The calls are the cost. The results are what you bought, and they arrive in two layers: what happened to each call, and what each call found out.
Everything below is on the campaign at /campaigns/….
The results summary
The part worth opening first. Results summary aggregates your Dialog Insight fields across the whole run — each value with a count and a percentage, so you read the campaign rather than the calls.
Two notes it gives you itself: “Only the most common values are shown”, and “No conversations from this campaign have been analysed yet” when it is too early. If it stays empty, the cause is almost always no insight attached — see below.
What happened to each call
Every record ends with a Disposition, and the list filters by All Dispositions and All Statuses. Read them as three groups rather than nine values:
| Group | Outcomes | What it tells you |
|---|---|---|
| Reached a person | answered_human | Your real contact rate |
| Reached nothing | no_answer, busy, voicemail | Worth retrying |
| Reached something wrong | disconnected, wrong_person, ivr_system | A data problem in your list |
That third group is the one people mistake for poor performance. A campaign with a
low contact rate and a pile of disconnected did not fail — it told you your list
is stale, and calling again will not fix it.
Click a record for Call Details: the transcript, the Call Recording with Play recording and Download recording, and the fields extracted from that call.
What each call found out
The structured fields — the answers, not the transcript. This is the difference between results you can act on and results somebody has to read. Sorting a hundred calls by a field takes a second; listening to a hundred calls takes a day, and nobody does it twice.
Export CSV gives you the whole thing. The Field key on each Dialog Insight field is “used as the CSV column header”, so name fields the way you want the export to read — before the run, not after.
Where the insight comes from
Worth knowing, because it explains an empty summary. A campaign either has its own or inherits the agent’s:
- “Inherited from the agent. This campaign has no override of its own.” — the usual and usually right case.
- “Overrides the agent default …” — this campaign asks different questions.
- “This campaign has its own extraction fields. They still run, but no other campaign or agent can use them.” — the Custom (not reusable) state. One press of Save as a reusable Dialog Insight fixes it.
- “This Dialog Insight no longer exists” — a dangling attachment, and nothing will be extracted.
If the summary is empty and the calls connected, this list is where to look.
Retry the outcomes worth retrying
A retry policy can be restricted to specific outcomes, and it should be. Calling back a no-answer is normal. Calling back someone who hung up on you is not, and calling a disconnected number again is spending money on the same failure.
Set the outcomes explicitly — the “retry everything” default is what turns a campaign into a complaint.
The callback list is the point
Our recommended pattern, and the one we run: define a field that marks a call as needing a person, then work that list.
Better than a live transfer for the same reason it is better on inbound — a transfer needs somebody free at that exact moment, and when nobody is, the person has been handed to a ringing phone. A flagged call is a complete conversation plus a callback queue already sorted by priority.
Auto-triage prioritises alongside it, and a field set to become a label turns your own value into something the inbox can filter on.
It does not have to end there
A finished call is an event a workflow can start from, so the follow-on is not a fixed list of features. Update the contact. Email the result to whoever owns it. Add the interested ones to a list and the rest to a suppression list. Book the follow-up. Write it into your own system.
The same event covers a chat conversation closing, so a workflow written for calls usually works for your other channels unchanged.
Where it all lives
Campaign results also land in /conversations with everything else — transcript, recording, fields, priority. A campaign is not a separate report to go and find, which matters because a separate report to go and find is a report nobody reads twice.