How to build a zero-tolerance red-flag list for outbound telesales

Zero-tolerance red flags are the compliance events that should never be averaged away into a “pretty good” score. If the customer demanded Do Not Call and the agent kept selling, the call fails — regardless of how polished the rest of the pitch sounded.

For AI-assisted QA to help, those flags must be defined per campaign, not buried inside a vague soft-skills rubric.

Start with events that force a zero

Build your list from outcomes that already trigger coaching, chargebacks, or compliance escalation today:

1. Do Not Call / stop-contact requests — customer language that should end the pitch
2. Forbidden claims — guarantees, absolute language, or unapproved product promises
3. Disqualifying statements — health, age, residency, or financial facts that should stop the sale under your script
4. Missing critical disclosures — when absence itself is a hard fail (treat carefully; some shops put these on the checklist instead)
5. Profanity or abusive conduct — if your policy treats it as automatic fail

Write each flag in language a reviewer would recognize on a real call — short phrases and clear intent, not legalese essays.

Separate red flags from the QA checklist

A practical pattern used on regulated floors:

• QA checklist — scored criteria (disclosures, qualification questions, process steps). Partial credit may exist.
• Red-flag list — zero-tolerance. Any observed violation forces the call score to zero, independent of checklist performance.

Keeping them separate prevents a strong opening from “saving” a DNC mishandle in an average.

Make flags reviewable, not just detectable

Detection without evidence creates noise and distrust. Every red flag your system raises should ship with:

• The quoted transcript line that triggered it
• Speaker context (agent vs customer — including warm-transfer edge cases)
• A one-click jump into the recording
• Human accept / reject with comment

That last step matters. Models mishear. Diarization mislabels. Your QA lead still owns the call.

Tune with a real batch, not a workshop whiteboard

Draft the list with compliance and campaign owners, then validate on production audio:

1. Process a pilot batch with your draft flags
2. Review false positives (especially mislabeled company script after transfers)
3. Tighten phrasing and add campaign-specific synonyms
4. Re-run before you scale to full dialer volume

Call QA is designed for this loop: configure checklist + red flags per campaign, score the batch, and let reviewers work the exception queue.

Keep ownership with the campaign

Red-flag lists drift when scripts change. Assign an owner (campaign manager + compliance) and revisit when:

• The script or offer changes
• A new client requirement lands
• QA finds repeated false positives or misses

AI can score every call. Only your team can define what “never acceptable” means for that floor.

Need help mapping your first campaign list? Book a Call QA pilot.