Reading your call history
Every call carries a transcript, a recording, sentiment and its own cost breakdown. Which view answers which question, and how to find the call that explains a bad week.
6 min readUpdated 26 August 2026
Four pages sit under Analyze and they answer different questions. Being in the wrong one is the usual reason people say they cannot find something.
| Question | Page |
|---|---|
| How are things overall, right now | Overview |
| How is this trending, and what is the sentiment | Analytics |
| What happened on that specific call | Call Log |
| What did callers actually tell us | Collected Data |
Overview
Described as your dashboard at a glance, over a period you choose. It carries number of calls, average duration, average latency, total cost, average cost, successful calls and top sentiment, with a chart of the overall success rate across the period.
Average latency is the number worth watching here and the one people ignore. It is the pause before your agent replies, and it is the single strongest driver of whether a call feels natural.
Analytics
The fuller view, with a start and end period you set. Alongside the same headline figures it adds three things that matter.
- Calls by Date, the daily volume across the period.
- Call Sentiment, the distribution across negative, neutral, positive and very positive.
- Call Disconnect Reasons, which is how calls ended.
Tip
Disconnect reasons is the most under-read chart on the platform. A rise in calls ending one particular way is usually the first sign something changed, and it shows up there before it shows up in sentiment.
Call Log
The list of individual calls, filterable by agent, status, duration, date and disconnect reason, with an export.
Each row opens a panel with the agent, status, call type, duration, disconnect reason and timestamps, and a link through to the full detail.
What is in a call's full detail
This is the most useful screen on the platform and the least visited.
- The audio recording, with a download.
- Call analysis, including the disconnect reason.
- A cost breakdown: total units, the flat platform fee, the total cost, and then one line per provider showing the model, the units consumed, the unit type and the cost.
That breakdown is what makes cost debuggable rather than mysterious. A call that cost more than you expected tells you which component did it, so you know whether to change a model, shorten a prompt, or accept that the call was simply long.
Finding the call that explains the week
The habit that works is to start wide and narrow. Find the day or the agent that stands out in Analytics, then open one call from it and read the transcript.
Aggregates tell you where to look. Only a transcript tells you what actually went wrong, and it is almost never what the numbers suggested.
- Volume up, duration flat: you are simply busier.
- Volume flat, duration up: something is making calls drag. Usually a prompt that does not close them out.
- Both flat, cost up: a model changed.
- Sentiment down with no other change: read the negative calls. There will be a pattern.
Common questions
- Where do I see what one call cost?
- Open the call in Call Log and go to the full detail. The cost breakdown itemises the platform fee and every provider line.
- What is average latency and why does it matter?
- The pause before your agent replies. On a phone call that gap is the most noticeable thing in the conversation, which is why it drives how natural a call feels more than answer quality does.
- Can I export call data?
- Yes, from Call Log, and separately from Collected Data for the structured fields captured on each call.
- Why is my success rate lower than I expected?
- Check disconnect reasons first. Success rate is an outcome; disconnect reasons tell you the mechanism behind it.
- Do web widget conversations appear here?
- Yes. They are conversations with the agent, so they sit alongside phone calls with their own transcripts and costs.
Was this helpful?