints that cost far more than the impressions they add.
Suppression. Hard bounces come out immediately, complainers get suppressed across every stream and every platform, and unsubscribes propagate within hours rather than at the end of the month.
Volume and frequency. New domains ramp gradually, sends are capped per stream, dormant cohorts never receive a large one-off blast, and the campaign calendar is aligned so marketing, lifecycle and product mail don't stack on the same day.
The legal floor is lower than these rules and still binding. The FTC's CAN-SPAM compliance guide requires a working opt-out honored within 10 business days, and notes that each separate violating email may incur a penalty of up to $53,088. Consent capture and preference design are covered in more depth in email list building.
Close with the written rule: who may add an address, and who may release a suppressed cohort.

Weekly Monitoring and the Incident Playbook

The weekly review exists to catch a problem while it's still an annoyance rather than a provider block. Google's bulk sender rules supply the thresholds that matter most: senders should keep the user-reported spam rate below 0.1% and prevent it from reaching 0.3% or higher, and the rate is calculated daily. Since June 2024, bulk senders above 0.3% become ineligible for mitigation and regain eligibility only after staying below 0.3% for seven consecutive days, as set out in Google's bulk sender guidelines.
That last detail changes how a team reads its dashboard. A monthly average hides the single day that breaches the line, so complaints get checked daily and discussed weekly. Placement gets compared by provider, never in aggregate, because a drop at Microsoft 365 or a corporate domain can hide behind open rates that look flat. A sustained rise in deferrals from one provider is throttling or a volume change, not a copy problem.
notion image
Metric
Review cadence
Trigger
First action
Complaint rate
Daily, reviewed weekly
Approaching 0.1% of sends; 0.3% or higher is a stop
Find the segment or campaign driving it and suppress or pause it
Bounce rate
Weekly
Hard bounces above 2%
Clean the list source and suppress hard bounces at once
Deferrals by provider
Weekly
Sustained rise at one provider
Check throttling and any volume change for that provider
Placement by provider
Weekly
Any provider-level drop
Compare with stream, campaign and infrastructure changes
Blocklist status
Weekly
Any new listing
Isolate the affected stream, then request delisting after the cause is fixed
DMARC pass rate
Weekly
Drop in aligned passes
Look for a new sender or a changed vendor
The 2% bounce line is common guidance rather than a provider rule, so treat it as a tripwire. Something is wrong with the source, not with the recipients.

The six-step incident playbook

  1. Confirm the signal and its scope. Pull the stream, the provider and the start date. A complaint rise that begins on a Tuesday and touches one provider usually points at whatever launched on Monday.
  1. Pause the segment, not the program. Stop the campaign or cohort that lines up with the change and let the rest of the stream keep sending.
  1. Check transactional mail separately. Confirm resets, receipts and invoices are still landing on their isolated stream before anything else.
  1. Find the cause. A new blocklist listing, a dropped DMARC pass rate, or deferral text that names reputation or volume.
  1. Fix the source. Remove the risky cohort, correct the failing authentication, or reduce the volume going to that provider.
  1. Watch seven consecutive days below the threshold before scaling back up.
Say example.com mails a promotional campaign to an old imported list, which is an example rather than a client story. Complaints climb at one provider and stay up for three days. Password resets keep landing the whole time, because they never shared that domain or that stream. That's the architecture paying for itself.
Write the warning and stop levels down in advance and attach an owner to each step. Skipping that turns repeated incidents into provider-level blocks that take weeks to clear.

Common Mistakes That Undo a Deliverability Strategy

Four habits break more strategies than any technical failure.
  • Reporting delivery rate as placement. Accepted isn't seen. A team that reports only delivery will watch the spam folder fill up and describe it as healthy volume.
  • Treating list growth as the scoreboard. New sign-ups look excellent while dormant cohorts keep generating complaints, and the revenue from those cohorts rarely covers the reputation they cost.
  • Assigning deliverability to everyone. When marketing, engineering and IT each assume another team is watching, nobody acts in the week a trigger fires. A small bounce spike becomes a block.
  • Waiting for the provider to tell you. Bulk-sender enforcement runs on daily rates. A team reviewing the dashboard monthly learns about the problem after the account has already lost eligibility.
Each one has a bill attached: lost revenue, support tickets from missing resets, or a sender reputation that takes months to rebuild. None of them are dramatic in the moment, which is exactly why they persist.

Quarterly Review and When to Bring in an Expert

The quarterly review is where the register gets refreshed and the strategy gets adjusted rather than inherited. Six items belong on the agenda, each with a named owner and a date:
  • Stream register. New streams, retired tools, changed platforms and any vendor that started sending since the last review.
  • DMARC stage per domain. Confirm which domains sit at monitoring, which at partial enforcement, and what blocks the next step.
  • Sunset results. Which cohorts were suppressed, how many re-permissioned, and whether the window still matches real engagement.
  • Trends by provider. Complaint and placement movement over the quarter, not the week.
  • Open incidents. Close them formally, with a note on what changed.
  • New vendors. Log them before their first send rather than after their first complaint.
The weekly rhythm feeds it. The operator updates the metrics table, and the accountable executive reviews changes and priorities at the scheduled weekly call, where fixes, risks and next actions get worked through with whoever is doing the sending.
Four signals suggest the internal model has run out of room: repeated incidents inside one quarter, a planned ESP or domain migration, provider blocks or deferrals that persist after the first fix, and no clear owner even once the register is written. At that point the choice is between a snapshot and a system. A one-time deliverability audit identifies issues at a specific moment in time, while ongoing deliverability management keeps the setup working through incidents, vendor changes and migrations, with an expert embedded in the team and a weekly review. If none of those four signals applies, the register and the review cadence are probably enough.

Frequently Asked Questions About Deliverability Strategy

How much should a company budget before email is a revenue channel? Price it against blast radius and the revenue mail at risk. The first spend is ownership time and monitoring for every stream; add outside help when password resets, receipts or outbound replies would be hurt by one failing campaign. A tool that only reports a drop is cheaper than an incident, but it does not reduce the cost if nobody is assigned to act.
What team size needs this model? The trigger is stream count and vendor count. Once a second team, platform or list source can send, the register and owner are needed even if the same person wears both hats. A single operator can run the weekly checks for a small setup; an accountable executive becomes necessary when risk crosses marketing, product and sales.
What if a vendor will not move off the primary domain? Treat refusal as a risk decision with a date. Cap its volume, keep it away from transactional and outbound streams, require complaint and bounce reporting, and review it at renewal. If it cannot isolate sends or share data, plan a replacement rather than let its reputation set the ceiling for receipts.
Should sending pause during an incident? Protect transactional mail first. Slow or stop the flagged stream, keep each change small and reversible, and document which provider, domain and campaign moved. Chasing a weekly target by pushing more volume through a falling stream usually lengthens recovery.
What does a healthy month look like? No unlogged sender has mailed, every stream still has an owner and domain, incidents close with causes, and list sources remain approved. The proof is isolation: one bad campaign should stay contained without touching password resets or receipts.
When is an expert cheaper than another dashboard? When the checks are known but the company lacks a named operator to run them weekly, outside help buys labor and accountability. MailAdept places a dedicated deliverability expert inside the team to audit the setup, fix what hurts placement and review risks every week. Start with the free audit at MailAdept; the operating question stays the same: can a failing stream be isolated, and will somebody act that week?

Fix Your Email Deliverability Before It Costs You Revenue

Get expert insights on why your emails go to spam and how to consistently reach the inbox.

Get a Free Deliverability Audit