What a Sending Domain Warm-Up Actually Looks Like Week by Week

A new sending domain has no enemies and no friends. Mailbox providers score reputation per domain and per IP, not per campaign, so the first message you send from a fresh domain arrives with nothing behind it. That absence is not neutrality. Filters treat unknown senders with more suspicion than known mediocre ones, because unknown is exactly what a throwaway spam domain looks like on day one. Warm-up is the work of building history in a shape those systems recognise as normal traffic.

What the receiving side watches is a short list: the volume curve over time, the hours you send, bounce rate, complaint rate, and whether people open, click and reply. Sudden volume from an unfamiliar domain is the single strongest trigger for a filter to act. And no, warm-up is not a trick or a bypass. It removes the reasons to distrust you, one signal at a time - the same reasons that legitimate mail into the spam folder .

Week Zero: Technical Setup Before the First Send

Nothing is warmed up until authentication is correct. SPF must list every source that sends for you. DKIM needs a dedicated selector for this domain, not a shared one inherited from a platform. DMARC starts at p=none with a rua address someone actually reads. Check the Return-Path sits on your sending domain - otherwise SPF alignment fails and DMARC passes on DKIM alone, if at all. This is also the minimum bar set by the Gmail and Yahoo rules for bulk senders, so there is no version of warm-up that skips it.

The IP needs a PTR record that resolves back to matching forward DNS. On our own sending fleet we set this before an IP ever touches customer traffic. Your tracking and link domain belong on the same organisational domain as the From address. A shared shortener drags someone else’s reputation into your mail, and you inherit whatever the last tenant did with it.

1. Publish and verify SPF, DKIM, DMARC, PTR.
2. Create seed inboxes at the large providers.
3. Wire bounce and complaint processing into suppression.
4. Register for available feedback loops.
5. Segment the list by engagement recency.

Tip: send to your seeds first and read the full headers, not the rendered message. Authentication results live there.

Week One: Small Volume, Your Most Engaged Contacts Only

Begin with people who opened or clicked in the last thirty days. If you have transactional-style recipients waiting on something they asked for, they belong in the first tranche. Keep daily volume low and constant. Resist climbing inside the first week - you are establishing a baseline, not proving how fast you can go.

Send at the same hours every day. Irregular bursts read as automation running loose, and that pattern gets scored. Split the opening days by provider so a single filter never sees your whole list at once. Two hundred messages to one large provider teaches it more than two hundred spread thin across twenty domains.

Watch bounces from the first send. A hard bounce rate above a low single-digit share means your list is the problem, and no ramp fixes a bad list. Stop, clean, restart.

Hold the newsletter. One-to-one style content, plain text, from a real person, earns replies. Replies are among the strongest positive signals a mailbox provider can observe, and week one is when they count most.

Weeks Two and Three: Increasing Volume Without Breaking the Curve

The ramp has a rough shape: roughly double daily volume every two to three days while metrics hold. The moment anything moves the wrong way, hold flat. Flat is a valid week. Going backwards costs less than pushing through a deferral wall.

Growth is measured per provider, never globally. I have watched a domain sit comfortably at one large provider and get throttled at another the same afternoon. Treat each as a separate account with its own budget.

Read deferrals as instructions. A 4xx response means slow down. Retry harder against a temporary failure and you convert a soft signal into a block. Keep the engaged-first ordering as you grow, because every new tranche is slightly colder than the one before it.

Bulk folder placement here is normal and recoverable. It is not a block. The two need different responses.

Tip: keep a plain log of daily volume, deferral text and bounce codes per provider. The SMTP response string tells you which filter reacted and to what.

Weeks Four to Six: Reaching Real Volume and What Usually Goes Wrong

This is where most warm-ups break. Three weeks of careful curve, then one full campaign to the entire list, and the domain is back where it started. The failures repeat themselves:

• An imported list nobody verified.
• A re-engagement blast to addresses collected two years ago.
• A content or template change that trips content filters mid-ramp.
• Sending from a second subdomain that was never warmed at all.

Complaint rate gets you blocked faster than anything else on that list. Unsubscribe must be one click and honoured immediately, not at the next export.

Blocks look like three things from our side: a 5xx with a policy URL, a sudden deferral wall, or silent bulk placement with no error at all. Getting out follows one order. Stop sending to that provider. Fix the cause. Wait. Resume at a fraction of previous volume. Delisting forms work only once the cause is gone - submit one before that and you have wasted the request.

Consent belongs in this technical picture, not in a legal footnote. RODO and PKE artykuł 398 require it, B2B included, and a list built without it produces complaints no ramp can absorb.

After the Warm-Up: Keeping the Reputation You Built

Reputation decays with silence. A domain that sends nothing for several weeks is partly cold again, and the next campaign pays for the gap. Keep a regular rhythm instead. Suppress inactive recipients rather than mailing them to wake them up - that tactic reliably produces complaints from people who forgot you ever existed. If you want to go deeper on the mechanics behind these signals, our deliverability articles cover them one by one.

Separate your streams by subdomain: transactional, marketing, notifications. One bad campaign then cannot take down password resets. Costs a little setup, saves an entire category of incident.

Once DMARC reports come back clean, move from p=none toward quarantine, then reject. In that order, reading the aggregate reports between steps.

Teams that would rather not manage this alone can run it on our dedicated sending server fleet, where IP allocation and PTR are handled before the first message leaves. Being honest about limits: we automate the curve and the suppression logic. Reading SMTP logs during an active block still needs a person. Automated per-provider throttling based on live response codes is standard practice in the field, and it is on our roadmap rather than in the panel today.

Summary: The Warm-Up in One Page

• Week zero: SPF, DKIM, DMARC, aligned Return-Path, PTR, seeds, suppression wiring.
• Week one: small constant volume, most engaged contacts only, same hours, reply-friendly content.
• Weeks two and three: controlled ramp per provider, hold flat on any bad signal, log every deferral.
• Weeks four to six: full volume, watch complaints, treat blocks as cause-and-fix rather than retry.
• After: steady rhythm, stream separation, DMARC enforcement.

Two rules decide the outcome. List quality first, curve second. Get the order wrong and the schedule stops mattering, because a clean ramp to bad addresses still ends in a block.

No warm-up compensates for a list gathered without consent. That is arithmetic, not compliance theatre.

Treat warm-up as a habit, not a project with a finish date. The domain you warmed in six weeks stays warm only while you keep sending like someone worth delivering to.