Call analysis

One call, fully opened up: the audio, the transcript, what was said and felt on both sides, and where the score came from.

Opening a call

Three routes lead here:

The Preview (eye) action, or the row itself, in Recent Calls on the dashboard.
A notification — when a colleague tags you in a comment, the bell takes you straight to that call.
A direct link. The URL carries the workspace and call ID, so it can be pasted into a ticket or a chat and it will open the same call for anyone with access.

Header and Player

The header names the audio file and, underneath, the call direction and duration. Inbound or Outbound is inferred from the filename prefix, so it appears when your recorder’s naming convention includes it.

The waveform player below it gives you:

Play / pause and stop;
playback speed presets, for skimming a long call;
a seekable waveform — click anywhere to jump, and quiet stretches are visible as flat sections, which is often how you spot dead air before you read about it;
the elapsed and total time.

Call preview header with the audio filename, direction, and the seekable waveform player with speed presets — The filename and direction, then the seekable waveform. Speed presets — 0.8x, 1x, 1.2x — are the difference between reviewing four calls in an hour and reviewing ten.

The three headline stats

Stat	Meaning	How to read it
Overall Sentiment	Sentiment of the call as a whole, from the transcript.	Green positive (High), amber neutral (Medium), red negative (Low), with a three-bar strength meter.
Overall Call Score	The call’s score out of 100, built from the opening, handling and closing weightage set on the workspace.	90+ Excellent · 75–89 Good · 50–74 Average · under 50 Poor.
Talk Ratio	Share of the call the agent spent talking.	Very high suggests the agent talked over the customer; very low can mean a passive agent or a long complaint. Compare against your workspace’s percent-talk goal.

Text sentiment

Sentiment is reported three ways: for the whole call, for the agent alone, and for the customer alone — each as a donut with the dominant sentiment in the centre. The split is what makes it useful: a call that is negative overall but positive on the agent’s side is usually a legitimately upset customer being handled well, which is a very different coaching conversation from an agent who turned the call negative.

Call Sentiment donut with agent and customer rings beside the Average Emotion During Call bar chart — Both panels sit side by side on the call. Sentiment gives the whole call in the large donut and the agent and customer as separate rings; emotion does the same across the six classes. **Achievement**, top right of the emotion card, compares the call against the workspace's targets.

Emotions

Emotion comes from the audio, not the words, across six classes — happy, sad, neutral, fear, disgust and surprise — again for the full call and per speaker. Where the workspace has emotion targets configured, a benchmark strip shows how this call compares against them.

Why both

Sentiment reads what was said; emotion reads how it sounded. Polite words in a fearful or disgusted voice are the pattern that manual sampling misses most often, and it only shows up when you look at the two side by side.

Call summary

A short written summary of what happened on the call — enough for a supervisor to get the gist without listening. The Listen button reads it aloud, which is handy when you’re working through a batch. Use the summary to triage, then use the audio and transcript to verify before acting.

Talk, Silence and Cross talk

The aggregation panel covers the mechanics of the conversation:

Percent Talk How much of the call was the agent speaking.
Percent Silence Dead air. Long silences point at hold behaviour, tool problems or an agent hunting for information. The threshold that counts as dead air is set per workspace (3–10 seconds).
Percent Cross Talk Both parties speaking at once — interruption. High cross talk on negative calls is one of the most reliable escalation signals in the platform.

Call Aggregation line chart of duration, dead air, cross talk and talk minutes beside the % Talk donut — Call Aggregation plots the mechanics minute by minute — duration, dead air, cross talk, monologue count, agent and customer talk — while **% Talk** gives the talk, silence and cross-talk split for the call as a whole.

Customer Interaction Indicator

This is the soft-skills checklist, scored per call. Each row is a skill, a score bar, and the reference words the platform listened for. The header chip reports how many skills were met — for example 7 of 10 skills met — and the column is sortable, so the misses float to the top.

Skill	What it looks for
Welcome	A proper greeting at the top of the call.
Name Verification	Confirming who the agent is speaking to.
My Name	The agent introducing themselves.
Courtesy	Please, pardon, excuse me — basic politeness markers.
Active Listening	Signals the agent was engaged rather than reading a script.
Cushioning	Reassurance — “I’m here to help you”.
Probing	Questions that establish the real problem.
Assistance	Offering concrete help.
Cust Education	Setting expectations — what happens next, and when.
Interaction Closing	A clean close: thanks, appreciation, next steps.

Customer Interaction Indicator table with each skill scored as a percentage bar alongside the reference words — Each row is a skill, its score, and the reference words the platform listened for. Sorting by score floats the misses to the top.

Call score

The score card breaks the number into its three stages, each with the parameters it draws on:

Stage	Parameters
Opening Score	Welcome · Name Verification · My Name
Handling Score	Courtesy · Cust Education · Active Listening · Assistance · Cushioning · Probing
Closing Score	Interaction Closing

The Overall Call Score combines the three using the weightage on the workspace, so a collections floor can weight handling above opening while a sales floor does the opposite. An info control on the card shows the calculation method with your workspace’s current weights filled in. Change the weights in Goals & targets.

Opening, Handling, Closing and Overall Call Score cards, each drawn as a dial out of 100 — The three stage scores and the combined result. A strong opening and close with a weak middle — as here — points the coaching straight at handling.

Transcript tab

Speaker separated — every turn is labelled Agent or Customer.
Click to jump — selecting a line moves the player to that moment. This is the fastest way to verify a flag: read the line, hear the tone.
Coaching chips — flagged turns carry a chip that opens the detail: the Issue, what was Flagged, and a Say instead suggestion you can hand to the agent verbatim.
PII masked — where masking is enabled on the workspace, names, account numbers and ID numbers are redacted in the transcript. See PII masking.

Speaker-separated transcript: agent turns on the left, customer turns on the right in dark bubbles, each with a play control, a timestamp and an emotion marker — Agent turns sit left, customer turns right, each with its own play control and timestamp so a line can be heard without hunting for it in the waveform. The face markers down the middle are the per-turn emotion read, and highlighted spans are the phrases that drove the scoring.

Comments tab

Comments keep the QA conversation attached to the call instead of scattered across email. Write a note, tag colleagues from the selector, and post — the people you tagged get a notification whose click-through opens this exact call. The tab is available to admins and users; agents don’t see it.

A practical review workflow

Start from the negatives Filter the dashboard to negative sentiment over your date range.
Read the summary Thirty seconds tells you whether the call is a real problem or an angry customer handled correctly.
Check the split Agent-side sentiment and emotion separate "customer was upset" from "agent made it worse".
Verify at the source Jump from the flagged transcript line into the audio. Never coach on a flag you haven't heard.
Look at the mechanics Talk ratio, silence and cross talk usually explain why the call went the way it did.
Leave the coaching note Comment on the call and tag the supervisor, then bookmark it so it's in your next session's list.