Warming up a new sending domain or IP in 2026: the schedule, the math, and when you don't need to.
Warmup is the slow, deliberate way you teach the mailbox providers to trust a new sender. In 2026 it is less about volume and more about engagement, and the first honest question is whether you need it at all. Below is how it works, a ramp you can generate for your own numbers, and the limits of the tools that promise to do it for you.
Email warmup is the practice of starting a new sending domain or IP address at a low volume and increasing it gradually, so the receiving mailbox providers build a reputation for you instead of treating a sudden flood of mail as the work of a spammer. It matters when you are on a dedicated IP or a new domain, and it matters less, or not at all, when you are on a healthy shared pool. The mechanics in 2026 have shifted: providers now weigh how recipients engage with your mail as much as how much of it you send, so a warmup that earns real opens and replies succeeds where a careful volume curve to an unengaged list fails. The work has three parts that have to be right together: the technical setup that comes before you send anything, the gradual ramp itself, and the engagement that turns volume into trust.
What warmup is, and what changed.
The idea has not moved in years; the thing the providers measure has, and that is what makes old warmup advice quietly wrong.
A receiving provider decides where your mail goes by asking, in effect, whether it can trust the sender. A brand-new domain or a fresh IP has no track record, so the provider has no reason to extend that trust, and a large first send reads as exactly the pattern a spammer produces: lots of mail, from nowhere, all at once. Warmup answers that by building a history. You begin with a small volume the provider can evaluate without risk, you keep the signals healthy, and you increase the volume in steps the provider has already learned to accept. Done properly, by the time you reach full scale the provider has weeks of evidence that your mail is wanted.
What has changed is the evidence that counts. A few years ago, warmup was mostly a volume exercise: ramp slowly and hope engagement held. Through 2025 and into 2026 the major providers tightened their filters and moved the weight onto behaviour, whether people open, reply to and keep your mail, and whether the sending pattern looks human rather than scripted. The effect is visible in the numbers. Industry measurement put Office 365 inbox placement down sharply year over year, into the low fifties as a percentage, and authenticated mail routinely lands in spam when engagement is thin. The practical consequence is that you cannot warm up to a list that will not engage. The volume curve is the easy part; the engagement underneath it is what the providers are reading.
First, the question most guides skip: do you even need to?
A surprising amount of warmup advice is sold to people who do not need to warm up anything. The honest filter takes two questions.
The first question is whether you are on a shared or a dedicated IP. A shared IP is already warm, kept that way by the steady traffic of every other sender on it, so there is no reputation for you to build on the IP and no ramp to run. A dedicated IP is yours alone, which is the point of it, but it also means the address starts from zero and has to be warmed. The second question is volume. A dedicated IP only earns its keep above roughly 100,000 emails a month sent consistently; below that, the traffic is too sparse to hold a reputation and you are better served by a well-run shared pool. So the test is simple: dedicated IP with consistent volume over about 100,000 a month, you warm up; otherwise, usually you do not.
This matters because warmup is often reached for as a cure when the disease is something else. A sender on a shared IP with mail going to spam does not have a warmup problem; almost always the cause is authentication that does not pass or a list full of addresses that bounce and complain. Pointing a careful ramp at a broken setup just fails more slowly and at greater cost in time. The one nuance worth keeping is that domain reputation is separate from IP reputation: a new domain on a shared, warm IP still has to build its own domain reputation through gradual, engaged sending, even though the IP underneath it needs nothing. The IP can be warm while the name on the mail is still a stranger.
Before you send anything: the prerequisites.
Warmup is the last step in setting up a sender, not the first. Begin it before these are in place and you are ramping a reputation that has nothing solid to attach to.
Authentication comes first and is not optional in 2026, since the major providers now reject unauthenticated bulk mail outright. SPF must be a single, correct record within the ten-lookup limit; DKIM must sign every message; and DMARC must be published and aligned, starting at a monitoring policy so you can see what is failing before you tighten it. The sending host needs reverse DNS that resolves and a HELO or EHLO name that matches, because a mismatch there is a classic spammer tell. The sending identity itself should sit on a dedicated subdomain kept apart from the mail your staff send internally, so that a problem on one stream cannot poison the other, and that subdomain should point at a real website rather than a parked page. Finally, the list you intend to warm up to has to be verified and genuinely consented, because everything that follows depends on those recipients engaging.
None of this is the warmup, and all of it decides whether the warmup can work. A ramp built on authentication that does not pass will stall the moment volume rises, and a ramp pointed at a list that was bought or scraped will produce the bounces and complaints that end a reputation rather than build one. The unglamorous truth is that most warmups that go wrong were lost before the first send, in the setup, and the schedule took the blame for a problem that lived upstream of it.
Generate a ramp for your own volume.
There is no single correct schedule, because the right pace depends on your target, your starting point and how clean your list is. Use this to model a curve, then let your bounce and complaint rates tell you whether to hold or push.
Reaches target on day —.
| Day | Send that day | Cumulative |
|---|
A model, not a promise. The numbers the providers care about are your bounce rate, kept under 2%, and your complaint rate, kept under 0.1%. If either climbs, hold or drop back a step; if they stay clean, you can push toward the faster end. Each dedicated IP runs its own ramp.
Domain warmup and IP warmup are two different things.
They get conflated constantly, and the conflation leads people to warm one while neglecting the other.
IP reputation belongs to the address your mail leaves from, and it is what a dedicated IP has to build from scratch. Domain reputation belongs to the name in your From address and your authentication, and it travels with you: change your ESP or your IP and your domain reputation comes along, for good or ill. The two interact but are scored separately, which produces the cases that confuse people. A new domain on a warm shared IP inherits the IP's standing but still has to earn its own domain reputation through gradual, engaged sending. A warm domain moved onto a fresh dedicated IP carries its name's good reputation but still has to warm the new address. And because each IP is scored on its own, you cannot pour one ramp across several addresses; every dedicated IP is warmed independently, on its own curve, which is why the usual planning figure is around one dedicated IP per million emails a month and why adding capacity means warming the new IP from the bottom while the established ones hold their volume.
Moving to a new IP without starting from zero.
Warmup is not only a new-sender problem. The same mechanics decide whether a migration goes smoothly or throws away reputation you already earned.
A common and avoidable disaster is the hard cutover: a sender moves to a new ESP or a new dedicated IP, points the full volume at the cold address on day one, and watches deliverability collapse exactly as it would for any unwarmed IP. The reputation built over weeks on the old address does not transfer to the new one, because IP reputation is tied to the address itself. The right approach is a staged migration that runs both addresses in parallel. You stand up the new IP and begin warming it on its own ramp while the old, warm IP continues to carry the majority of the traffic, then move volume across in increasing steps as the new IP proves itself, until the old one can be retired. For a few weeks you are effectively warming the new IP and draining the old one at the same time.
Domain reputation makes this easier than a true cold start, because the name in your From address and your authentication carry their standing with them regardless of which IP the mail leaves from. A sender keeping the same domain across a migration starts the new IP with a known-good name attached, which the providers weigh alongside the unknown IP. That does not remove the need to warm the new address, but it shortens the climb. The mistake to avoid is assuming the domain's good reputation excuses a cold IP from any ramp at all; it lowers the difficulty, it does not eliminate the step.
The ramp math, and how long it really takes.
The generator gives you a curve; this is how to choose its shape and read the time it implies.
The shape of a warmup is a small starting volume, a daily increase while the signals stay healthy, and a ceiling at your target. A new domain typically starts around fifty messages a day; a dedicated IP with a genuinely engaged list might open at a hundred. From there a daily increase in the range of twenty to thirty percent is a sound default, with a clean and engaged list supporting a faster pace and a colder or larger list demanding a slower one. The percentage is a starting assumption, not a rule, because the real governor is whether the providers keep accepting the rising volume without deferrals or complaints. The duration falls out of the arithmetic: the further your target is from your start, the more doubling steps the ramp needs, and the longer it runs.
| Target / day | Typical window | Note |
|---|---|---|
| Under 10,000 | 3–4 weeks | Most senders land here |
| 10,000–100,000 | 4–6 weeks | Watch per-provider limits |
| 100,000+ | 6–8 weeks minimum | Often multiple IPs |
| New domain, aggressive | 4–6 weeks before scale | No shortcut on a new name |
The number that should never be rushed is the time itself. A warmup that completes a week early and burns the reputation costs months to undo, while a warmup that runs a week long costs only that week. The asymmetry is the whole reason to err slow: the downside of patience is small and the downside of haste is large, so when the generator and your metrics disagree, the metrics win and you hold.
Engagement is the signal the curve is really carrying.
If the providers are weighing behaviour, then the order in which you send matters as much as the volume you send.
The single most effective move in a modern warmup is to send to your most engaged recipients first. Sort your list by how recently and how often people have opened and clicked, and start the ramp with the top of that list, the people almost certain to open and reply. Those interactions are the evidence the provider is looking for, and they teach it to trust the sender before any colder segment is introduced. As reputation builds, you widen the audience outward, adding less-engaged recipients only once the established ones have done the work of vouching for you. This is why a warmup to a freshly bought list fails immediately: there is no engaged core to start from, only strangers, and strangers produce silence or complaints rather than the opens the ramp depends on.
The same logic explains a few practices that look superstitious but are not. Sending at varied, human times rather than in a single nightly burst reads as a person rather than a script. Keeping early content genuinely useful, the kind people open and act on, manufactures the engagement the provider rewards. And resisting the urge to add your whole list at once protects the signal: a small, engaged send beats a large, indifferent one at every stage of the ramp. The volume is the visible part of warmup, but engagement is the part being scored, and a sender who organises the ramp around engagement is working with the grain of how the providers now decide.
Reading the warning signs while it runs.
A warmup is a feedback loop, not a fixed plan. The providers tell you when you are pushing too hard; the skill is in listening.
Two numbers govern whether you hold or advance. Hard bounces should stay under two percent, and a figure above that is a list-quality signal, not a pacing one: it means unverified or stale addresses are in the send, and no change to the schedule will fix a list problem. Complaints should stay under one in a thousand, and a complaint spike is the fastest way to lose ground, because it tells the provider directly that recipients did not want the mail. Watch both daily rather than weekly, because the value of monitoring is catching a problem on the day it begins instead of the week it becomes a crisis.
The other signal is the provider's own response to your volume. A 4xx deferral is a soft no, a request to slow down rather than a rejection, and the correct answer is to hold at the last volume the provider accepted cleanly, spread the sending across the day instead of bursting it, and let the rate settle before climbing again. Deferrals that clear on retry are an ordinary part of warmup. Deferrals that harden into 5xx rejections mean you have pushed past what the provider will tolerate and should drop back a step and rebuild from there. Because each provider sets its own thresholds, it is normal for Gmail, Microsoft and Yahoo to accept different volumes on the same day, which is the argument for ramping each of them on its own curve rather than a single global rate.
The big three do not want the same things.
A single global ramp treats Gmail, Microsoft and Yahoo as one audience. They are not, and the differences are large enough to plan around.
Gmail leans hardest on engagement and is the most transparent about it. Postmaster Tools shows you the domain and IP reputation it has assigned, your spam complaint rate and your authentication results, and the number that governs everything is the complaint rate: keep it under 0.10% and never let it approach 0.30%, the level at which Gmail's tolerance ends. A Gmail warmup lives or dies on opens and replies, so the engaged-first ordering matters most here. Microsoft, covering Outlook, Hotmail and Live, is historically the hardest mailbox to warm. It throttles unfamiliar senders aggressively through SmartScreen, weights IP reputation heavily, and will defer or reject a ramp that climbs faster than it likes, returning the rejections that bulk senders learned to recognise after its 2025 enforcement tightened. Its feedback comes through SNDS and its junk-mail reporting programme rather than a dashboard as clear as Google's, so warming Microsoft means watching deferrals closely and climbing patiently. Yahoo, together with AOL, mirrors Gmail's authentication and one-click-unsubscribe requirements but maintains its own complaint threshold scored independently, so a clean Gmail reputation does not guarantee a clean Yahoo one.
There is a measurement trap that touches all of them: Apple's Mail Privacy Protection pre-fetches images for a large share of recipients, which inflates open rates and makes opens an unreliable signal on their own. A warmup that judges engagement by opens alone can be reading noise, which is why replies, clicks and the absence of complaints carry more weight than a raw open rate. The practical conclusion is that one curve cannot suit all of these at once: ramped together, you either crawl at the pace Microsoft tolerates or trip it trying to keep Gmail interested. Splitting the ramp by provider, which a self-hosted stack makes straightforward, lets each climb at the rate it will accept.
Cold outreach, bulk mail, and the honest limits of warmup tools.
Warmup means different things for different kinds of sending, and the tools that promise to automate it are useful in a narrower band than their marketing suggests.
For bulk marketing or transactional mail from a dedicated IP, warmup is the gradual, engaged ramp described here, and there is no tool that replaces it; the reputation comes from real recipients wanting real mail. For cold outreach sent from many individual mailboxes, a different pattern is common: automated warmup tools connect each mailbox to a network of other inboxes that open, reply to and rescue each other's messages, manufacturing engagement signals before real campaigns start. These tools have a place for low-volume cold senders, but two honest caveats apply. The mailbox providers have grown markedly better at recognising artificial inbox networks, so the lift these tools give has been shrinking, and over-reliance on them can itself become a risk signal. And for cold outreach the binding constraint is rarely warmup at all; it is volume per mailbox, where the safe ceiling sits around thirty to fifty sends a day per inbox, warmup messages included, which means scale comes from more mailboxes rather than more mail per mailbox.
There is also a European dimension that the tool-centric guides skip. Warmup presumes you are sending to people who want your mail, and in the EU you cannot lawfully send marketing to recipients who never consented. A warmup pointed at a non-consented list is both a legal exposure and a deliverability dead end, since it generates the bounces and complaints that bury a sender, a point that connects warmup directly to the consent and list-hygiene discipline we cover in the bulk sender requirements. The tool can automate the sending; it cannot manufacture consent, and it cannot turn an indifferent list into an engaged one.
After the ramp: maintenance and decay.
Reaching full volume is not the finish line. A reputation that is not maintained quietly erodes, and the senders who get caught out are the ones who treat warmup as a one-time event.
Sender reputation decays when sending stops. A domain or IP that goes quiet for several weeks loses some of the standing it built, so resuming at full volume after a long pause can look, to the provider, like a dormant sender suddenly waking up at scale, which is the pattern of a compromised or sold address. Senders with seasonal or bursty patterns, the retailer who mails heavily in November and barely in February, run into this most, and the fix is to keep a baseline of engaged mail flowing through the quiet periods rather than letting the channel go cold. Where a long gap is unavoidable, a shortened re-ramp on the way back up is safer than assuming the old reputation is still intact.
Maintenance is also a matter of holding the signals that earned the reputation in the first place. Lists decay as addresses go stale, so ongoing verification and the removal of recipients who stopped engaging keep the bounce and complaint rates that the providers watch from drifting upward over time. A sender that ramps cleanly and then lets its list rot will see deliverability slide months later for reasons that look mysterious but trace straight back to neglected hygiene. The discipline that completes a warmup is the same discipline that sustains it: consistent volume, clean lists, and engagement kept healthy rather than assumed.
From the side of people who run the infrastructure.
Most warmup advice is written from the buyer's seat. This is the view from the sending stack, where the ramp is something you build and throttle rather than something a dashboard hides.
On a self-hosted sending stack, warmup stops being a fixed schedule and becomes a set of controls. A system like PowerMTA lets you throttle per receiving domain and per virtual IP, so you can hold Gmail at one rate, Microsoft at another and Yahoo at a third, matching each provider's tolerance rather than averaging across them and tripping the strictest one. You can stage the ramp by virtual MTA, keep transactional and marketing streams on separate IPs with separate curves, and watch the bounce and deferral feedback directly in your logs instead of waiting for an ESP to summarise it days later. That control is the advantage of owning the infrastructure; the cost is that the monitoring and the tuning are yours, which is precisely the work most senders underestimate.
That is the lens we bring, because we run sending infrastructure rather than only advise on it, and warmup is one of the moments where running it and understanding it are the same skill. A good warmup is unremarkable in hindsight: authentication that passes, a list that consented, a ramp paced to the providers' feedback, and engagement organised so the right recipients vouch for the sender first. Where it helps to have that built and watched by people who do it for a living, that is the work our email infrastructure and deliverability practice exists to do, and for agencies and resellers who need it under their own brand, the same discipline runs through our white-label email infrastructure.
Questions senders are asking in 2026.
Do I really need to warm up my IP?
How long does email warmup take in 2026?
What volume should I start with, and how fast can I increase it?
What is the difference between domain warmup and IP warmup?
What bounce and complaint rates are safe during warmup?
Can I warm up several IPs at once by splitting my volume?
Do automated warmup tools work?
Should marketing and transactional mail share a warmup?
What do I do when an ISP starts deferring my mail mid-warmup?
Can I warm up to a purchased or scraped list?
Why does engagement matter so much now?
Do I need to warm up again after a pause?
What has to be in place before I start warming up?
Does shared infrastructure mean I never think about warmup?
How do I monitor a warmup while it runs?
Can self-hosted senders warm up the same way as ESP customers?
Do Gmail, Microsoft and Yahoo expect different things during warmup?
How do I move to a new IP or ESP without re-warming from zero?
A warmup is only as good as the setup underneath it.
We get the authentication right, separate your streams, build the ramp around your most engaged recipients, and throttle each provider on its own curve, then watch the bounce and reputation signals so a problem is caught on day one rather than week three. Whether it runs on a dedicated IP or your own PowerMTA stack, the warmup is the easy part once the foundation is sound.