Cold Outreach That Does Not Burn the Sending Domain
Cold outreach fails at the infrastructure layer long before anyone reads the first line. Copy gets blamed. The damage actually sits in the sending domain. We run our own fleet of sending servers, and the pattern repeats itself: a campaign goes out too fast from a domain with no history, the big filters react, and every message after that one pays for it. Reputation attaches to the domain and the sending IP, not to the campaign that caused the mess. A bad week in March follows you into June.
Symptoms arrive in a fairly predictable order. Soft bounces climb first. Acceptance slows down, messages sit in the queue longer than they did last week. Then placement shifts to the bulk folder, and eventually the big filters start handing back outright 5xx rejections. Recovery runs slower than the damage did. Filters release a domain on their own schedule, and no support ticket speeds that up.
Keep the corporate domain out of this. The address your invoices go out on should not share reputation with the address your outreach goes out on. Separate the sending identity and a burnt outreach domain stays a contained problem.
Authentication Before the First Send
Set up authentication before volume, not after the first block. SPF has a hard lookup limit, and records creep past it quietly as vendors get added and never removed. Audit the record, list only what actually sends, and drop the mail service you stopped using two years ago.
DKIM should sign with a key belonging to the sending domain. Rotate selectors on a schedule you decide in advance. Verify the signature survives forwarding, because a broken signature after a mailing list hop looks exactly like no signature at all from the filter’s side.
Start DMARC at p=none with reporting turned on. Then actually read the aggregate reports. They name every source sending under your domain, including the ones you forgot about (there is always at least one). Move to quarantine, then reject, once those sources are clean.
Return-Path alignment catches people out. A bounce address on a vendor subdomain breaks DMARC alignment even when DKIM passes cleanly. Check it. And confirm PTR with forward-confirmed reverse DNS on the sending IP - a missing PTR is one of the cheapest rejections you can possibly earn at the largest providers. The rest of our email deliverability guides go deeper into each of these records.
Tip: send a test message to a seed address you control and read the full headers before the first campaign, not after the first block.
Warmup Is a Schedule, Not a Switch
A new domain and a new IP both start with no history. Filters exist to catch volume spikes from unknown senders, and an unwarmed launch is exactly that. Ramp by consistent daily increments and hold the pattern steady. Irregular bursts undo three weeks of trust in an afternoon.
Warm the domain with mail people actually open before you warm it with outreach. Transactional messages, internal correspondence, anything that generates genuine engagement teaches the filters something true about the sender.
Across our fleet we keep dedicated IP ranges per client segment. One aggressive sender then cannot drag down the neighbours sharing that space.
- Verify SPF, DKIM, DMARC and PTR, then read a seed message’s headers.
- Send to a small seed set and confirm placement across providers.
- Open a verified segment at low volume, one provider at a time.
- Increase daily volume in even steps, never doubling overnight.
- Hold at the level your reply rate genuinely supports.
Reciprocal warmup networks inflate engagement signals without producing a single conversation. We do not run them.
List Hygiene Decides the Bounce Rate
Bounces burn domains faster than anything else on this page. Scraped lists carry dead addresses and spam traps in the same file, and you cannot tell them apart by looking.
Verify before the send, not after: syntax, MX presence, catch-all detection, role accounts. Role addresses like info@, office@ and biuro@ accept everything and complain more than named mailboxes, so treat them as a separate risk tier with its own decision. Recycled spam traps sit in old exports, quietly waiting. A two-year-old file is not a list, it is a liability. The same discipline runs through everything we write about cold email outreach.
Suppress hard bounces permanently, across every campaign, not just the one that produced them.
- Hard bounces from any prior campaign, on any list.
- Unsubscribes, including ones handled outside the platform.
- Role accounts you cannot justify contacting.
- Addresses with no verified MX record.
- Contacts older than your consent record actually covers.
Consent Is Not Optional in B2B
Two separate legal layers apply here. RODO gives the basis for processing the data. The Polish electronic communications act adds a distinct requirement for the message itself: article 398 requires prior consent for direct marketing sent to a terminal device, and a company mailbox is a terminal device.
The myth that B2B is exempt costs more than any bad subject line ever will. A company address tied to a named person is personal data. Keep the consent record properly: source, timestamp, the exact wording shown to the person, and a way to produce all of it on request.
Complaints correlate with missing consent. Filters see the complaint and adjust reputation. Regulators ask for the record and adjust something else. Both readings point the same direction.
So treat compliance as a deliverability feature rather than legal overhead. A list built on real consent bounces less, complains less, and replies more.
Sending Patterns That Filters Read as Human
Volume per mailbox per day matters more than total campaign size. Spread across mailboxes and domains instead of pushing a single one harder. Throttle by recipient provider too, because the big filters differ in tolerance and in retry behaviour.
Reply handling closes the loop. An outreach domain that never receives mail looks like a broadcast channel, and filters notice the asymmetry. Keep the message plain: minimal HTML, no tracking-heavy template, no link shortener sharing reputation with every other sender using it. One link per message where you can manage it, pointing at a domain you control.
We run our own campaigns on the same infrastructure we sell, which is why the MailCraft email sending platform is built around per-domain controls rather than a single global throttle, and why per-domain sending limits are part of every plan we offer rather than an upsell.
Tip: monitor bounce codes per provider, never as one average. A rising 5xx rate at a single provider is a block forming; an average hides it until the whole list is affected.
When the Block Arrives
Stop sending to the affected provider first. Continuing into a block deepens it, and every retry adds evidence against you.
Read the SMTP response text carefully. The big filters name the reason and usually the remediation path, in plain words, right there in the rejection string. Delisting requests work when the underlying cause is fixed. They fail when the same list is still queued behind them.
Establish what is actually blocked: the IP, the domain, or the content signature. Each needs a different fix, and guessing wastes a delisting attempt you only get so often. If the IP is the problem, start by checking it against the DNSBL blacklists and working through their delisting forms.
From experience, recovery is a ramp rather than a resume. Go back to warmup volumes for that provider specifically and rebuild.
Automated per-provider reputation dashboards are on our roadmap and not in the panel today. For now this is manual log reading, and we would rather say so.
Summary: Boring Infrastructure Outlasts Clever Copy
Authentication, warmup, hygiene and consent do more for reply rates than any subject line test you will run this quarter. A domain kept clean for a year is worth more than a campaign that converts once and takes the sender down with it.
No provider can promise inbox placement. Anyone offering to bypass filters is selling the fastest available route to a block, and you own the consequences after they cash the invoice.
Build the sending setup you can still use next year.
