Most call centers coach the same way: pull an agent aside after a bad call, review a random recording once a week, or wait until someone's numbers are so bad they're about to get fired. That's not coaching. That's damage control.
VICIdial has every tool you need for real-time coaching — silent monitoring, whisper, barge-in, screen monitoring, performance dashboards. The problem is that most operations never set it up properly and never build a repeatable workflow around it.
Three Modes, Three Purposes
Silent Monitor lets a supervisor listen to a live call without anyone on the call knowing. Uses Asterisk's ChanSpy under the hood. No beep, no audio artifact — the agent's channel isn't modified. Use it for routine quality checks, evaluating new agents, and pre-coaching reconnaissance (listen to an agent's calls before a coaching session to identify specific behaviors to address).
Whisper lets a supervisor talk to the agent during a live call without the caller hearing. The caller only hears the agent. Use it for feeding rebuttals, guiding new agents through their first live calls, and recovering calls that are going sideways. Important: the agent needs to be comfortable receiving instructions while simultaneously talking. Start with short prompts ("ask about the warranty," "slow down") before attempting full rebuttal feeding.
Barge-In puts the supervisor on the call as a three-way conference. All parties hear each other. This is the emergency brake, not the steering wheel. If supervisors are barging into calls regularly, your training program has a problem. Over-using barge-in damages agent confidence and creates a culture where agents lean on supervisors instead of developing skills.
Getting It Working
Supervisors need User Level 7+ with Monitor and Barge-In permissions enabled. Their phone (hardware, softphone, or WebRTC) must be registered in VICIdial under Admin > Phones.
From the real-time report at /vicidial/realtime_report.php, locate the agent, click their row, and select MONITOR, WHISPER, or BARGE. The supervisor's phone rings — answer it to begin. You can escalate mid-session: start with silent monitor, escalate to whisper if the agent needs help.
If agents hear a beep when monitoring starts, the ChanSpy q (quiet) flag is missing from the dialplan. Fix it — that beep immediately defeats the purpose of silent monitoring.
Data First, Monitoring Second
Don't listen to calls randomly. Use the data to identify who needs coaching and what kind.
Conversion rate by agent is the revenue metric. Run a 7-day query grouping by user, filtering to 50+ calls minimum. Include standard deviation of talk time — an agent with high stddev is inconsistent, and coaching should focus on replicating what a good call looks like.
Agent utilization separates real productivity from logged-in time. An agent with 35% pause time when the team average is 15% is either taking too many breaks, slow-rolling between calls, or having system issues. Sort by pause percentage descending to find your biggest coaching opportunities.
Talk time distribution reveals more than averages. Two agents with 180-second averages could have very different distributions — one consistently 170-190, the other bouncing between 60 and 400. The consistent agent follows a process. The other is winging it. Agents with the highest standard deviation in talk time are your primary coaching targets for process consistency.
Disposition pattern analysis catches gaming. An agent with 40% "Not Interested" when the team average is 25% might be giving up too quickly. Unusually high DNC dispositions might mean the agent is using DNC to remove leads they don't want to call. Both are coaching conversations.
The Coaching Session That Works
Keep it to 15-30 minutes and follow this structure:
1. Data review (3 min). Show the agent their metrics compared to team averages. No judgment — just numbers.
2. Call review (10-15 min). Play specific recordings you identified. Pause at key moments. "On this call at 2:14 PM Tuesday, you got 'I need to think about it' and said 'okay, thanks for your time' — let's talk about handling that differently." That's coaching. "Your conversion rate is low" is not.
3. Skill building (5-10 min). Focus on ONE behavior. Not five things. Role-play the scenario until the agent can deliver the new approach naturally.
4. Action items (2 min). One measurable goal for next week. "Increase conversion from 3.2% to 4.0%" is measurable. "Do better on calls" is not.
Follow-Up Is Where Most Programs Fail
Within 24 hours: silent monitor 3-5 calls, check if the behavior change is being applied. If yes, send a positive note. If not, have a brief follow-up.
At 3 days: pull updated metrics, compare to baseline.
At 7 days: full metrics comparison, pre-coaching week vs. post-coaching week.
At 30 days: sustained improvement, partial improvement, or no change. Document it. Use that data to refine your approach.
New Agent Coaching Schedule
Days 1-3 (nesting): supervisor monitors every call via whisper. Active coaching on every call. End-of-day call review. Goal: agent gets through the script without major intervention.
Days 4-7 (supervised independence): silent monitor 50% of calls. Whisper only when the agent is about to lose a call. Goal: handles routine calls independently.
Days 8-14 (spot monitoring): silent monitor 10-20% of calls. No whisper unless requested. Goal: performing within 80% of team average.
If an agent is still significantly below team average after 30 days of intensive coaching, have the honest conversation about fit.
For the full setup including SQL queries for coaching dashboards and Grafana integration, see the complete coaching guide on ViciStack.
Originally published at https://vicistack.com/blog/vicidial-agent-coaching/
Top comments (0)