Cold Email 2026/07 · Sending
Module 07 · Sending

Sending is orchestration, not a button.

Inbox rotation, daily caps, schedule timing, variant split, stop-on-reply. These are the controls that turn good infrastructure and good copy into actual results. Get one wrong and the whole pipeline above leaks.

The one idea

Five controls, all of which matter

ControlCorrect settingWhy
Inbox rotationSpread sends across all inboxesNo single mailbox exceeds its safe cap
Daily cap per inbox30–40 / day (ramp from low)Matches warmup; abrupt volume burns rep
ScheduleRecipient's business hours, local time3am sends tank reply rate
Variant split2–3 copy variants for email 1Tests copy + breaks fleet fingerprint
Stop-on-replyOnA follow-up after a reply kills trust instantly
Schedule

When to actually send

For most B2B audiences in 2026: Tue–Thu, 9am–11am or 2pm–4pm in the recipient's local time. Monday morning is the second-highest reply window. Avoid Friday afternoon (replies plummet) and weekend sending (most filters dock you for it).

2026 update Most sending tools now auto-detect the recipient's timezone from their company HQ or LinkedIn profile and schedule locally. Confirm this is on. A campaign sent to the East Coast at the West Coast's 9am is a campaign sent at 6am — landing in their inbox at the bottom of the morning stack.
Test slices

Always send a 50-lead test before scaling

A test slice is a small portion of your list (50–100 leads) that you send first, then wait 24–48 hours, then check: did emails inbox? Did merge fields render? Did any inbox flag warnings? If anything's off, you fixed it at 50 leads instead of 2,000.
  1. Pick 50 leads from your list at random.
  2. Launch the campaign at 50 only. Watch over 24 hours.
  3. Open 3–5 of the sent emails in your inbox provider's "Sent" view. Are they rendering right? Are merge fields populated correctly?
  4. Check Google Postmaster or mail-tester for any flags.
  5. If everything looks clean, release the rest.
A / B variants

Test one variable at a time

The rule: change one thing. Subject line, or first line, or CTA — never two at once. Otherwise you can't tell which change moved the result. Two variants is plenty; three only if your list is 1,000+.
A clean A/B test Variant A subject: Quick question about your new pricing page
Variant B subject: 11% revenue lift idea for your pricing page
Same body, same CTA, same everything else. Send to a random 50/50 split. Run for 2 weeks. Compare reply rate.
Stop-on-reply

The setting people forget that ruins everything

If a prospect replies "thanks, will look" and your sequence sends them follow-up email 2 three days later, you've turned a positive reply into "this is clearly a bot." Every sending tool has a stop-on-reply toggle. Always on. No exceptions.
Do this now

Configure one campaign end to end

  1. Import your verified list (from Module 02) into your sender.
  2. Map custom fields for any merge variables in your copy.
  3. Attach all your warmed inboxes from Module 04. Confirm rotation is on.
  4. Paste the 4-email sequence from Module 05 with the 3 / 4 / 5-day delays.
  5. Enable stop-on-reply.
  6. Set per-inbox daily cap at 30. Set the campaign-wide cap to match your funnel.
  7. Schedule to recipient's local business hours (Tue–Thu, 9–11am).
  8. Send to 50 leads first. Check 24 hours. Then scale.
Don't do this

The sending mistakes that waste good pipelines

Skip the test slice. One bad merge field renders "Hey {firstName}" in 2,000 inboxes. Permanent reputation damage.
Ramp too fast. Per-inbox cap is a reputation limit, not a suggestion.
Identical copy from every inbox. Use variants and spintax or the fleet looks like one blast.
Send all timezones at once. Schedule locally.
Turn stop-on-reply off "just for this campaign." No.
What this rules out

Now read your numbers

If you've done modules 1–7 cleanly, any remaining problem will show up as a specific row in your funnel deviating from the model. Module 08 walks you through reading those numbers.