Monitoring Cold Email Deliverability: Detecting Spam

Most “deliverability” dashboards show you a proxy, not the truth. No mail provider tells the sender “this email went to spam.” That’s exactly why good campaigns often die quietly — deliverability tips over, and nobody notices until replies stop coming. This guide is the monitoring perspective: not how to set up deliverability (that’s covered in the deliverability fundamentals article), but how to recognize that it’s tipping — in time.

The 7 things most teams get wrong

1. You don’t know whether you’re in the inbox. Bounces, opens, and replies don’t measure placement.
2. The warmup score is a vanity metric. 98-100 points while real prospects never reply — the most common false sense of security in cold email.
3. The biggest lever isn’t copy or warmup, it’s ESP matching. Gmail→Gmail, Microsoft→Microsoft.
4. Aggregated metrics hide the problem. A healthy overall bounce rate can mask a Microsoft collapse.
5. Don’t trust your tool’s provider labels. Detect the real ESP from the DNS MX records, not from the label.
6. Microsoft is a different, harder game — lower volume ceilings, stricter filters, slowest recovery.
7. Monitor daily and act automatically on thresholds. A Monday problem becomes a burned domain by Friday if a human has to notice it first.

”Are we in spam?” — harder to answer than you’d think

The uncomfortable truth: bounces, opens, and replies say nothing about placement. A bounce is a rejection at the door, not spam foldering. Opens are unreliable (no-pixel best practice + Apple Mail Privacy). What’s left:

Signal	What it proves	Detects the spam folder?
Bounce rate (by recipient ESP)	Hard rejection at the door	Strongest early indicator — but rejection, not junk
Reply drop-off vs. baseline	People aren’t seeing the email	Best practical inbox proxy for cold email
Warmup-pool spam rate	Placement in the warmup network	❌ Vanity — not your real audience
Seed/placement test	Inbox vs. spam via a seed panel	✅ Direct — but synthetic
Google Postmaster	Real Gmail complaint rate + reputation	✅ Direct — Gmail only
Microsoft SNDS	Outlook reputation	✅ Direct — IP-based, often unavailable on managed mailboxes

Takeaway: The free signals you already have (bounce + reply drop-off, split by ESP) are an excellent early-warning system. If you want to know rather than infer, you need a placement test or Postmaster.

The warmup-score trap

Warmup scores measure placement within the warmup pool. This number can show 98-100 while your real Microsoft prospects see nothing. Never auto-pause or auto-throttle based on the warmup score alone. Log it, alert on it — but treat it as the vanity metric it is. Trust bounce-by-ESP and reply trends instead.

ESP matching only works if both sides exist

“Google→Google, Microsoft→Microsoft” is the most effective deliverability move there is. But it has a prerequisite teams overlook: you need Microsoft sending mailboxes to route Microsoft recipients to. If your fleet is 100% Google, ESP matching does nothing for your Microsoft segment — there’s nothing to match.

The subtle trap after that: how the mailboxes are connected determines whether the matching actually takes effect. Mailboxes connected via native Microsoft OAuth are recognized as Microsoft. Mailboxes bolted on via generic IMAP/SMTP often land in the “other” bucket — you then own Microsoft mailboxes that are never used as Microsoft. Always check the connection type.

Split everything by recipient ESP

Your sequencer’s analytics are aggregated. Aggregation hides exactly the failure mode that kills German/EU B2B cold email: Gmail placement fine, Microsoft placement collapsing. Build the split yourself — map every recipient domain to its real provider (a static map for the common ones, MX lookup for the rest) and calculate bounce/reply per ESP. This one view turns an invisible problem into an obvious one.

And: classify the ESP from the MX records, not from the tool’s label. We’ve found managed mailboxes whose provider codes didn’t match the vendor’s own documentation. DNS doesn’t lie.

The numbers that matter (threshold cheat sheet)

Metric	Target	Warning	Act
Inbox placement	>80%	<80%	<60% → investigate
Bounce rate (campaign)	~1%	>1%	>2% → pause
Bounce rate (domain)	<2%	>2%	>4% → stop sending, rotate to a healing domain
Microsoft segment bounce	<3%	>3%	route to Microsoft senders / pause segment
Spam complaints	<0.1%	>0.05%	>0.1% → pause (Google/Yahoo limit: 0.3%)
Human reply rate	>10%	<5%	drop >50% vs. average → investigate

Sending limits per provider (cold/day per inbox)

Provider	Max cold/day	Note
Google Workspace	~15	Warmup ratio ~1:1.75
Microsoft 365	~10	Strictest filter, slowest recovery
Azure tenant (50 mailboxes/tenant)	~5 max	Exceeding this risks a tenant-wide burn (4-6 weeks recovery)

Scale via more domains/inboxes, never via more volume per box. Rule of thumb: 8 domains × 5 inboxes × ~20/day, not 2 inboxes × 250/day.

The staged action ladder

Automate the safe, gate the destructive:

1. Alert — Slack notice with entity + metric + fix.
2. Throttle — automatically lower the daily limit toward policy.
3. Pause — automatically stop the campaign/mailbox at danger thresholds.
4. Rotate — swap a burned domain for a healing domain — gated by a human.

Run steps 1-3 automatically with a full audit log; require a human for rotation and for resuming anything. Speed where it’s safe — a hand on the wheel where it isn’t. (The same human-in-the-loop principle as in Multi-Channel Outbound.)

The 10-point pre-flight checklist

• SPF: one record, ≤10 lookups, ends in ~all
• DKIM signed and passing on every domain
• DMARC: p=none → quarantine (4-8 weeks) → reject (4+ months)
• One-click unsubscribe (RFC 8058)
• Never send cold from the primary company domain
• 2x-domains rule: half sending, half healing, rotate monthly
• ≤15/day Google, ≤10/day Microsoft, per inbox
• Sender ESP mix roughly matches the lead ESP mix
• <100 words, plain text, no tracking pixels, no links in first contact
• Daily monitoring with automatic thresholds — no human staring at dashboards

Treat any provider’s “95% inbox guarantee” as marketing, not a metric. We break down how many emails one meeting ultimately costs here: Cold Emails per B2B Meeting.

---

CegTec runs cold-email-deliverability monitoring as part of the outbound system — bounce/reply split by ESP, automatic threshold actions with human-in-the-loop for domain rotation. The full threshold config + monitoring blueprint is available as a lead magnet.

Start your free trial · 4 weeks free, no credit card. Prefer to see it running first? Book a demo.