Cold Email Deliverability: The Rules I Follow in 2026

Deliverability is not just an infrastructure problem. If the offer is weak and people mark the email as spam, perfect inboxes will not save the campaign. I start with a relevant list and a message people have a reason to read, then I build the sending system around that.

These are the rules I use after monitoring more than 75,000 inboxes.

Reputation Starts With the Domain and the Inbox

New inboxes sending cold email immediately look suspicious. Warm-up gives the inbox conversation history before it carries campaign volume. For Google Workspace, I treat two weeks of warm-up as essential. For Microsoft, domain age matters even more, although I still prefer to warm the inbox.

That difference matters because Google and Microsoft do not filter the same way. Gmail gives more room for a traditional cold email. Outlook is far more sensitive, so I keep its copy extremely short and plain.

Match the Infrastructure to the Market

For smaller companies, such as ecommerce brands, Google inboxes are usually the practical fit. Larger companies are more likely to use Microsoft, where the filtering is tougher. I segment campaigns by the prospect's email provider so I can see performance differences instead of treating every inbox the same.

I also avoid contacts protected by systems such as Barracuda or Proofpoint where possible. Those tools are built to block the kind of outbound email I am trying to send.

Keep Per-Inbox Volume Conservative

The send limits are not interchangeable. For Google, I use a small number of inboxes per domain and work toward roughly 15 to 20 cold emails per inbox once warmed up. For Microsoft or Azure setups, I use more inboxes per domain but send only two to five cold emails per inbox.

Warm-up volume should stay above projected cold volume. As a working rule, I run roughly 1.3 to 2 times the cold-send volume in warm-up activity. The goal is not to force volume through a mailbox. It is to scale incrementally while the reputation remains healthy.

Write for the Provider Receiving the Email

For Outlook, I use short copy, often 15 to 40 words, and strip out anything that makes the message look like an automated cold email. No elaborate signature, no long greeting, no unnecessary formatting. The message should read like a direct note to a coworker.

For Gmail, I still keep the copy clean, but I do not treat it as tightly constrained. In both cases, I use one-to-three-word plain-text subject lines, avoid emojis, numbers, and dollar signs, use a short CTA, and keep the signature basic.

I also use variation in copy. For Microsoft in particular, unique AI-generated lines can help prevent every message from carrying the same obvious structure. The goal is not fake personalization. It is to avoid making every email look machine-produced.

Diagnose the Right Failure

Do not change the whole system because one campaign slows down. Start with the symptom.

If inbox-health tests look good but replies disappear, the copy or offer may be the problem. A Microsoft campaign with no bounces and no engagement can be shadow-filtered, so test a different structure and angle. If bounces spike or an IP-pool issue appears, replace or investigate the sending inboxes. If Gmail performance drops, reduce volume, improve the offer, and test the copy before assuming the infrastructure is broken.

I monitor bounce rate daily, run weekly deliverability tests, and investigate an inbox-health score below eight out of 10. This connects directly to the wider Cold Email Full Playbook: list quality, offer, copy, and infrastructure are one system. Deliverability breaks when one of those layers is ignored.