Infrastructure architecture
Cold email infrastructure has five layers: identity (domains and DNS), sending (mailboxes), orchestration (sequencer and suppression), record (CRM) and observability (monitoring and attribution).
Each layer can silently fail. Only the last one tells you when the others have.
Domains
Use dedicated sending domains, separate from the domain your product, billing and support depend on. Spread volume across several so reputation problems stay contained.
Keep the domains recognisably related to your brand — plausible to a human reader, isolated from your primary estate.
DNS authentication
Configure SPF, DKIM and DMARC on every sending domain, then verify alignment with a test send rather than trusting the provider's checklist.
Set DMARC to a monitoring policy first, review reports, then tighten.
Mailboxes and warm-up
Provision mailboxes with full profiles and signatures, then ramp volume gradually. New identities that jump straight to target volume are the most reliable way to land in spam.
- Complete profile and signature
- Gradual ramp schedule
- Conservative daily ceiling
- Scale by adding mailboxes, not by raising caps
Deliverability practice
Verify every list before launch, keep formatting plain, avoid heavy link usage in first touches, and monitor bounce and complaint rates continuously with automatic pauses at defined thresholds.
Sending patterns
Distribute volume across mailboxes, randomise intervals, and send within working hours for the recipient's region. Human-shaped sending patterns matter as much as content.
Sequencing platform
Pick the sequencer that integrates cleanly with your CRM and supports per-mailbox limits, suppression rules and reliable reply detection. Feature lists matter less than accurate reply handling.
CRM integration
Outbound activity should log to the CRM automatically, with segment, persona, trigger and campaign fields on the record. Without them you can't attribute anything later.
Tracking
Measure what changes decisions: replies, sentiment, meetings booked and opportunities created. Consider whether pixel and link tracking is worth its deliverability cost on your domains.
Monitoring
Weekly checks across the estate so problems surface before a quarter is affected.
- Authentication still valid
- Blacklist checks
- Bounce and complaint rates by mailbox
- Reply-rate drift
- Inbox placement tests
Troubleshooting
When performance drops, isolate the layer before changing copy. Check bounces first, then placement by provider, then per-mailbox performance, then data freshness, then messaging.
Most sudden drops are infrastructure or data problems wearing a messaging costume.
Pre-launch checklist
Nothing sends until every line is true:
- Dedicated domains registered and isolated
- SPF, DKIM and DMARC verified per domain
- Mailboxes provisioned, profiled and warmed
- Per-mailbox volume ceilings set
- Lists verified and suppression loaded
- Sequencer connected and logging to CRM
- Monitoring and auto-pause thresholds live
FAQ
Enough that each one sends a conservative daily volume. The right number depends on your target volume, provider limits and warm-up state — calculate it from the ceiling per mailbox, never by raising the ceiling.
Tracking that rewrites links or embeds pixels on a fresh domain can contribute to filtering. Many teams reduce or drop it and rely on reply-based measurement instead. Test rather than assume — provider behaviour changes.
Not sure this is your constraint?
Fix your outbound foundation