Cold Email Warmup in 2026, An Operator Playbook

Cold email warmup is the process of gradually ramping a new mailbox's sending volume while simulating real conversations, so Gmail and Outlook learn to trust the sender before any prospect ever sees a message. Done right, it takes four to six weeks per mailbox and pushes inbox placement past eighty percent.

What cold email warmup actually does, and what it does not

A warmed mailbox is a mailbox that mailbox providers have already classified as a normal correspondent. Gmail and Outlook do not store a single global score for your domain. They store a behavior profile per sending address, built from how the address engages with other addresses, how often it gets replies, how often it gets opened, how often it gets marked as spam, and how steady its volume looks over time.

Warmup is the work of producing that profile before a single cold message goes out. The tool you choose, automated or manual, does not change what the providers are watching for. It changes how cheaply you can produce the signals. The mistake newer operators make is treating warmup as a checkbox the platform handles. It is the gate to inbox placement, and the math behind it is what cold email deliverability actually depends on.

What warmup does not do is fix a sloppy list, a generic message, or a domain that already carries a spam complaint history. A warmup tool cannot wash a damaged reputation. It can only build a clean one on top of clean infrastructure. If your last campaign earned a 1.2 percent spam rate from a recycled domain, no thirty day curve fixes it. You retire the domain, register a fresh one, and start warmup on day one again.

Mailbox count math for a target send volume

Most operators pick a sender tool and let the default tell them how many mailboxes to buy. That math runs the wrong direction. Decide your weekly send target first, then work backwards.

The ceiling per healthy mailbox in 2026 is roughly twenty to thirty cold sends a day, not forty or fifty. The reason is the Google and Yahoo bulk sender rules that took effect in February 2024. Senders above five thousand messages a day to Gmail addresses must authenticate with SPF, DKIM, and DMARC, offer one click unsubscribe, and hold a spam complaint rate under 0.3 percent, ideally below 0.1 percent. Even below the bulk threshold, the major providers throttle a mailbox that consistently sends in volume to addresses that never engage. Twenty five a day per mailbox is the operator ceiling that keeps you well inside the safe band.

Work it out. A team that wants five hundred cold sends a day at twenty five per mailbox needs twenty mailboxes. A team that wants two thousand a day needs eighty. At an Instantly Hypergrowth seat around 97 dollars a month, plus four to seven dollars per Google Workspace seat per month per Instantly's pricing page, eighty mailboxes is a real line item. The right move is to question the send target before the mailbox count. Most pipelines do not need two thousand sends a day. They need eighty well targeted ones.

The thirty day ramp curve, by week

A clean ramp moves three to five sends a day in week one to twenty five to thirty a day in week four, across every mailbox. Each week roughly doubles or adds about forty percent on top of the prior week. Anything faster trips filters that watch for unnatural ramp shapes. InboxKit's protocol data puts full warmup inbox placement at 88 percent against 54 percent for skipped or abbreviated warmup, a thirty four point gap that decides whether the campaign was worth running at all.

Week 1, three to five warmup messages a day. The tool sends to a private network of mailboxes that open, reply, and move messages out of spam. No cold sends, none.

Week 2, ten to fifteen a day. Still warmup only. Start watching your provider's reputation dashboard for the first daily signals.

Week 3, twenty a day. You can begin a handful of true cold sends per mailbox in the second half of the week if reply and bounce metrics look clean. Five real sends per mailbox per day at most.

Week 4, twenty five to thirty a day total. About half is warmup, half is real cold. By the end of the week the ratio shifts to ten warmup and twenty real, and you hold there. For category context, the 2026 warmup tools landscape breaks down the credible options and what each one optimizes for.

The single fastest way to break this curve is to import a five thousand prospect list and queue all of it on day fifteen. Providers do not see ambition. They see a step function in your sending pattern that did not exist on day fourteen, and the throttle applies before lunch.

Provider choice, Google Workspace vs Outlook vs custom SMTP

Google Workspace is the default for a reason. The deliverability ecosystem is built around Gmail's signal model. Filters are aggressive but predictable. Mailbox cost runs around seven dollars a month per seat for Business Starter, or about four dollars on certain partner resellers, which is the price most cold email infrastructure stacks plan against.

Microsoft 365 mailboxes route through Outlook, and Outlook filters work on a different model that punishes new senders harder and rewards them slower. The advantage is reach. Many enterprise buyers are on Outlook, and a campaign sent from a Workspace domain to an Outlook recipient passes through both filter stacks. Operators running large enterprise outbound run mixed pools, Workspace for volume and 365 for matching the recipient side. A primer on the broader plumbing sits in cold email infrastructure.

Custom SMTP services like Mailgun, SendGrid, and Postmark, plus the various reseller "private" warmup networks, are not the right pick for cold outbound in 2026. They produce strong transactional reputation and weak relational reputation, and Gmail learned to distinguish years ago. Use them for product email and receipts. Use real Workspace or 365 mailboxes for cold.

Domain and subdomain strategy for warmup

Never warm up your primary domain for cold outbound. The first principle of survivable cold email is that the sending domain is disposable, and the brand domain is not. The pattern is to register two to four lookalike domains (yourbrand.co, getyourbrand.com, tryyourbrand.io) and run mailboxes off those, never off yourbrand.com.

A typical setup is four lookalike root domains, three mailboxes per root domain, twelve mailboxes total, with each mailbox warmed independently. If one root domain trips a spam rate threshold, you pause it and the other three keep running. Burning a lookalike domain costs the registration fee plus four weeks of warmup. Burning yourbrand.com costs every transactional email your product sends. The economics are not even close.

Inside the lookalike domain, the same rule that governs outbound prospecting holds. Match the sending name to a real person whose photo is on LinkedIn, set the reply to address to a monitored inbox, and never send cold from a generic name like sales or hello.

SPF, DKIM, DMARC, and the DNS gate

Before any warmup message goes out, the DNS records have to be clean. SPF tells receiving servers which IPs are allowed to send for your domain. DKIM signs the message with a private key so the receiver can verify the sender. DMARC tells receivers what to do with messages that fail the first two and reports back on what is happening at scale.

The Google bulk sender rules made all three required for any sender above five thousand daily Gmail messages, and the threshold for the rest of the world is now effectively zero. A mailbox without DMARC published gets quietly down ranked. The minimum policy that ships is v=DMARC1; p=none; rua=mailto:dmarc@yourdomain.com. The p=none is a starting position. Move to p=quarantine once you have read two weeks of DMARC reports and confirmed no legitimate sender is being misclassified.

DKIM keys should be 2048 bit. Workspace and 365 both default to 1024, which still passes but gets weaker every quarter as the major providers tighten. Rotate to 2048 in the admin panel before warmup begins, not after.

Postmaster Tools and Microsoft SNDS, the only signals that matter

Warmup tool dashboards are useful, but they show what the tool itself observed. The signals that decide your campaign live at the provider. Google's Postmaster Tools and Microsoft's Smart Network Data Services are the two free dashboards that show how Gmail and Outlook actually score your sending domain and IP.

Postmaster Tools surfaces spam rate by user, domain reputation, IP reputation, encryption rate, authentication rate, and delivery errors. A healthy Workspace domain reads low or no spam rate, high domain reputation, and 100 percent authentication on SPF, DKIM, and DMARC. Anything below "medium" reputation in week three is a sign the warmup curve was too steep or the message content tripped filters.

SNDS is harder to read but worth the effort. It reports the spam complaint rate Microsoft observed from the sending IP, broken down by day. Outlook's filter is opaque, and SNDS is the closest look you get.

The discipline most operators skip is checking both dashboards twice a week, every week, for the first three months. The data is free. The teams that read it consistently outperform the teams that buy more warmup credits. A clean dashboard plus a clean message stack is what scales into the rest of a sales engagement platform without burning the infrastructure underneath it.

Manual vs automated warmup, and what to actually trust

Manual warmup is sending a handful of real personalized messages a day from the mailbox to real people who reply. It works. It also does not scale past about three mailboxes per operator. Useful as a refresher for a single executive's mailbox. Not useful as the foundation of a cold outbound stack.

Automated warmup is what every credible sender tool now ships. The mechanism is a private network of mailboxes that the tool controls, which open, reply, and move your warmup messages out of spam on a randomized schedule. Lemlist bundles Lemwarm into its base Email plan at 55 dollars a month on the annual rate per Lemlist pricing. Instantly bundles warmup across every paid plan. Most stand alone warmup tools sit between three and five dollars per mailbox per month on top of the sender platform.

The trust test for any automated warmup is whether the receiving network is real Workspace or 365 mailboxes, or simulated traffic from headless browsers. Real mailbox networks produce signals that look exactly like real correspondence to a filter. Headless networks can be detected, and the filter response to "we caught you faking engagement" is harsh. Ask the vendor before you wire it in. If they cannot answer, do not buy it.

When to stop warming

The third year of cold email tooling produced one quiet finding the older guides miss. Permanent warmup damages deliverability after a mailbox is established. After roughly six to eight weeks at full ramp, the warmup pattern becomes a tell. Provider filters expect a normal mailbox to have variable, human shaped traffic, not a constant background flow of identical engagement.

The operator move is to keep automated warmup at maintenance volume after week six, around five to eight messages a day per mailbox, instead of running it at week four levels forever. Then taper to zero on mailboxes that are sending real cold volume above twenty a day and getting positive engagement on those sends. A mailbox earning a five percent reply rate from real prospects no longer needs synthetic engagement, and the synthetic engagement starts to make its pattern look engineered. This is also the slot where the Lemlist campaign skill takes over, so real targeted sends replace the warmup curve and the operator runs the whole motion from one prompt.

Common warmup mistakes

A short list of the failures that show up in almost every audit.

• Ramping too fast in week two, doubling instead of adding forty percent.
• Buying mailboxes faster than warmup capacity. Twenty new mailboxes is twenty four week clocks.
• Warming up on the brand domain. The damage is uncapped.
• Skipping DMARC because "Google did not block us yet." It will.
• Treating the warmup tool dashboard as the truth, while Postmaster Tools shows medium reputation and a 0.4 percent spam rate.
• Running warmup forever and producing a synthetic engagement signature.
• Sending cold sequences from a generic alias like sales@ or hello@. Both filters and humans flag it.

The thirty day plan to run this month

Run this plan once and you have a working sending stack. Run it once a quarter for every new domain you add and you have an outbound program that does not collapse the first time a list turns sour. This is the foundation under any modern outbound motion, the layer the AI SDR tools you bolt on top all assume is already true.

Days 1 to 3. Register two to four lookalike domains. Set up Google Workspace or Microsoft 365 mailboxes, three per domain. Publish SPF, DKIM at 2048 bit, and DMARC at p=none with a reporting address. Connect each mailbox to your sender platform and start warmup at three to five messages a day.

Days 4 to 7. Hold week one volume. Verify Postmaster Tools is reporting authenticated traffic. Set up SNDS for the sending IP. Read both dashboards on day seven and write down the numbers.

Days 8 to 14. Ramp to ten to fifteen warmup messages a day. Draft the first real cold sequence offline. No real sends in this window.

Days 15 to 21. Ramp warmup to twenty a day. Late in the week, send five real cold messages per mailbox per day to a tight list of fewer than two hundred prospects. Track reply, bounce, and spam complaint daily.

Days 22 to 30. Ramp warmup to twenty five a day. Cold volume grows to ten to twenty per mailbox per day. By day thirty you are running half warmup, half real cold, and Postmaster Tools should read high domain reputation.

Day 30 onward. Drop warmup to maintenance, five to eight a day per mailbox. Cold volume holds at twenty to twenty five per mailbox. Move DMARC to p=quarantine once two weeks of clean reports are in.