Cold email still works, and it still burns domains. The difference between a team that gets replies for years and one that is in the spam folder by week three is almost entirely operational: how the mailboxes are set up, how fast they ramp, how much they send, and whether anything stops them when things go wrong. This is the checklist we run on.
1. Domain and authentication
- Send from separate domains, not your main one. Use look-alike domains (yourcompany-team.com, getyourcompany.com) so a reputation hit never touches the domain your company's own mail depends on.
- SPF, DKIM and DMARC on every sending domain. Without all three, Google and Microsoft will throttle or reject. Start DMARC at
p=none, watch reports, then tighten. - A real website on each domain that redirects to your main site, with a real contact page. Filters and humans both check.
- Age before volume. A brand-new domain that sends 500 emails on day one is spam by definition.
2. Mailbox warm-up
- Warm every new mailbox before it sends a single cold email. Weeks of low, rising, conversational volume with replies, so the mailbox has a history.
- Ramp per campaign, not just per mailbox. A warmed mailbox still needs a slow start on a new sequence.
- Rest mailboxes after trouble. A mailbox that bounced or was flagged needs time off and a fresh warm-up, not a lower cap.
3. Caps and gaps that are actually enforced
- A daily cap per mailbox. Low double digits for cold mail is normal. Spread across a fleet of mailboxes rather than pushing one hard.
- A minimum gap between sends. Minutes, not seconds, with randomness.
- Per-domain, per-company and per-campaign limits on top of per-mailbox caps, so you never hit the same company with twenty emails on a Tuesday.
- Enforce caps in the system, not in the plan. If a cap can be overridden with a spreadsheet, it will be. Rookbound enforces caps, gaps and ramps in the database so a campaign physically cannot exceed them.
4. Match senders to recipients
Mail travels best between the same providers. Google recipients get Google senders, Microsoft gets Microsoft. Addresses behind security gateways such as Proofpoint, Mimecast or Barracuda need extra-careful sending: lower volume, cleaner content, and fast pausing on any sign of trouble.
5. Verify the list, twice
- Verify on import. Invalid addresses bounce, and bounces are the fastest reputation killer.
- Verify again before sending if the check is older than a few weeks; people change jobs.
- Skip catch-all domains unless you choose to accept the risk, because you cannot tell a real address from a dead one.
6. Content that doesn't look like a blast
- Plain text. No HTML templates, no images, no tracking pixels.
- No links in the first touch. A link in a cold email from a new domain is a spam signal; earn the click in a reply.
- Short sequences. Up to six emails, stopped the moment someone replies.
- One-click unsubscribe on every email, and honour it before the next send.
- Replies, not opens. Open tracking requires a pixel, which hurts deliverability and lies anyway. Make reply rate the headline number and leave open and click tracking off unless you need them.
7. Automatic pausing
The teams that last are the ones whose system stops itself. Pause a mailbox or a domain automatically on:
- a bounce rate rising above a low threshold,
- any spam complaint,
- a blacklist hit,
- inbox placement falling in your seed tests.
Reroute that mailbox's queue to healthy ones, rest it, and warm it back up before it sends again.
8. What happens after the reply
A reply is the whole point, and the moment most systems get wrong. Sort replies (interested, meeting, not interested, out of office, wrong person, unsubscribe), let the team correct a label in one click, and act on the label: stop the sequence, move interested people into a long-running drip from the same mailbox they wrote to, and make sure nobody who has ever responded is blasted again. Keep the CRM honest: sends, replies, bounces and unsubscribes should flow back as activity, and interested replies should become a task for the person who owns the conversation.
The one-page version
| Area | Rule |
|---|---|
| Domains | Separate look-alike domains, SPF + DKIM + DMARC, real site |
| Warm-up | Every new mailbox, then a slow ramp per campaign |
| Caps | Per mailbox, domain, company and campaign; enforced by the system |
| Matching | Google to Google, Microsoft to Microsoft; careful with gateways |
| Lists | Verified on import and again before send; skip catch-alls |
| Content | Plain text, no links first touch, six emails max, one-click unsubscribe |
| Tracking | Replies yes; opens and clicks off by default |
| Safety | Auto-pause on bounces, complaints, blacklists, placement drops |
| After reply | Sort, stop, drip from the same mailbox, sync the CRM |
Quick answers
How many cold emails can one mailbox send per day?
Keep it in the low double digits for cold mail and scale with more mailboxes, not a higher cap. A warmed mailbox with a history tolerates more than a new one.
Should I use open tracking in cold email?
No, by default. The tracking pixel is a spam signal and the data is unreliable. Measure replies.
Why does mail to some companies always land in spam?
They are probably behind a security gateway (Proofpoint, Mimecast, Barracuda). Those need lower volume, plainer content and immediate pausing on any bounce or complaint.
When should I pause a sending domain entirely?
On a blacklist hit, on a complaint, or when placement drops across several mailboxes on that domain at once. Pause, reroute, rest, re-warm.