Customers who waited longest and gave up don't count against your average. A contact center with 30% abandonment at the 5-minute mark will report better AWT than one with 5% abandonment, even if the latter serves customers faster overall.
The formula: AWT = Total Wait Time / Number of Customers Served
Critical decision: Do you include abandoned contacts? Most calculations exclude them, which flatters performance
Calculate two versions: AWT (Answered) and AWT (All Contacts) - the gap tells you how much the metric is flattering you
Segment by cohort: Blended averages hide queue-specific problems
Staffing is core: AWT is fundamentally a supply-demand problem
Last updated: June 2026
Average Wait Time measures the total duration customers spend waiting before being served by an agent or system. It answers: how long do customers sit in queue before someone helps them?
Lorikeet is an AI customer support platform that removes queue pressure entirely by resolving issues instantly, with no hold time and sub-1-second response latency on voice, improving AWT for the contacts that still need a human.
How to Calculate It
The core formula:
AWT = Total Wait Time / Number of Customers Served
Numerator: Cumulative time all customers spent waiting, typically from queue entry until agent connection.
Denominator: Number of customers who actually received service. This is where measurement gets tricky.
Critical decision - abandoned contacts:
Most AWT calculations only count answered contacts. This creates a perverse incentive: customers who waited longest and gave up don't count.
Calculate two versions:
AWT (Answered): Wait time for customers who reached an agent
AWT (All Contacts): Includes abandoned contacts, using their abandon time as their wait time
Data Collection and Measurement
Data sources:
For voice: ACD systems log queue entry and answer timestamps
For chat: Your chat platform records conversation start and first agent response
For email: Less meaningful since email is asynchronous
Segment by: Channel, time of day, customer segment, issue type, and region. A blended average hides problems in specific queues.
Want to reduce wait times with AI that actually resolves issues? See how Lorikeet handles contacts without queueing.
Worked Example
A B2B SaaS company tracks chat AWT for a Tuesday afternoon:
Customer A: 3 min wait, answered
Customer B: 3 min wait, answered
Customer C: 12 min wait, answered
Customer D: 3 min wait, answered
Customer E: 8 min wait, abandoned
AWT (Answered only): (3+3+12+3) / 4 = 5.25 minutes
AWT (All contacts): (3+3+12+3+8) / 5 = 5.8 minutes
The 10% gap is relatively small here. If abandonment were higher with longer waits, the gap would be dramatic.
Common Pitfalls
Excluding abandoned contacts from the calculation. The most common form of unintentional metric gaming.
Fix: Track both answered and all-contacts AWT. Report the gap as a quality signal.
Ignoring IVR and pre-queue time. Your metric says 45 seconds; your customer experienced 3+ minutes.
Fix: Measure total elapsed time from first contact attempt.
Treating blended averages as meaningful. Company-wide AWT might hide that billing runs at 30 seconds while technical runs at 8 minutes.
Fix: Segment by queue, channel, and customer tier.
Optimizing AWT at the expense of resolution quality. Rushing customers off hold to hit targets, only to transfer them multiple times.
Fix: Pair AWT with FCR and CSAT.
Lorikeet's Take
At Lorikeet, we've learned that AWT is a supply-demand problem at its core. Staffing model improvements and contact volume reduction through effective self-service are higher-leverage than incremental process optimizations. An AI concierge that resolves issues end-to-end, rather than deflecting them, takes contacts out of the queue entirely, so the wait time you measure only applies to the cases that genuinely need a human.
The most honest way to measure AWT includes abandoned contacts. The gap between answered-only and all-contacts AWT tells you how much your headline metric is flattering you. We recommend reporting both.
Finally, AWT doesn't tell you whether the wait was worth it. A customer who waited 5 minutes and got their problem solved is happier than one who waited 30 seconds and got transferred three times. Pair AWT with resolution quality metrics.
Key Takeaways
The denominator decision (answered only vs. all contacts) dramatically affects what AWT means.
Cross-platform comparisons are unreliable because wait time starts and stops at different points.
Blended averages hide queue-specific problems. Segment by cohort.
AWT is a supply-demand problem. Staffing and volume reduction are highest-leverage fixes.
Pair with resolution quality metrics. Fast answers that lead to poor outcomes create worse experiences.
