Optimizing VICIdial Inbound Queues: Fixes for the 5 Worst Default Settings

Why Your VICIdial Inbound Queue Loses Calls (And How to Fix the 5 Worst Settings)

VICIdial’s inbound engine provides over 80 settings per inbound group but ships with defaults that lack hold music or overflow routes. This skeleton setup frequently leads to callers hanging up after 45 seconds of dead silence and unanswered rings.

Why This Matters

While VICIdial offers powerful features like skill-based routing and weighted ring strategies, the gap between its potential and the default configuration is vast. In a technical reality, running the default ‘No Agent No Queue’ setting as ‘Y’ results in immediate caller disconnects when no agents are logged in, a failure that actively destroys lead conversion pipelines despite the underlying system’s capability.

Key Insights

• The ‘ring_all’ strategy triggers simultaneous SIP INVITEs for every available agent, which can destabilize Asterisk servers if team sizes exceed 10 agents.
• Skill-based routing is achieved via Agent Rank (0-9) within inbound groups, where rank-9 agents are prioritized by the engine over lower-ranked staff.
• Implementing the CALLERID_CALLBACK option allows high-volume queues to maintain caller positions while offering a key-press callback, drastically lowering abandonment rates.
• Blended campaigns require the ‘Available Only Ratio Tally’ set to ‘Y’ to prevent VICIdial from over-dialing outbound leads when inbound spikes occur.
• SLA monitoring can be performed via direct SQL queries on the ‘vicidial_closer_log’ to track the percentage of calls answered within specific thresholds like 30 seconds.

Working Examples

Custom Music on Hold class definition in /etc/asterisk/musiconhold.conf

[sales_hold]
mode=files
directory=/var/lib/asterisk/moh/sales
sort=random

Query to monitor live inbound queue status and longest wait times

SELECT campaign_id as inbound_group,
COUNT(*) as calls_waiting,
MAX(TIMESTAMPDIFF(SECOND, call_time, NOW())) as longest_wait
FROM vicidial_auto_calls
WHERE call_type = 'IN' AND status = 'LIVE'
GROUP BY campaign_id;

SQL for tracking Service Level Agreement (SLA) performance based on a 30-second answer threshold

SELECT campaign_id,
COUNT(*) as total_calls,
SUM(CASE WHEN queue_seconds <= 30 THEN 1 ELSE 0 END) as within_sla,
ROUND(SUM(CASE WHEN queue_seconds <= 30 THEN 1 ELSE 0 END) / COUNT(*) * 100, 1) as sla_pct
FROM vicidial_closer_log
WHERE call_date >= CURDATE()
GROUP BY campaign_id;

Practical Applications

• Use case: Multi-language routing using an IVR to direct callers to ‘SALES_EN’ or ‘SALES_ES’ groups based on user input. Pitfall: Failing to configure overflow from specialized groups back to a general pool, leaving callers stranded if specific language agents are unavailable.
• Use case: Tiered support escalation where calls move from ‘SUP_T1’ to ‘SUP_T2’ after 180 seconds to ensure high-level technical oversight. Pitfall: Neglecting to update holiday routing in ‘Admin > Call Times’, causing calls to ring into dead queues on closed business days.
• Use case: Priority routing by DID where a VIP line is assigned a priority of 99 to jump ahead of general inquiry lines. Pitfall: Setting Inbound Queue Priority to 0 in blended environments, which forces inbound callers to compete equally with outbound leads.

References:

• https://dev.to/gamlin/why-your-vicidial-inbound-queue-loses-calls-and-how-to-fix-the-5-worst-settings-5973