Reply classification is the quietly highest-leverage automation in cold email. Done right, it turns 500 inbound replies per month into a clean queue of interested prospects, pauses sequences for objections, and removes unsubscribes instantly — without a human reading every message. This playbook covers the categories that matter and the rules that should fire on each.
The 7 reply categories that matter
- Interested — "Yes, let's talk," "Send me more info," "What are your rates?"
- Objection — "We use [competitor]," "Not right now," "Budget is tight."
- Not the right person — "Try our VP of Sales," "I'm not in charge of this."
- Out of office — auto-reply with return date or vacation note.
- Unsubscribe / do-not-email — "Take me off your list," "Stop emailing me."
- Job change — "I no longer work at [company]," "Please try [new email]."
- Bounce / delivery failure — handled automatically.
Related: Do AI First-Line Generators Work in Cold Email?
What should happen on each
- Interested: pause sequence, Slack-ping the AE, create CRM task, insert TidyCal link in reply template.
- Objection: pause sequence, tag objection type, enrol in nurture drip, notify rep.
- Not the right person: pause sequence, capture suggested contact, re-enrol the suggested contact if ICP match.
- Out of office: pause sequence until return date (parse the date from the auto-reply).
- Unsubscribe: add to global suppression list, never email again, remove from all active and future sequences.
- Job change: update contact record, LinkedIn real-time re-verify, route to new email + new company if known. LinkedIn RTV.
- Bounce: mark address invalid, re-run through waterfall, decide whether to retry in 90 days.
Why natural-language classification beats keyword rules
Old-school keyword rules ("contains the word 'unsubscribe' = unsubscribe") miss 30–40% of true positives. Real replies are messy: "please stop," "take me off," "no thanks," "I'm good," "just don't," "removing this asap." A natural-language classifier captures all of those — and flags ambiguous replies for human review instead of guessing wrong.
Related: How to Plan a Cold Email Sequence That Earns Replies: Five Steps From List to Follow-Up
Confidence thresholds
- 90%+ confidence: Auto-action fires.
- 70–89%: Auto-action fires but reply is also flagged for rep review.
- Below 70%: No auto-action; reply lands in a "needs human" queue.
CRM integration pattern
Every reply should become a native CRM activity — not a synthetic task or a note in a custom field. Happierleads pushes every classified reply to HubSpot / Salesforce / Pipedrive / Zoho / GoHighLevel as a normal email activity with the classification as a custom property. CRM integrations.
Unified inbox vs. scattered mailbox
When you rotate across 10 mailboxes, replies scatter across 10 inboxes. A unified inbox pulls every reply into one view, preserves the classification, and lets reps respond without context-switching. Inbox Engine unified inbox.
Metrics to track
- Classification accuracy (spot-checked).
- Time from reply → rep action.
- Unsubscribe honoring rate (should be 100%).
- OOO pause accuracy (does the sequence actually resume on the right day?).
- Job-change detection rate (cross-check with LinkedIn).
Turn on classification
Start a trial — reply classification is on by default in Happierleads' Sequencer.
