Skip to content
Thirdlane

Contact Center Analytics

Contact Center is the queue and agent half of Organization Analytics. It answers how long callers waited, which queues miss their target, what agents did with their shift, and where the work actually arrives during the week.

Open it from Reports > Organization Analytics, then choose Contact Center from the category dropdown. This is the contact center reporting surface; Grafana is not used for queue or agent reports.

Typical questions

  • Which queues miss their service level, and by how much?
  • Are we losing callers to waiting, or to having nobody logged in?
  • Who is logged in but not taking calls?
  • How much of the shift goes to pauses, and for what reasons?
  • When do we actually need people on the floor?
  • Are callers coming back because their problem was not solved?

How the page is arranged

Sections are grouped in the left-hand list. Queues covers what happened to callers, Agents covers what happened to staff, Performance covers speed and demand, Channels covers the non-voice mix, and Interactions browses individual records.

Every section respects the date range and queue filters at the top, with two deliberate exceptions noted where they occur.

Where queues begin

This page owns the call from the moment it enters a queue. Anything before that - an auto-attendant menu, or a ring group that cascades into a queue - is reported under Call Analytics in the Routing group.

That boundary matters when a hunt group feeds a queue. The hunt group report covers the ringing that happened before the handoff and this page covers the wait afterwards, so the same call legitimately appears in both with different numbers. Neither is wrong; they measure different stretches of the same call.

Queues

By queue compares entered, answered and abandoned across queues, with answer rate and average wait per queue and a daily trend. The twelve busiest queues are charted individually and the rest are combined as Other, so a long tail of low-volume queues does not make the chart unreadable.

SLA reports service level as the share of answered queue calls picked up within the threshold, against the goal - the classic 80/20 formulation, meaning 80% of calls answered within 20 seconds. Both the threshold and the goal cascade from the queue, to the organization, to the platform default, so a queue that sets neither still gets scored.

The same page also reports inclusive service level, which scores the identical calls against everyone who called rather than only those who were answered. The difference between the two numbers is the callers who gave up. Standard service level can look healthy while a substantial share of callers hang up, because abandoned calls simply are not counted; inclusive service level is the number that notices them.

SLA Breaches lists the individual calls that waited past their queue threshold, whether they were eventually answered or given up on. A service level percentage tells you that you missed; this tells you who waited, on which queue, and by how much.

Callbacks covers callers who accepted an offer to be called back instead of waiting. Rejected means the number failed validation rather than that anyone declined. Called back means the system dialled the caller and put them back into the queue - so a callback that reaches a caller who has since gone out still consumes an agent.

Agents

Agents and Utilization are the volume view: interactions per day, how many agents were on queue, conversation time, and the hourly relationship between load and staffing.

Occupancy is the one to understand properly. It is talk time as a share of the time an agent was logged in and not paused - so it measures how busy someone was while available, not how much of their day was spent on the phone. Time logged into several queues at once counts once rather than being multiplied. Sessions that began before the period, or had not ended when it closed, are clamped to the period, which means occupancy for today mid-shift will read differently once the shift ends.

The Where Staffed Time Went chart splits each agent’s staffed time into talking, paused, and waiting for a call, for the twenty agents with the most staffed time. Reading occupancy next to that split is more useful than either alone: a low occupancy with high waiting means overstaffing for the demand, while a low occupancy with high pause time is a different conversation entirely.

Login Sessions lists every login and logout with its duration, plus staffed hours per day and how many agents contributed. A blank logout is not missing data - it means the agent was still logged in when the period ended.

Ring No Answer counts calls the queue offered to an agent that the agent did not pick up. This is the section that catches a problem no other report will: an agent who is logged in and never answers looks entirely fine in every count-based report, because they simply do not appear in them. Here they appear at the top.

Pause Reasons shows how much of the shift went to pauses and why. A pause carries a code only when the agent used a pause feature code or was paused with a reason supplied; every other pause is counted under No code given. A large “No code given” share usually means agents are pausing by a route that cannot carry a reason, not that they are hiding anything - check how they are actually pausing before reading intent into it.

Dispositions reports the outcome codes agents assigned, as a mix, a trend, and a detail table.

Performance

Wait Time is a histogram of how long callers waited, mixing answered and abandoned calls together, plus the trend over the period.

Answer Speed deliberately excludes abandoned calls. Because Wait Time mixes both, a queue where many callers give up early can show a flattering average wait; Answer Speed shows only how quickly the calls that got through were picked up. The bars are calls per wait band and the line is the running share answered by that point, so you can read off what proportion were answered within any given number of seconds.

Abandoned covers the callers who gave up: per day, how long they waited first, which queues they were in, and a recent list with click-through to the call.

Both Abandoned and SLA Breaches offer a short abandon threshold. Setting it excludes abandons shorter than that many seconds, on the basis that a caller who hangs up after three seconds has usually misdialled or changed their mind rather than been failed by the queue. It is off by default and capped at ten minutes. Turning it on makes the numbers kinder, so agree the figure before reporting on it rather than adjusting it afterwards.

Disconnect Causes answers a different question from Abandoned: not how many callers gave up, but how every queue call ended. Each call leaves the queue exactly once, counted as handled, abandoned, timed out, no agents available, or transferred - abandoned means the caller gave up while waiting, timed out means the queue gave up on finding an agent, and no agents available means nobody was logged in to ring. The How Calls Ended chart shows the share each accounts for, and the daily trend shows whether that mix is shifting. A rise in “no agents available” is a staffing or log-in problem rather than a patience problem, which the abandon rate alone would not tell you apart.

Traffic Patterns answers when the work arrives: volume by weekday, a weekday-by-hour heatmap, and the peak number of callers queued at once per hour. The queue-length figure is reconstructed from arrivals and departures rather than sampled, so it is the true peak within each hour rather than whatever a periodic snapshot happened to catch.

FCR is a first-contact-resolution proxy, and worth reading with its method in mind. It assumes a caller who called once was resolved and a caller who called more than once was not. The headline rate compares unique callers against those who called repeatedly anywhere in the period.

Two consequences follow. A repeat caller about an entirely different matter counts against the rate, so the figure is better read as a trend than as an absolute. And the headline rate counts repeats across the whole period while the daily chart counts repeats within each day, so on a multi-day range the daily line will normally sit above the headline figure - a caller who called Monday and Friday is a repeat overall but resolved on both days. Compare like with like: headline against headline, day against day.

Anonymous callers are excluded, since a caller with no number cannot be recognised on a second call.

Channels

Channels shows the voice, chat and SMS mix and how it moves over time. Messaging covers outbound messaging volume and delivery outcomes, broken down by number, status and agent. Recordings and Transcription report volumes rather than content, including how much of what was recorded was actually transcribed - the gap between the two is the usual reason a search for a phrase returns less than expected.

For the messages themselves rather than the counts, use Message Detail Records.

Interactions

Interactions charts closed interactions by channel and outcome. By agent / DID / client groups the same records by whichever dimension you choose from Group by.

The grouped table loads one batch at a time and Load more appends the next, so a total at the bottom of the visible rows is a total of what has loaded, not of the period. For a full record-level browse, use the Contact Center Interactions category instead.

Access

The category is available to administrators whose role includes call reporting. Organization administrators see their own organization only.

Individual sections can be hidden per organization from Limits on the Tenant form, and sections whose underlying feature is off - messaging, recordings, transcription - hide themselves.