Zendesk Voice Analytics Report

Voice is the channel most support dashboards forget. Ticket-based metrics like first response time and backlog describe asynchronous work well, but they say almost nothing useful about a queue where the customer is on the line and hangs up if you are slow.

A voice analytics report fixes that gap. It measures the parts of a call that decide whether the contact succeeds — how long the customer waited, whether they abandoned, how long the agent talked, and how much time went into wrap-up afterwards. This guide covers how to build the report in Zendesk Explore, the call-versus-leg trap that quietly breaks most first attempts, and how to read voice numbers next to the rest of your queue.

What this report should answer

A useful voice report answers:

How long do customers wait before an agent picks up?
What share of calls are abandoned before they are answered?
Where does call time go — IVR, queue, talk, hold, consultation, or wrap-up?
Which hours and groups create the worst wait, and does it match staffing?
When voice fails, where does that demand reappear?

For definitions, link to the glossary instead of restating them on the dashboard: average speed of answer, call abandon rate, handle time, queue wait time, and channel mix. Keep the report next to the support metrics dashboard so voice is reviewed as part of total demand rather than a separate silo.

Understand calls and call legs before you build anything

Zendesk models a call as one call made up of one or more call legs. A leg is a single participant’s interaction with the call.

End-user legs begin after the call is answered and the caller hears the welcome message, and end when either side hangs up. A callback request creates a second end-user leg.
Agent legs begin when an available agent’s phone or browser starts ringing, and end when the agent declines, transfers, or completes the call — including any configured wrap-up time.

A transferred call with a consultation can easily contain four legs. That matters because of the single most common Explore mistake on this dataset: do not mix call-level and leg-level metrics or attributes in the same report. If you do, call-level metrics are multiplied by the number of legs in the call, and your totals silently inflate on exactly the complicated calls you most wanted to study.

Build two reports instead — one at call level for customer experience, one at leg level for agent behaviour — and read them side by side.

How to build the report in Zendesk Explore

Start a new report on the Talk > Talk - Calls dataset. Before customising, check the prebuilt Zendesk Talk dashboard; several of the views below already exist there.

1. The customer experience view (call level)

Use these call metrics:

Metric	What it tells you
Call answer time	Time from the customer connecting to the first conference with an agent. Excludes voicemail and post-transfer waits.
Call wait time	Time the customer waited to talk to an agent after being routed where they wanted to go.
Call IVR time	Time spent in the IVR menu.
Call talk time	Time in conference with an agent, from the customer’s perspective.
Call on-hold time	Hold time from the customer’s perspective.

Slice those by the attributes that explain them: Call completion status, Call type, Call direction, Call outside business hours, Call exceeded queue wait time, Call IVR destination group, and Call Talk number.

Report answer time and wait time as medians, not averages. Voice distributions are heavily skewed by a handful of very long waits, and the same logic that makes median first reply time more honest applies here.

2. The abandon view

Abandonment is the metric that turns voice reporting into an operational tool. Build it from Call completion status, which distinguishes calls completed by an agent, sent to voicemail, and abandoned before either could happen.

Two cuts make it actionable:

Abandon rate by hour of day, compared against staffing. This is where coverage gaps become undeniable.
Abandon rate against wait time buckets. Most teams find a threshold — often somewhere between 60 and 120 seconds — past which abandonment climbs steeply. That threshold, not an industry benchmark, is your real answer-time target.

Also filter on Call exceeded queue wait time to separate customers who hit your configured limit from those who simply gave up early. They are different problems.

3. The agent behaviour view (leg level)

Switch to a leg-level report and use Leg duration, Leg talk time, and Leg wrap-up time, split by Leg agent name, Leg completion status, Leg agent available via, and Leg instance.

The most useful things this view reveals:

Wrap-up time as a share of leg duration. Long wrap-up is often the real handle time problem, and it is invisible in talk time alone.
Declined and missed legs by agent. Leg completion status shows whether an agent declined the call, missed the prompt, or accepted it. A rising decline rate usually means concurrency or status hygiene problems, not unwillingness.
Legs per call. A creeping average means more transfers and consultations, which is a routing signal. Compare with Zendesk Group Reassignment Rate Report.

4. Connect voice to the rest of the queue

Voice demand does not vanish when a call is abandoned. Pair the report with:

ticket volume on email and messaging in the hour after abandon spikes
repeat contact rate for the same requesters
channel mix trend over the same period

If abandoned calls at 9am are followed by an email spike at 9:30am, you do not have a voice problem and an email problem. You have one coverage problem being counted twice.

Business hours matter more on voice than anywhere else

The Call outside business hours attribute is true or false based on the schedules configured in your account, and it should be a standing filter on every voice report. Blending in-hours and out-of-hours calls destroys answer time and abandon rate at the same time: out-of-hours calls route to voicemail or go unanswered, which drags the blended numbers toward a conclusion nobody can act on.

Report in-hours performance as the operational number, and out-of-hours volume as a separate coverage question. For the wider distinction, see business hours vs calendar hours and the Zendesk After-Hours Ticket Report.

How to interpret the patterns

Answer time is fine but abandon rate is high

Look at the distribution, not the median. A healthy median with a heavy tail means a subset of callers wait far longer than the typical case — usually during a narrow peak or in one IVR destination group. Cut by hour and by group before touching staffing.

Talk time is rising while abandon rate falls

This is often a genuine improvement being misread as a problem. Longer conversations with fewer abandons can mean agents are resolving more on the first call. Confirm with first contact resolution before treating talk time as waste.

Agents are doing post-call admin that could be templated, automated, or moved into the call. This is one of the fastest capacity gains in most voice teams because it needs no headcount change.

High IVR time and high abandon rate together

Callers are getting lost before they ever reach a queue. Use IVR transitions and Call IVR action to find the menu step where they drop, and treat it as a self-service design problem rather than a staffing one.

Voicemail volume is climbing

Check whether Caller with requested voicemail is true. Requested voicemail is a customer choice; forced voicemail from an unanswered queue is a coverage failure. The two should never sit in the same number.

Common mistakes

Mixing call and leg metrics in one report. Call-level values get multiplied by leg count and every conclusion drawn from that report is wrong.
Reporting averages on skewed wait data. Use medians and percentiles; a single 40-minute wait can move an hourly average enough to hide a good day or invent a bad one.
Leaving out-of-hours calls in the operational number. Filter on Call outside business hours before you report anything.
Treating abandon rate as a voice-only metric. Abandoned calls become emails, chats, and callbacks. Track the spill-over or you will understate demand.
Counting transfers as agent failure. Legs per call is a routing and skills metric first. Look at Zendesk Routing and Assignment Metrics before drawing performance conclusions.
Ignoring wrap-up in handle time. An agent leg is not complete until wrap-up ends, and excluding it understates true capacity consumption.

What to do when wait time and abandons climb

Confirm the problem is in-hours by filtering on business hours.
Find the shape: is it all day, one peak hour, one group, or one Talk number?
Separate queue delay from IVR delay using Call IVR time and Call wait time.
Check leg-level decline and missed rates — sometimes the agents are scheduled but not reachable, which is an adherence question rather than a headcount one. See Zendesk Schedule Adherence Report.
Compare demand against forecast for that interval before adding heads. See Zendesk Forecast Accuracy Report.
Re-measure after two weeks using the same in-hours filter and median basis.

Where this report fits in your dashboard

Voice reporting works best beside:

Together they show when demand arrives, whether the queue could absorb it, and where unanswered voice demand resurfaces.

FAQ

Which Explore dataset should I use for call reporting? Start with Talk > Talk - Calls. Build a second report for leg-level analysis rather than adding leg metrics to the call report.

Why did my call totals double after adding an agent attribute? You mixed a call-level metric with a leg-level attribute. Call metrics are multiplied by the number of legs, so any call with a transfer or consultation is counted more than once.

What is a good average speed of answer? There is no universal number worth copying. Find the wait threshold where your own abandon rate starts climbing and set the target just below it.

Should abandoned calls count as tickets? For volume reporting, no — but they should count as demand. Track them alongside ticket volume so capacity planning reflects contacts you never handled.

Do voice conversations appear in AI agent reporting? Yes. Since the 2026 reporting change, AI agent conversation statuses are unified across messaging, email, and voice, and the Channel group ticket field separates Digital from Voice. See Zendesk Verified vs Contained Resolution Report.

See voice wait, abandons, and talk time next to the rest of your queue - start free