Definitions

What each one actually means

A dedicated IP address carries only your mail. A shared IP pool carries yours alongside other senders'. The differences are mechanical rather than a matter of opinion, and each row below turns into a practical consequence for one side or the other.

Dedicated IPShared IP pool
Who else sends from itNobodyOther customers of the same provider
Whose behavior moves the reputationYours onlyEvery sender in the pool, collectively
Warming required before full volumeYes, weeks of gradually increasing volumeNo, the pool is already warm on your first message
Effect of a gap in sendingReputation is reassessed on recent behavior, so the ramp is partly undoneOther senders keep the volume flowing, so the pool stays warm
Volume needed to maintain itConsistent volume from you, indefinitelyNone from you specifically
Control over sending patternsComplete, including the freedom to get it wrongBounded by the pool operator's rules
PTR hostname in your own domainPossible, one sending domain per IP address. Rarely asked forNo, the PTR names the operator, which is what receivers ask for
Separating marketing from transactionalAnother set of addresses per stream, each warmed and maintained separatelySeparate pools per stream, run by the provider
Typically billed asA monthly add-on, often priced per IP addressIncluded in the platform
Who manages the warmingYou, or your provider on your behalfThe provider, as a condition of running the pool

Why dedicated IPs are usually recommended

The standard advice exists for real reasons, and the best statement of them is not a vendor's. M3AAWG's Sender Best Common Practices, version 4.0 of August 2026, sets out when a dedicated environment is the right provisioning decision. Its list is the case for a dedicated IP, made by a body whose membership includes the mailbox providers doing the filtering.

  • Isolating mail of unknown or below-average quality. A sender or provider may want the mail kept away from everyone else's while its quality is established or its changes tracked.
  • Isolating mail of higher-than-average quality. The same reasoning in reverse: protecting a strong sender from the reputational effects of other people's mail.
  • Control over volume. M3AAWG notes that combining mail from more than one entity "can create irregular traffic volumes", and that "consistency of volume, whether high or low, plays an integral part in the determination of IP reputation and deliverability outcomes".
  • Not being identified with the provider. An entity may want its mail attributed to itself rather than to the ESP, for reputation and authentication reasons, and may require outbound MTA settings that differ from the ones a shared environment implements.
  • Third-party services that require it. Certification, allowlisting or similar programs may name a dedicated environment as a prerequisite.

Two more sit outside that list and are worth adding. Infrastructure isolation is a contractual or regulatory condition in some environments, binding whether or not it improves placement. And a sender large enough to be most of a pool's volume gains little from the pool and a good deal from isolation: cleaner signal, cleaner diagnostics, no need to reason about anyone else.

Reverse DNS, and who actually asks for alignment

An IP address has exactly one PTR record, so a reverse lookup can only ever return one hostname. M3AAWG's definition of a dedicated IP builds that in: it is an address used by one sender, which "usually identifies itself in rDNS as being associated with that brand". Its shared IP definition is the mirror image, an address "usually identified with the ESP that owns the IP rather than one of the brands sending from that IP". The best practice it gives for dedicated environments is that reverse DNS "should be entity-specific rather than ESP-specific". So the capability is real, and only a dedicated address can provide it.

The question is who wants it. Google does not. Its sender guidelines require that "the public IP address of a sending SMTP server must have a corresponding PTR record that resolves to a hostname", and that the hostname resolve back to the same address. That is forward-confirmed reverse DNS, it applies to shared pools and dedicated addresses equally, any competent provider satisfies it for you, and nothing in it asks for the PTR hostname to match your sending domain. The claim that Gmail wants PTR alignment circulates widely. It is not in Google's published requirements.

The provider that has actually published something in this area is Deutsche Telekom, for t-online.de, and what it asks for is narrower than alignment with your From domain. Its sender requirements state that the hostname for the delivering system's IP address must let an affected party easily research and identify the operator's domain and website, with a direct means of contact, citing RFC 1912, forward-confirmed reverse DNS and Article 5 of EU Directive 2000/31/EC. That is a transparency requirement about who runs the mail server. It is not an authentication signal about who the mail is from.

Which is why it bites where the operator and the sender are the same party: an organization running its own MTA. On an established ESP network the operator is the ESP, its hostname identifies it, and the requirement is met at that level. It is met more comfortably still for an accredited one. Omnivery is a certified member of the Certified Senders Alliance, the European accreditation scheme that carries particular weight with German and wider European receivers, so the sending identity is not an unknown quantity a reverse lookup has to establish from scratch.

PTR alignment per sending domain is a genuine capability that only a dedicated IP can give you, and it is one of the least frequently requested things in email.

If it is offered as a headline reason to upgrade, that deserves the same scrutiny as the rest of the upsell. Ask which receiver is asking, and for what.

The rest of it is worth conceding plainly: for a sender with high, steady, daily volume, a dedicated IP is the right choice, and Omnivery provisions them for that case. The question is what happens to everyone else, which is most senders.

A dedicated IP still has neighbors

The first item in the case above is the one everybody buys: on a dedicated address, the reputation is entirely yours. It is also the weakest of them, because an IP address is never assessed as an isolated object. It sits in a /24, in a network, in an autonomous system, and reputation systems look at all of those.

This is not an inference. It is how the widely used blocklists are built:

  • Spamhaus DROP lists netblocks, not addresses. It is described as an advisory "drop all traffic" list of "netblocks that are leased or stolen by professional spam or cyber-crime operations", published in CIDR notation.
  • ASN-DROP goes a level higher again, listing "autonomous system numbers (ASNs) that are hijacked or leased by professional spam or cyber-crime operations". The autonomous system is a unit of reputation in its own right.
  • Spamhaus presents a network owner's space by /24. Its Reputation Portal "lists all your ranges by /24, along with the number of SBL, XBL, CSS, and BCL listings", with a heatmap of the individual addresses inside each one. The unit of attention offered to the party responsible for the space is the range.
  • Listings can widen. Spamhaus CSS lists IPv4 addresses individually, but notes for IPv6 that a large number of spam-emitting addresses across different /64 blocks "could cause listings to extend to larger blocks".
  • Reputation services report the block, not just the address. Cisco Talos shows the owner of the largest IP block an address belongs to, and its score aggregates more than 25 public blocklists alongside its own global data.

M3AAWG's own conditions for bringing your own IP space to a provider point the same way: the minimum range it will accept is a /24, and it requires that "the addresses in the IP address range must have a clean history". Not the address you intend to use. The range.

What that means for the purchase

A dedicated IP is not a plot of land you own outright. It is an address inside somebody else's range, announced by somebody else's autonomous system, sitting next to whatever else that party has provisioned. Buying a dedicated address moves you out of a shared reputation at the IP layer and leaves you in one at every layer above it.

So the question comes back to the same place. You are choosing a network to sit in either way. The choice is not whether to have neighbors; it is whether the party admitting them applies any standard, and whether it controls the space at all. A provider renting capacity from a hyperscaler is offering you a dedicated address inside a range it does not own, announced by an autonomous system it does not run, alongside tenants it did not select and cannot remove. Vetting is not available to it as an option, because there is nothing there for it to vet.

Making every layer resolve to the same entity

Omnivery owns its IP address space and operates its own networks and autonomous systems, on its own physical hardware, with no hyperscaler anywhere in the delivery path. That is not a procurement detail. It is what allows the layers a reputation system actually looks at to line up:

  • The address sends mail from a vetted customer.
  • The /24 around it holds addresses used by other vetted customers, because the same admission standard applies to all of them.
  • The network is operated by the party that set that standard.
  • The autonomous system announcing the route belongs to the same party.

Every reputational signal points at the same clean entity. A receiver widening its assessment from the address to the range, to the network, to the AS finds the same answer at each step rather than a different party at every layer, most of whom have made no commitment about who they let in. That is the version of a dedicated IP's promise that survives contact with how reputation is actually computed - not an address isolated from its neighbors, which is not a thing that exists, but a neighborhood that reads the same at every level of zoom.

There is a second-order version of this worth knowing if you are being sold several dedicated addresses at once. Spamhaus CSS targets snowshoe patterns, meaning volume spread thinly across many addresses with weak sending history. A bank of lightly used dedicated IPs, warmed on a deadline for one campaign a quarter, is not snowshoe spam and does resemble its shape. The e-commerce case below is the one most likely to produce it.

The problem nobody mentions: a dedicated IP has to stay warm

A dedicated IP is sold as a purchase. It behaves like a standing commitment to a sending habit. The argument runs in four steps, and only the first of them is widely discussed.

  1. Warming takes weeks of gradually increasing volume

    A new IP address has no delivery history, no complaint history and no engagement record. Sudden volume from it is indistinguishable from a compromised host, so the mail is throttled, deferred or filtered. The remedy is a slow, consistent ramp - M3AAWG suggests considering six weeks as an average, and pacing it against how receivers respond rather than against a calendar. The IP and domain warming guide covers the mechanics.

  2. Reputation is a rating of current behavior, not a certificate you earn once

    Google defines it in those terms: reputation is "a rating of the quality of domains and IP addresses used to send email, and is determined by the sending behavior of a domain or IP address". Behavior is present tense. Receivers are scoring a rolling record of recent sending, and an address that stops sending stops supplying one.

    Google's own tooling shows where that ends. Postmaster Tools withholds data for a day when "the total number of messages for a given day is too low", which is a plain statement that below a certain volume there is not enough signal to report on. A sender who cannot fill that record is asking receivers to form an opinion from very little.

  3. So the ramp is not a setup cost, it is an ongoing volume floor

    Reaching full volume once does not finish the job. Holding the reputation means continuing to send at a comparable rate, more or less indefinitely. Every quiet period is a partial reset, and the deeper the trough the more of the ramp it undoes.

  4. Most senders are sold one against the wrong number

    The volume threshold at which a dedicated IP gets recommended is almost always measured against peak or projected sending. The number that decides whether the address survives is the baseline: what you send on an ordinary Tuesday in your quietest month. Those two figures can differ by an order of magnitude, and the gap between them is where dedicated IPs fail.

The question is not whether you send enough email to justify a dedicated IP. It is whether you send enough email, consistently enough, to keep one warm between your peaks.

A sender who can answer yes should have a dedicated IP. A sender who cannot is buying an address that will spend most of the year cold.

Why this hits e-commerce hardest

E-commerce sending is spiky by construction. Black Friday, the seasonal sale, a product launch, an end-of-line clearance: the calendar is built around a handful of days that carry a disproportionate share of the year's revenue. Between those days, volume drops to a fraction of the peak.

Apply the standard dedicated-IP advice to that shape and it produces a specific, repeatable failure.

  1. The peak needs more than one IP address

    A single IP can only push so much mail per hour before receivers start deferring it. Getting a large campaign delivered inside the window it was timed for therefore takes several addresses sending in parallel, not one.

  2. The baseline cannot maintain even one of them

    Ordinary weeks do not produce enough volume to keep one dedicated IP warm, let alone spread across several. Divide a modest baseline across a bank of addresses and each individual address looks quieter still.

  3. So the warming gets scheduled around the campaign

    The addresses get ramped in the weeks before the peak, because that is when they are needed. Warming becomes a project with a deadline attached to the most commercially important date in the year.

  4. And the campaign leaves from addresses that are not warm yet

    Freshly ramped IPs meet their largest ever send on the day it matters most. Receivers see an unfamiliar address producing its highest volume to date, which is the exact pattern warming exists to avoid.

Applied to a spiky sending profile, dedicated-IP advice produces the worst deliverability on the most valuable sending days of the year.

That is not an edge case or a hypothetical. It is the normal outcome of the standard recommendation meeting a normal retail calendar.

The same shape turns up wherever sending follows events rather than a steady rate: ticketing, travel around a season, utilities around a billing run, real estate around a launch. E-commerce is simply where the peaks are largest and the gaps between them longest. What that pattern needs from an email platform is covered in transactional email for e-commerce.

And one set of addresses is not enough

Marketing and transactional mail must not share a reputation record. A campaign that draws complaints should not be able to spend the standing your password resets and order confirmations depend on. That much is not controversial, and it is the reason most senders end up running two streams.

Separating them properly takes several signals at once, and the order matters. The primary boundary is the DKIM signing key, because receivers weight a DKIM-signed domain above the IP address a message arrived on. Beneath that sit separate IP pools, separate sending domains with transactional mode enabled only on the transactional one, and different message headers per stream. The warming guide covers why no single one of those signals is sufficient on its own.

M3AAWG names the underlying problem in its case for provisioning a dedicated environment, and it is worth quoting because it is an argument against mixing the streams rather than an argument for any particular product:

"Combining transactional and marketing email or mail from more than one entity can create irregular traffic volumes. Consistency of volume, whether high or low, plays an integral part in the determination of IP reputation and deliverability outcomes."

M3AAWG Sender Best Common Practices, version 4.0. Note that "more than one entity" and "transactional and marketing" sit in the same sentence: a receiver reads mixed streams the way it reads mixed senders.

So the streams have to be provisioned apart. The IP layer is one of the four ways that happens. It is also the layer where a dedicated setup multiplies.

  • Each stream needs its own addresses. Two streams means two sets, not one set used carefully. Addresses cannot be shared between streams without giving up the separation you bought them for.
  • Each set warms separately. Every stream you separate is a reputation record you have to build. Two streams means two warms running in parallel, each with its own ramp and its own monitoring.
  • Your baseline divides before it reaches any address. This is the part that catches senders out. Whatever consistent daily volume you have is split across the streams first, so the number each individual address sees is smaller than the number you used to justify the purchase. A sender who could just about keep one address warm cannot keep two.
  • Peak multiplies too, and the streams cannot lend to each other. A marketing peak needs several addresses to deliver inside its window. The transactional stream still needs its own throughout, and borrowing the marketing addresses for a busy transactional hour would merge exactly the two reputations you separated.
  • The settings differ per stream, so the addresses are not interchangeable. Transactional mode changes suppression behavior and is enabled per sending domain rather than per message. One-click unsubscribe under RFC 8058 belongs on the marketing stream; the transactional stream carries List-Help instead. A domain and address set configured for one stream is not one you can point at the other.
  • The bill multiplies per address. Dedicated IPs are priced per address per month, so the line item is the number of addresses each stream needs multiplied by the number of streams, every month, whether or not the addresses were busy.

Stream separation doubles the warming problem and doubles the invoice, while halving the baseline volume that has to solve it.

The three effects compound in the same direction, which is why senders who separate their streams correctly are often the ones least able to sustain dedicated addresses.

On a vetted shared pool none of that reaches the customer. Omnivery separates the streams for every customer running both, as the way the platform is set up rather than as a tier or an add-on: distinct DKIM signing keys as the primary boundary, distinct IP pools beneath them, separate sending domains with transactional mode enabled per domain by the Omnivery team, and distinct headers per stream. Suppressions and reporting are held per domain, so the two streams stay separate in the data as well as in the reputation. There are no addresses to buy per stream and no second warm to run, because every pool involved was already warm.

How a shared pool solves the volume problem

The mechanism is not complicated, and M3AAWG states it directly. From the same version 4.0 that sets out the case for dedicated environments:

"IP reputation is sensitive to changes in sending volume. Mail volumes from more than one entity can be combined in a shared environment to sustain consistent overall average sending volume from the shared environment to establish and maintain IP reputation."

M3AAWG Sender Best Common Practices, version 4.0, on when provisioning to a shared environment is appropriate.

That is the whole argument in two sentences, and it is worth noticing which way round it runs. The shared environment is not a discount version of a dedicated one that you tolerate until you can afford better. It is the provisioning decision that solves a problem a dedicated environment has and a shared one does not.

What it buys, in practice:

  • Aggregate volume keeps the addresses warm. Many senders' mail combines into one continuous flow, so the pool's addresses always have recent behavior for receivers to score. No individual sender has to generate maintenance volume, because everybody's ordinary sending is somebody else's.
  • Your peak lands on capacity that is already trusted. A campaign leaves from addresses with an established record instead of from addresses that were ramped in a hurry for the occasion. The peak is absorbed rather than announced.
  • Reputation is set by collective behavior, which is far steadier than yours. A single spiky sender's pattern is noisy. The sum of many senders' patterns is not, and a stable pattern is the easier one for a receiver to trust.
  • There is no ramp before you can operate. Sending starts at normal volume on the first day, because the addresses were warm before you arrived.
  • A single mistake is diluted rather than amplified. M3AAWG again: "By provisioning a number of entities within a shared environment, mistakes made by any single sender in the environment can dilute the overall reputational impact." On a dedicated address, one bad send is the whole record.

Note what the pool does not do. It warms the IP side and nothing else. Your sending domain is new to every receiver regardless of which address it leaves from, and its record still has to be built - M3AAWG recommends domain warming even for a new subdomain of an organizational domain that already has a reputation. That is the most common misreading of "the provider handles warming", and it is the subject of the IP and domain warming guide.

The catch: a shared pool is only as good as its worst sender

All of that depends on the other tenants, and the risk that comes with them is real. M3AAWG puts it more bluntly than a vendor would:

"ESPs that send large volumes of email on behalf of their clients are at the mercy of their worst clients' worst practices."

M3AAWG Sender Best Common Practices, version 4.0, opening the section on vetting.

One bad sender damages inbox placement for everyone sharing those addresses. A complaint spike, a purchased list, a spam trap hit, a compromised account: any of them degrades the pool's standing, and every other sender in it carries the consequence without having done anything and often without ever learning the cause. Diagnosing it from the inside is close to impossible, because the symptom is your mail performing worse for reasons that are not in your data. M3AAWG's own word for what one entity's poor sending or list-building does to a shared environment is "poison".

This is precisely why the standard advice says to get a dedicated IP, and as a response to this risk the advice is reasonable. It is solving the problem in the wrong place. Moving to a dedicated address removes you from the bad neighbors; it does not remove the bad neighbors, and it charges you the entire warming problem described above for the privilege of leaving.

The alternative is to fix the pool. That is only possible if the pool operator controls who gets in, which most of them have decided not to do.

Which is why who is in the pool is the whole question

A shared pool is a bet on your neighbors. Most providers ask you to place that bet blind, on a pool anybody can enter with an email address and a credit card, or with neither. The controls that decide whether the bet is a reasonable one are the entry controls, and they exist before anyone sends a message.

This is not an Omnivery opinion about how a pool should be run. It is what the standards body requires of anyone running one:

"All ESPs must have some type of pre-send vetting process to proactively identify malicious senders before they mail and they must all have a post-send vetting process to monitor clients after they mail."

M3AAWG Sender Best Common Practices, version 4.0, section 5. It also recommends "extra diligence in vetting entities to be provisioned in a shared environment", because of the greater chance that one entity's practices poison it for the others.

Both halves of that are load-bearing, and the second is the one usually missing. Vetting at signup catches the sender who was always going to be a problem. Vetting after the fact catches the sender whose list quality decays, whose account is taken over, or whose business model changes. Omnivery runs both halves, and neither is automated end to end:

  • No free tier, by design. Free plans are the front door for anonymous senders, and an anonymous sender is one with nothing to lose from a suspension. Removing the free tier is an anti-abuse measure before it is a pricing decision.
  • Every customer vetted before contract. Reviewed by staff, normally within one to three working days. There is no path into the pool that skips a person. This is the pre-send half.
  • Every sending domain reviewed by staff. Not only the account. The domain that will be signing the mail is looked at before it sends any.
  • Ongoing monitoring by deliverability analysts. People who watch the pool's behavior and contact senders about a developing problem, rather than waiting for reputation damage to show up in someone else's inbox placement. This is the post-send half, and it is done by people rather than only by thresholds.
  • Every customer signs with its own DKIM key. M3AAWG recommends that mail from each entity in a shared environment be authenticated with DKIM under a unique domain or subdomain, so receivers can tell the individual senders apart when making automated decisions, so each can run its own DMARC, and so a sender's earned reputation travels with it. A shared pool where everyone signs as the provider gives receivers nothing to separate.

Vetting also decides who is turned away, and the answer is more specific than "spammers". Cold prospecting does not pass. Neither does synthetic warming, the practice of manufacturing opens, clicks and replies between networks of controlled accounts, which the same M3AAWG document calls artificial warming and says is "not recommended and may be illegal", which its Position on Cold Email calls "not acceptable in any manner", and which now attracts blocklists that attach to the sending domain and the brand on it. The warming guide covers why.

Vetting, the absence of a free tier and staff review of every domain make it substantially less likely that a damaging sender is in the pool at all, and monitoring is what catches the ones whose sending deteriorates after they are admitted. None of that is a guarantee that no sender in a pool will ever cause a problem, and any provider claiming otherwise is describing a pool they do not operate. What it changes is the odds, and the odds are the entire substance of the shared-versus-dedicated decision.

Why the standard advice persists

A dedicated IP is a billable line item. Across the industry it is sold as a monthly add-on, frequently priced per address: Mailgun lists additional dedicated IPs at $59 per IP per month, with one included from its higher-volume plans upward. That structure is normal, and variations on it are near-universal.

It also creates an incentive problem worth naming. A provider recommending a dedicated IP is recommending a product it profits from, at a volume threshold the provider itself sets. The recommendation may well be correct. But the party giving it is not disinterested, and the threshold behind it is rarely tested against the number that actually decides the outcome, which is whether the sender can keep the address warm between peaks rather than what the sender does at peak.

None of that requires anyone to be acting in bad faith. The incentive is structural, which means it operates on honest advice as readily as on dishonest advice, and it is a reason for a sender to ask a specific question rather than to distrust the answer: does this recommendation account for my baseline volume, or only my peak?

Omnivery offers dedicated IPs. We recommend them when the deliverability case supports it, and we tell senders not to buy them when it does not. A dedicated IP that cannot be kept warm is worse than no dedicated IP at all, and selling one anyway is not deliverability advice, it is a revenue decision.

That position has a cost attached. It means declining revenue on the basis of a sender's baseline volume, which is a figure that has to be asked for rather than one that appears on a pricing page.

When you should still choose a dedicated IP

A dedicated IP is genuinely the right answer in five situations, and Omnivery provisions one where any of them holds.

  • Volume is high and consistent, per stream. Enough to warm the address and then hold it warm through ordinary sending, not only through the peaks - and enough to do that separately for marketing and for transactional, since the two cannot share addresses. This is the case a dedicated IP was designed for.
  • A receiver has actually asked you for PTR alignment. Rare, and worth pinning down before acting on it. An IP address has one PTR record, so a hostname in your own domain means a dedicated address per sending domain - but Google does not ask for this, and Deutsche Telekom's published requirement is about identifying the mail server's operator, which an established ESP already satisfies. If a specific receiver has named this to you, it is a real reason. If a provider raised it unprompted, ask who is asking.
  • A regulatory or contractual requirement mandates infrastructure isolation. Whether or not it improves placement, a term in a contract is a term in a contract.
  • Sending patterns are unusual enough that pool norms are a poor fit. M3AAWG notes that entities in a shared environment should have similar content and metrics wherever possible, and that an entity may require outbound MTA settings a shared environment does not implement. A profile that sits far outside the pool's shape is better off with its own address.
  • A third party requires it. Certification, allowlisting and similar programs sometimes name a dedicated environment as a prerequisite. That is a condition to satisfy rather than a deliverability argument to evaluate.

Version 4.0 of the M3AAWG BCP also documents a third path that is neither a shared pool nor a provider-owned dedicated address: bringing IP space you already own onto a provider's platform. It avoids warming new addresses, because the addresses are not new, and it keeps the reputation you have already built in space you continue to control. M3AAWG sets conditions - verified ownership by RDAP or a signed letter of authorization, a minimum IPv4 range of a /24, a clean history for every address in it, reverse DNS pointing at the owner's domain rather than the provider's, and sending from the old provider stopped before the routing announcement completes - and notes that it "is generally not widely offered at the current time" because of the setup it demands on the provider side. Omnivery supports it, in both its EU and US data centers. Worth knowing the option exists before assuming the choice is binary, particularly on a migration where the addresses you already have are the asset you are trying not to lose.

The test to apply

The useful question is not what volume tier you have reached. Providers publish thresholds and those thresholds are chosen by the party selling the address. The question is about the shape of your sending rather than its size:

Between your peaks, in your quietest month, can you generate enough consistent daily volume to keep the address warm on its own?

If yes, a dedicated IP will do what it promises. If no, it will spend most of the year cold, and a cold dedicated address performs worse than a well-run shared pool rather than better.

Answering that honestly usually needs your own numbers rather than a rule of thumb, which is what a deliverability analyst is for. Omnivery's white-glove deliverability team works the question from your actual sending history and per-receiver placement, and will tell you not to buy a dedicated IP when the numbers do not support one.

At a glance

Dedicated vs shared IP at a glance

  • A dedicated IP is a sending IP address used by a single sender, and a shared IP pool is a set of sending IP addresses used by many senders at once.
  • A dedicated IP address must be warmed with gradually increasing volume before it can carry full sending volume.
  • Google defines sender reputation as a rating determined by the sending behavior of a domain or IP address, which makes it an assessment of recent behavior rather than a permanent status.
  • A dedicated IP address requires ongoing consistent volume to stay warm, so warming is a continuing commitment rather than a one-time setup cost.
  • Google Postmaster Tools withholds a day's data when the message volume for that day is too low, which shows that a low-volume sender can leave a receiver with too little signal to assess.
  • Most senders are advised to move to a dedicated IP address on the basis of peak volume rather than baseline volume.
  • E-commerce sending is typically spiky, with a small number of high-volume peak days and much lower volume between them.
  • A peak campaign may require several dedicated IP addresses to deliver inside its window, because one IP address can only send so much mail per hour before receivers defer it.
  • Baseline volume for a spiky sender is often insufficient to keep even one dedicated IP address warm, let alone several.
  • Applying dedicated-IP advice to a spiky sending profile can produce the worst deliverability on the most valuable sending days of the year.
  • A shared IP pool stays warm through the aggregate volume of many senders, so no individual sender has to generate maintenance volume.
  • A shared IP pool absorbs a sender's peak using capacity that is already warm and already trusted.
  • A sender joining a shared IP pool can operate at normal volume immediately, because no IP warming period applies.
  • A single poor send is diluted by the collective history of a shared IP pool, where on a dedicated IP address it is the entire record.
  • An IP address is not assessed in isolation: it sits within a /24, a network and an autonomous system, and reputation systems operate at those levels as well as at the level of the individual address.
  • A dedicated IP address is announced from a provider's network and autonomous system, so its reputation remains exposed to the reputation of the space around it.
  • Spamhaus DROP is an advisory list of netblocks leased or stolen by spam or cyber-crime operations, published in CIDR notation rather than as individual addresses.
  • Spamhaus ASN-DROP lists autonomous system numbers hijacked or leased by spam or cyber-crime operations, which makes the autonomous system a unit of reputation in its own right.
  • The Spamhaus Reputation Portal presents a network owner's address space by /24 with listing counts for each range, so the range rather than the single address is the unit of management offered.
  • Spamhaus notes that a large number of spam-emitting IPv6 addresses across different /64 blocks could cause listings to extend to larger blocks.
  • Cisco Talos reports the owner of the largest IP block an address belongs to, and aggregates more than 25 public blocklists alongside its own global data when scoring an address.
  • M3AAWG sets the minimum range for bringing your own IP space to a provider at a /24 and requires that every address in the range have a clean history, which treats the range rather than the address as the unit of reputation.
  • Omnivery owns its IP address space and operates its own networks and autonomous systems, so the sending address, the surrounding range, the network and the announcing autonomous system all belong to the same vetted entity.
  • A provider that rents capacity from a hyperscaler offers a dedicated address inside a range it does not own, announced by an autonomous system it does not operate, alongside tenants it did not select and cannot remove.
  • Spamhaus CSS targets snowshoe patterns, meaning volume spread thinly across many addresses with weak sending history, which is the shape a bank of lightly used dedicated IP addresses can take.
  • M3AAWG defines a dedicated IP environment as one in which a single entity has exclusive use of and responsibility for the outbound mail through it, and a shared IP environment as one to which more than one distinct entity is assigned.
  • M3AAWG Sender Best Common Practices version 4.0, updated August 2026, states that mail volumes from more than one entity can be combined in a shared environment to sustain consistent overall average sending volume and so establish and maintain IP reputation.
  • M3AAWG states that provisioning a number of entities within a shared environment can dilute the overall reputational impact of a mistake made by any single sender in it.
  • M3AAWG states that combining transactional and marketing email can create irregular traffic volumes, and that consistency of volume plays an integral part in determining IP reputation and deliverability outcomes.
  • M3AAWG states that ESPs sending large volumes on behalf of clients are at the mercy of their worst clients' worst practices.
  • M3AAWG requires every ESP to operate both a pre-send vetting process to identify malicious senders before they mail and a post-send vetting process to monitor clients after they mail.
  • M3AAWG recommends extra diligence when vetting entities to be provisioned in a shared environment, because one entity's poor sending or list-building practices can poison the environment for the others.
  • M3AAWG recommends that mail from each entity in a shared environment be authenticated using DKIM under a unique domain or subdomain, so receivers can differentiate the senders and each entity can run its own DMARC.
  • M3AAWG recommends that reverse DNS for a dedicated IP address be entity-specific rather than ESP-specific.
  • M3AAWG suggests considering six weeks of warm-up as an average, and states that artificial warming practices are not recommended and may be illegal.
  • Bringing IP address space you already own onto a provider's platform is a documented third option that avoids warming new addresses, subject to verified ownership, a minimum IPv4 range of a /24, a clean history and reverse DNS pointing at the owner's domain, and it is not widely offered.
  • The main risk of a shared IP pool is that one sender's poor practice damages inbox placement for every other sender in the pool.
  • A shared IP pool is only viable if the senders admitted to it are controlled, which makes entry criteria the substance of the shared-versus-dedicated decision.
  • Marketing and transactional email should build separate reputation records, so a campaign that draws complaints cannot damage delivery of password resets and order confirmations.
  • Stream separation works across several signals at once, with the DKIM signing key as the primary boundary because receivers weight a DKIM-signed domain above the IP address a message arrived on.
  • Separating marketing from transactional on dedicated IP addresses requires a separate set of addresses per stream, because addresses shared between streams give up the separation they were bought for.
  • Every separated stream is a separate reputation record to warm, so two streams on dedicated IP addresses means two warms running in parallel.
  • A sender's baseline volume is divided across streams before it reaches any individual IP address, so stream separation makes the warmth test harder to pass per address rather than easier.
  • Marketing and transactional streams cannot lend IP addresses to each other at peak, because using the same addresses for both merges the reputations the separation exists to keep apart.
  • Transactional mode changes suppression behavior and is enabled per sending domain, and one-click unsubscribe under RFC 8058 belongs on the marketing stream while transactional mail carries List-Help, so address and domain sets configured for one stream are not interchangeable with the other.
  • Dedicated IP cost multiplies by the number of streams, because the addresses each stream requires are billed per address per month whether or not they were busy.
  • Omnivery separates every customer's marketing and transactional streams using distinct DKIM signing keys, distinct IP pools, separate sending domains with transactional mode enabled per domain, and distinct message headers.
  • Suppressions and reporting on Omnivery are held per sending domain, so separated streams stay separate in the data as well as in the reputation.
  • Dedicated IP addresses are commonly sold as a monthly add-on priced per address, which gives a provider a financial incentive to recommend them.
  • Mailgun lists additional dedicated IP addresses at $59 per IP per month, with one included from its higher-volume plans upward.
  • A dedicated IP address that cannot be kept warm through consistent sending performs worse than a well-maintained shared pool rather than better.
  • Omnivery operates no free tier, vets every customer before contract, and reviews every sending domain by staff, specifically to control who is in its shared pools.
  • Vetting, the absence of a free tier and staff review of every sending domain reduce the likelihood that a damaging sender is in a shared pool, and do not amount to a guarantee that none ever will be.
  • Omnivery offers dedicated IP addresses and recommends them only where the deliverability case supports it.
  • A dedicated IP address remains the right choice for a sender with high and consistent volume, or where a receiver has specifically asked for PTR alignment, or where infrastructure isolation is contractually or legally required.
  • Google requires every sending SMTP server's public IP address to have a PTR record that resolves to a hostname, and that hostname to resolve back to the same address, for shared and dedicated addresses alike.
  • An IP address has exactly one PTR record, so a PTR hostname in the sender's own domain requires a dedicated IP address for each sending domain.
  • Google does not require or recommend that a PTR hostname align with the sender's domain; its published requirement is forward-confirmed reverse DNS on the sending IP address, which applies to shared pools and dedicated addresses alike.
  • Deutsche Telekom is the mailbox provider that publishes a requirement in this area, and it asks that the hostname for a delivering system's IP address let an affected party identify the operator's domain and website with a direct means of contact, citing RFC 1912, forward-confirmed reverse DNS and Article 5 of EU Directive 2000/31/EC.
  • The Deutsche Telekom requirement identifies the operator of the mail server rather than the sender of the message, so it is satisfied at the ESP level and principally affects organizations running their own MTA, where operator and sender are the same party.
  • Omnivery is a certified member of the Certified Senders Alliance, the European sender accreditation scheme that carries particular weight with German and wider European receivers.
  • PTR alignment per sending domain is one of the least frequently requested capabilities in email, so a provider presenting it as a headline reason to buy a dedicated IP address deserves the question of which receiver is asking.
  • Omnivery supports bringing your own IP address space onto its platform, in both its EU and US data centers, which M3AAWG notes is generally not widely offered.
  • A shared IP pool warms the IP side only, and a sender's domain is new to every receiver regardless of which IP address the mail leaves from.
  • The test for whether a dedicated IP address is appropriate is whether the sender can generate enough consistent daily volume between peaks to keep it warm.

Questions

Dedicated and shared IP questions

What is the difference between a dedicated IP and a shared IP?

A dedicated IP is a sending IP address used by one sender only, so its reputation reflects that sender's behavior alone. A shared IP pool is a set of addresses used by many senders at once, so its reputation reflects the collective behavior of everyone in it. The practical difference is not privacy but volume: a dedicated address has to be warmed and kept warm by its single sender, while a shared pool is kept warm by everybody's ordinary sending combined.

Do I need a dedicated IP?

Probably not, and that is usually the right outcome rather than a compromise. The test is not what volume tier you have reached, it is whether you can generate enough consistent daily volume between your peaks, in your quietest month, to keep the address warm on its own. If you can, a dedicated IP will do what it promises. If you cannot, it will spend most of the year cold, and a cold dedicated address performs worse than a well-run shared pool.

At what volume should I get a dedicated IP?

There is no honest single number, and any threshold you are quoted was chosen by the party selling the address. Volume thresholds also measure the wrong thing. What decides whether a dedicated IP works is the shape of your sending rather than its size: a sender doing large peaks with quiet months between them can exceed any published threshold on annual volume and still be unable to keep an address warm. Ask about your baseline, not your peak.

What happens to a dedicated IP if I stop sending for a while?

Its reputation is reassessed on the basis of recent behavior, and there is no longer any recent behavior to assess. Google describes reputation as a rating determined by the sending behavior of a domain or IP address, which is a present-tense judgment rather than a status you earn once. A quiet period undoes a meaningful part of the ramp, and the longer and deeper the gap the more of it is lost. Resuming at full volume after a gap is the same pattern warming exists to avoid.

Is a dedicated IP really isolated from other senders?

Not fully, and this is the part the "your reputation is entirely your own" pitch leaves out. An IP address is never assessed on its own: it sits inside a /24, a network and an autonomous system, and reputation systems work at those levels too. Spamhaus DROP lists netblocks rather than addresses, ASN-DROP lists whole autonomous system numbers, and the Spamhaus Reputation Portal presents a network owner's space by /24 with listing counts per range. Cisco Talos reports the owner of the largest block an address belongs to. A dedicated address is announced from your provider's network, so you take on that network's standing whether or not the address itself is clean. The practical consequence is that you are choosing a neighborhood either way, and the useful question is who else the operator lets into the space.

Can one bad sender ruin a shared IP pool?

Yes. A complaint spike, a purchased list, a spam trap hit or a compromised account degrades inbox placement for every other sender using those addresses, through no fault of theirs and often with no visible cause. This is the genuine risk of a shared pool and the reason the standard advice recommends moving away from one. The alternative response is to control who is admitted to the pool in the first place, which is what vetting is for.

What does M3AAWG say about dedicated vs shared IPs?

Its Sender Best Common Practices, version 4.0 of August 2026, treats the choice as a genuine provisioning decision rather than a hierarchy, and sets out when each is appropriate. For shared environments it states that mail volumes from more than one entity can be combined "to sustain consistent overall average sending volume from the shared environment to establish and maintain IP reputation", and that provisioning several entities together means "mistakes made by any single sender in the environment can dilute the overall reputational impact". For dedicated environments it lists isolating mail of unknown or unusually high quality, controlling volume, not being identified with the provider, needing different outbound MTA settings, and third-party programs that require one. It also requires every ESP to run both pre-send and post-send vetting, and recommends extra diligence when vetting senders for a shared environment.

Is a shared IP pool safe?

That depends entirely on who else is in it, which makes the entry criteria the question rather than the architecture. A pool that anyone can join with a free account carries real and unmanageable risk. A pool where every customer is vetted before contract, every sending domain is reviewed by staff, there is no free tier, and analysts monitor behavior on an ongoing basis carries substantially less. No provider can promise that a sender in a pool will never cause a problem, and one that does is describing a pool it does not operate.

How many dedicated IPs do I need for a large campaign?

More than one, which is exactly where the trouble starts. A single IP address can only push so much mail per hour before receivers begin deferring it, so a large campaign that has to land inside a window needs several addresses sending in parallel. The number depends on the size of the send, the window and the recipient provider mix. The harder problem is not provisioning them, it is that a baseline which struggles to keep one address warm will not keep several warm.

Why do e-commerce senders struggle with dedicated IPs?

Because e-commerce sending is spiky and dedicated IPs need consistency. The peak days need several addresses to deliver quickly, while the ordinary weeks between them do not produce enough volume to maintain even one. The addresses end up being warmed on a deadline set by the campaign, so the largest send of the year leaves from addresses that were ramped in a hurry. The result is the worst deliverability on the most valuable sending days, which is the opposite of what buying the addresses was meant to achieve.

Do I need separate dedicated IPs for marketing and transactional email?

If you are on dedicated addresses, yes, and this is where the cost of the decision becomes clear. Marketing and transactional should build separate reputation records so a campaign drawing complaints cannot damage delivery of password resets, and addresses shared between the streams give up exactly that separation. So two streams means two sets of addresses, two warms running in parallel, and a baseline that divides across the streams before it reaches any single address. The bill multiplies the same way, since dedicated IPs are priced per address per month. Note also that the separation is not only about addresses: the primary boundary is the DKIM signing key, because receivers weight a DKIM-signed domain above the IP a message arrived on.

How does Omnivery separate marketing and transactional sending?

Across four signals, in order of how much reputation each carries: distinct DKIM signing keys as the primary boundary, distinct IP pools beneath them, separate sending domains with transactional mode enabled only on the transactional one, and different message headers per stream. Suppressions and reporting are held per sending domain, so the streams stay separate in the data as well as in the reputation. This is how the platform is set up for customers running both streams rather than a tier or an add-on, so there are no addresses to buy per stream and no second warm to run.

Does Omnivery offer dedicated IPs?

Yes. Omnivery provisions dedicated IP addresses where the deliverability case supports one: high consistent volume, a specific receiver asking for PTR alignment, a contractual or regulatory isolation requirement, or a sending pattern that fits the pool badly. Omnivery also supports bringing your own IP address space onto the platform, in both its EU and US data centers, which is the better route where you already hold addresses with a history worth keeping. It is not the default recommendation, and a sender whose numbers do not support one is told so. The default is the vetted shared neighborhood, which requires no warming management from the customer.

Is a dedicated IP worth paying for?

It is worth paying for when you can keep it warm, and a waste of money when you cannot. Dedicated addresses are sold as a monthly add-on, commonly priced per address, so the cost is recurring while the benefit depends on a condition that many buyers never meet. Before paying, work out your baseline daily volume in your quietest month and ask whether it is enough to sustain the address on its own.

Why do email providers recommend dedicated IPs?

Partly because the advice is genuinely correct for high-volume, steady senders, and partly because a dedicated IP is a billable line item. Mailgun, for example, lists additional dedicated IPs at $59 per IP per month. That does not make any particular recommendation wrong, but the incentive is structural and the volume threshold is set by the party selling the address. The question worth asking a provider is whether its recommendation accounts for your baseline volume or only your peak.

Do I need to warm up a shared IP?

Not the IP side. A well-maintained shared pool is already warm and you inherit that from the first message, which is why sending can start at normal volume immediately. The domain side is different and is still yours: your sending domain is new to every receiver regardless of which IP address the mail leaves from, and its reputation record has to be built. Confusing the two is the most common misreading of "the provider handles warming".

Does a dedicated IP improve deliverability on its own?

No. A dedicated IP address changes who is responsible for the reputation, not what the reputation is. On day one it has no history at all, which is worse than a good shared pool and better than a bad one. What improves deliverability is consistent volume, clean list practice, correct authentication and low complaint rates, and a dedicated address raises the cost of getting any of those wrong because there is nobody else's good behavior to dilute it.

What is PTR domain alignment and does it require a dedicated IP?

PTR domain is the hostname obtained by a reverse DNS lookup on the sending IP address. Alignment means that hostname sits in the sender's own domain rather than the provider's, and yes, it requires a dedicated IP address per sending domain, because an address has exactly one PTR record.

The more useful question is who asks for it, and the answer is almost nobody. Google does not: its requirement is that every sending IP have a PTR record resolving to a hostname that resolves back to the same address. That is forward-confirmed reverse DNS, it applies to shared pools too, and it says nothing about matching your sending domain. The provider that does publish a requirement here is Deutsche Telekom, and what it asks is that the hostname for a delivering system let an affected party identify the operator's domain and website with a direct means of contact, citing RFC 1912, FCrDNS and Article 5 of EU Directive 2000/31/EC. That identifies who runs the mail server, not who the mail is from, so an established ESP satisfies it with its own hostname. It principally affects organizations running their own MTA, where the operator and the sender are the same party.

Can I bring my own IP addresses to a provider?

Sometimes, and it is the option most often left out of the dedicated-versus-shared conversation. M3AAWG documents it in version 4.0 of its Sender Best Common Practices: rather than using addresses the provider owns, a sender can bring IP space it owns onto the provider's platform. The appeal is that the addresses are not new, so there is no warming to do and no reputation to rebuild. The conditions it sets are ownership verified by RDAP or a signed letter of authorization, a minimum IPv4 range of a /24, a clean history for every address in the range, reverse DNS pointing at the owner's domain rather than the provider's, and sending from the previous provider stopped before the routing announcement completes. M3AAWG notes the approach "is generally not widely offered at the current time" because of the work it requires on the provider side. Omnivery supports it, in both its EU and US data centers.

Sources

Every external claim on this page traces to one of these. Where a statement describes Omnivery's own platform or policy rather than a published standard, it is linked to the documentation instead.

  • Google, Email sender guidelines. The source of the PTR requirement quoted above: the public IP address of a sending SMTP server must have a corresponding PTR record that resolves to a hostname, and that hostname must resolve back to the same address. Applies to all senders, on shared and dedicated addresses alike.
  • Google, Postmaster Tools dashboards. The source of the definition used on this page: reputation is a rating of the quality of domains and IP addresses used to send email, determined by the sending behavior of a domain or IP address.
  • Google, Set up Postmaster Tools. The source of the statement that data may be missing for a day when the total number of messages for that day is too low.
  • M3AAWG Sender Best Common Practices, version 4.0, updated August 2026, replacing versions 2.0 and 3.0. The single largest source on this page. It supplies the definitions of dedicated and shared IP environments, the list of situations in which a dedicated environment is appropriate, the statement that mail volumes from more than one entity can be combined in a shared environment to sustain consistent average sending volume, the statement that provisioning several entities together dilutes the reputational impact of any one of them, the observation that combining transactional and marketing mail creates irregular traffic volumes, the recommendation that each entity in a shared environment be DKIM-authenticated under a unique domain or subdomain, the reverse DNS best practice for dedicated environments, the requirement that every ESP operate both pre-send and post-send vetting, the "poison" characterization of one entity's practices in a shared environment, the "bring your own IP" conditions, and the statement that artificial warming practices are not recommended and may be illegal. Linked through m3aawg.org/senderbcp, which is the reference URL the document names for itself and which redirects to the current version. M3AAWG is the Messaging, Malware and Mobile Anti-Abuse Working Group, whose membership includes the major mailbox providers.
  • M3AAWG, Position on Cold Email, published November 2025. The source of the quoted position that artificially simulating subscriber engagement is "not acceptable in any manner".
  • Validity, Deceptive Reputation Warming through Synthetic Messaging. The Heatwave blocklist's own description of synthetic warming, how it is detected, and the fact that a listing attaches to the sending domain and the brand on it.
  • Spamhaus, Don't Route Or Peer (DROP). The source of the statement that DROP lists netblocks rather than individual addresses, published in CIDR notation, and that ASN-DROP lists autonomous system numbers hijacked or leased by spam and cyber-crime operations.
  • Spamhaus, the Reputation Portal. The source of the statement that a network owner's space is presented by /24 with listing counts per range, and that the heatmap shows the individual addresses inside each /24.
  • Spamhaus, Combined Spam Sources (CSS). The source of the snowshoe characterization and of the note that a large number of spam-emitting IPv6 addresses across different /64 blocks could cause listings to extend to larger blocks. CSS lists IPv4 individually at /32; the widening behavior quoted here is the documented IPv6 case.
  • Cisco Talos, sender IP reputation and the Talos Reputation Center overview. The source of the statement that Talos reports the owner of the largest IP block an address belongs to, and that its score aggregates more than 25 public blocklists alongside its own global data.
  • Mailgun pricing. The source of the $59 per IP per month figure for additional dedicated IP addresses, and of the fact that one is included from its higher-volume plans upward. Checked September 2026; vendor pricing changes without notice.
  • Deutsche Telekom, sender requirements for t-online.de. The source of the only published mailbox-provider requirement in this area: the hostname for the delivering system's IP address must let an affected party easily research and identify the operator's domain and website with a direct means of contact, citing RFC 1912, forward-confirmed reverse DNS and Article 5 of EU Directive 2000/31/EC. Quoted here in translation from the German.
  • Certified Senders Alliance, the European sender accreditation scheme, of which Omnivery is a certified member.
  • Omnivery, IP and domain warming. The companion guide, covering the warming mechanics this page's argument rests on and the automated rate limits that replace a manual schedule.

Where this page says a warming ramp takes weeks, M3AAWG's own guidance is to "consider 6 weeks of warm-up as an average" - an average rather than a schedule, and one that depends heavily on list size, list quality and recipient provider mix. The number of IP addresses a large campaign requires depends on send size, delivery window and recipient mix, and no general figure is offered here for that reason. The claim that Gmail treats PTR domain as an authenticity signal for unauthenticated messages appears in several places including Omnivery's own documentation; it is not in Google's published sender guidelines and is not repeated on this page.

Related guides

This decision settles which addresses your mail leaves from. These cover what happens to the reputation afterward, and what decides whether a pool is worth being in.

  • IP and domain warming - the mechanism this whole argument rests on, including why domain reputation now matters more than IP reputation and why a shared pool does not warm your domain for you.
  • Double opt-in, and how bots break it - list quality is what a pool's collective reputation is made of, and an unverified list is the most common way a sender becomes the neighbor everybody else is worried about.
  • The List-Unsubscribe header and RFC 8058 - complaint rate is the signal that moves a pool's standing fastest, and a working one-click unsubscribe is what keeps a complaint from being registered as spam.
  • White-glove deliverability - the analyst work behind the recommendation, including the per-receiver placement review that answers the dedicated-or-shared question from your own numbers.
  • Transactional email for e-commerce - what a peaky retail calendar needs from a sending platform, which is the profile this page's central example is drawn from.