B2B email marketing practices that hold up in 2026

Two changes have invalidated a lot of B2B email advice. Apple Mail Privacy Protection prefetches images, so open rate is no longer a measurement of anything — and the Google, Yahoo and Microsoft sender requirements made authentication a gate rather than a best practice. Everything below assumes both.

1. Build the list, never buy it

A purchased B2B list is legal in some jurisdictions and ruinous in all of them. The addresses were not collected for you, they include role accounts and spam traps, and the complaint rate they generate ends the sending domain's reputation. Under GDPR you will also struggle to document a lawful basis: legitimate interest can cover B2B outreach in several European markets, but it requires a documented assessment, a clear relevance to the recipient's role, and an easy objection path — none of which survives "we bought it".

What works instead: content with a real reason to exchange an address, webinars, product trials, inbound demo requests, and events where consent is captured individually rather than by badge scan.

2. Treat list hygiene as a standing process

• Remove hard bounces on the first permanent failure, and enforce the suppression list at send time across every tool that mails as your domain.
• Verify imported and older lists before the first send. Syntax and MX checks are deterministic; SMTP probes are unreliable at Microsoft 365, which commonly accepts at RCPT TO and bounces afterwards, so hold unknown and catch-all verdicts out of bulk sends.
• Target hard bounces under 1% and spam complaints under 0.1%. Google's published threshold is 0.3%, which is a ceiling, not a target.
• B2B lists decay faster than consumer lists because people change jobs. Expect a meaningful share of a two-year-old list to be gone.

3. Segment past firmographics

Industry, headcount and revenue tell you who to talk to, not what to say. Add:

4. Separate your mail streams

Put product and transactional mail, marketing campaigns, and outbound prospecting on different subdomains with different From addresses, and keep the organisational domain for person-to-person mail. Reputation is evaluated per sending domain, so this is also the only way to stop a campaign's complaints from delaying a password reset. Outbound prospecting in particular should never share a domain with anything you cannot afford to lose.

5. Authenticate every sending domain, and verify on delivered mail

Publish one SPF record per domain, within the 10 DNS-querying mechanism and 2 void lookup limits. Publish DKIM at the selector your sender actually signs with. Publish DMARC with rua, read the reports, and move to enforcement. Then confirm it on a received message rather than in a DNS checker — spf=pass, dkim=pass, dmarc=pass, with header.from aligned. Rely on DKIM alignment as primary, because DKIM survives forwarding and SPF does not.

6. Stop reporting open rate as a result

Apple Mail Privacy Protection fetches tracking pixels through a proxy regardless of whether the recipient read anything, which inflates opens for any list containing iPhone and Apple Mail users. Gmail has long pre-fetched images through its own proxy too. The consequences:

• Open rate is unusable as a performance metric and worse than unusable as a trigger for automation. "Opened but did not click" sequences now fire on people who never saw the message.
• Use clicks, replies, meetings booked, and pipeline created. For outbound, reply rate is the only number that reflects a human.
• Keep open data for one purpose only: coarse engagement trends over long windows, understood to be inflated.

7. Write for a reply, not for a brochure

• One idea per email, one ask per email.
• Short. A B2B decision-maker reads on a phone between meetings.
• Plain-text-leaning design for one-to-one outbound; a one-image-and-a-button layout reads as bulk and is invisible when Outlook blocks remote images, which it does by default.
• Specific over general: name the problem in the reader's vocabulary, with a number you can defend.
• A closing line that names one next step. "Looking forward to hearing from you" asks for nothing.

8. Cap frequency and enforce it centrally

A prospect can sit in a nurture sequence, a webinar invitation list, a product announcement list and an SDR sequence simultaneously. Without a global per-contact frequency cap applied across tools, they receive four emails in two days from one company, and that is what produces the complaint. Suppress across systems too: a demo booking must stop the prospecting sequence.

9. Sunset disengaged contacts on a schedule

One re-engagement attempt at six to twelve months of no engagement, then suppression. This feels like losing reach and is the most reliable way to raise placement, because mail to people who never respond teaches providers that your mail is unwanted — and long-dormant addresses are where recycled spam traps live.

10. Measure placement, not delivery

Delivery rate is acceptance by the receiving server; it reads 99% while your campaign sits in spam. Placement is the folder. Measure it per provider, because Gmail, Microsoft and Yahoo disagree constantly and a blended number averages away the only actionable fact. Run it continuously so a decline appears as a trend rather than as a quarter of missing meetings.

What not to do

• Do not buy a list, enrich a bought list, or accept one from an agency.
• Do not run opened-based automation now that opens are synthetic.
• Do not send marketing from the domain your invoices come from.
• Do not jump to DMARC p=reject before your reports show every legitimate source authenticating.
• Do not increase volume by more than roughly 30–50% day over day on a young domain.
• Do not add a sending domain the week before a campaign.
• Do not test placement by mailing colleagues on the same tenant; internal mail skips the filters that matter.

MailPilot covers the infrastructure half of this list: warm-up over OAuth (Gmail, Outlook) or SMTP/IMAP including SendGrid, Amazon SES, Mailgun and Postmark, SPF/DKIM/DMARC verification against delivered mail, daily per-mailbox reputation, and real per-provider inbox placement from a seed-inbox network — so your reply-rate experiments run on mail that is actually arriving.