The Cold Email Deliverability Optimization Guide

Deliverability isn't one fix - it's a sequence. Get the order wrong (say, scaling volume before warm-up is done, or fixing copy before the domain itself is trusted) and you'll burn a domain chasing a problem that was never about the copy. This is the order we actually walk clients through, start to finish.

1. Before you send a single email

Everything below assumes these are already true. If any aren't, fix them first - nothing downstream matters if the foundation is broken.

• SPF, DKIM, and DMARC are all configured and actually passing, not just present in DNS
• The sending domain is separate from your primary corporate domain - a burned cold-outreach domain should never be able to take billing or support email down with it
• New domains and mailboxes are warming up on a schedule, not sending cold volume from day one
• Your lead list was verified in the last 30 days
• You have a way to notice if the domain lands on a blacklist, instead of finding out from a client asking why replies stopped

→ Run all five of these at once with our free Domain Health Check - paste a single domain or your whole sending list

2. Sending volume and the warm-up curve

A new domain or mailbox has no reputation yet - good or bad. Mailbox providers treat that as a blank slate to watch closely, not a blank check to trust. The volume curve is how you spend the first few weeks proving you're not a burst spam operation.

• Week 1: 10-20 emails/day per mailbox, heavy on warm-up traffic that gets opened and replied to
• Weeks 2-3: ramp toward 30-50/day, shifting the mix toward real cold volume as engagement holds
• Week 4+: scale toward your target volume in steps, not a single jump - each step should be smaller than the reputation you've already built can absorb

The ratio of warm-up to cold traffic matters more than the total number. A mailbox sending 40 cold emails and 5 warm-up emails a day looks very different to a mailbox provider than one sending 40 warm-up and 5 cold - even though the volume is identical. Watch your reply and bounce rates at each step before moving to the next; a curve is a guideline, not a schedule to follow blind.

3. When to lower volume instead of pushing through

This is the step most teams skip, because it feels like giving up ground. It's usually the fastest way to get it back. Cut volume - on the specific domain or mailbox, not the whole campaign - when you see any of: bounce rate climbing past 3%, spam-complaint rate above roughly 0.1%, or a sudden reply-rate drop with no change in targeting or copy. Those are early warnings, not the crash itself - by the time inbox placement visibly collapses, you're already a week or two into the reputation damage.

Cutting volume by 50% for a week is a minor setback. Rebuilding a domain's reputation after a full collapse is measured in months, not weeks.

4. Copy that doesn't trip filters

Spam filters aren't matching a static blocklist of words anymore, but certain patterns still reliably raise a message's spam score: urgency stacking (“act now,” “limited time”, all in one subject line), unrealistic promises, excessive exclamation marks, and links that don't match their anchor text. None of these alone will bury an email. Stacked together, on a domain that's already borderline, they're often the difference between inbox and spam.

→ Paste your subject line and body into our free Spam Word Checker to catch these before you send

5. Test copy the same way you'd test anything else

Run two variants of a sequence to a comparable, randomly-split segment of your list - not sequentially to different segments, which conflates copy performance with list quality. Hold everything else constant (send time, volume, targeting) and change one variable at a time: subject line, opener, or CTA. A test that changes three things at once tells you something changed, not what to do about it. Give each variant enough volume to be meaningful before calling a winner - a 20-email sample where one gets 3 replies and the other gets 1 is noise, not signal.

6. Clean HTML, plain text, and attachments

• Prefer plain text or very light HTML for cold outreach - heavy HTML templates read as bulk/marketing mail to filters, even when the content is a personal-sounding one-to-one message
• Never attach files to cold outreach - attachments are one of the strongest spam signals available to a filter, regardless of file type
• Link to a hosted document or landing page instead of attaching one
• If you must send HTML, keep the text-to-image ratio high and skip tracking pixels embedded as tiny 1x1 images where you can avoid them

7. Open tracking: the trade-off nobody mentions upfront

Open tracking works by embedding an invisible image that loads when the recipient opens the email - which is exactly the kind of hidden-content pattern spam filters are built to notice. Since Apple Mail Privacy Protection and similar features started pre-fetching images for a large share of inboxes, open rates were already unreliable as a metric. Turning tracking off on a domain that's still building reputation removes one more signal filters can flag, at the cost of a metric that wasn't trustworthy anyway. Keep it on for warm, engaged segments where you need the data; turn it off for a domain in its first month.

8. Ongoing hygiene, not a one-time setup

• Check bounce and reply rates daily during the first month on any domain, weekly after that
• Pause or rewrite a specific campaign the moment its numbers diverge from your other campaigns on the same domain - one bad campaign can drag down every other campaign sharing that domain's reputation
• If inbox placement drops sharply with no obvious cause, stop sending from that domain entirely rather than troubleshooting live - every email sent while the cause is unknown is more potential damage

9. The metrics that actually tell you something

In order of what we'd actually look at first when something's wrong:

• Reply rate - the strongest signal of both list quality and inbox placement combined; almost impossible to fake or misread
• Bounce rate - your earliest warning of a list problem, and one input mailbox providers weigh directly
• Spam-complaint rate - small in absolute numbers but disproportionately damaging; a handful of complaints on a small domain can undo weeks of clean sending
• Open rate - directionally useful for a domain's overall trend, but too distorted by privacy features and tracking-pixel blocking to trust on any single campaign

If you only have time to watch one number daily, watch reply rate. It's the metric hardest for a bad list or a burning domain to fake.

10. Scaling once it's actually working

A domain with healthy reply rates and low complaints for 3-4 consecutive weeks has earned real reputation - the mistake at this stage is scaling that one domain past what its history supports, instead of adding capacity. Add backup domains and mailboxes and run them through the same warm-up curve from section 2, rather than pushing an already-proven domain to volumes it hasn't earned yet. Reputation doesn't transfer between domains - a second domain sending from the same company still starts from zero.

None of these ten steps are exotic. Most of what separates a sender who scales confidently from one capped at 200 emails/day, afraid to push volume, is doing them in this order instead of skipping to the one that feels most urgent that week.