Section 3 — Call Summary Callback

This is the per-call analysis Zipteams sends back to you once a recording has finished processing. One call in via Section 1, one callback out.

It is the section most integrations need. Everything Zipteams inferred from that single conversation — intent, BANT, chapter summaries, qualification answers, pitch-quality scoring — arrives in one POST to a URL you own.

What you receive

If you are creating fields in your own CRM to hold this data, these are the outputs to plan for:

Output	Type	Notes for sizing your fields
Customer intent	string	A label such as `INTERESTED`. Short.
Customer intent justification	string	Free text, typically 3–4 lines. Use a long-text field.
Concern / low-intent category	string	The bucket behind a low intent, e.g. `Financial Constraints`, `Pricing`, `Functionality`. Delivered as `bant.concern_categories`. Short.
Call summary and action items	array of strings	Chapter-by-chapter. Can total up to roughly 600 words per call. Needs a long-text field, or one row per chapter.
Call quality score	number	0–100.
Qualification answers	array	One label + answer pair per topic discussed. Variable length.
Per-parameter quality scoring	array	One entry per configured audit parameter, with coaching suggestions.

How to enable it

Send callback_url in your call sync payload. That's it.

{
  "data": [
    {
      "call": { "id": "call_98231", "recording_url": "…", "start_time": "…", "phone_number": "…" },
      "agent": { "id": "AG-4471", "email": "priya.sharma@yourcompany.com" },
      "callback_url": "https://yourcompany.com/zipteams/call-analysis"
    }
  ]
}

This callback is enabled by default — no configuration on the Zipteams side is needed. If callback_url is present, you get the callback. If it is absent, you do not.

Never send an empty callback_url

"callback_url": "" is not a valid URL and will cause the entire call record to be rejected. If you do not want a callback, leave the key out of the JSON completely.

Don't want to build a webhook receiver?

Omit callback_url and view exactly the same analysis on the Zipteams Dashboard, or embed it in your own app with our embeddable components. No integration work required.

Delivery details

Property	Value
Method	`POST`
Content type	`application/json`
Destination	The exact `callback_url` you sent for that call
Timing	A few minutes after your call-sync request, depending on recording length
Expected response	`200 OK` — return it as soon as you have received the body, do not process synchronously
Retries	Up to 3 attempts with exponential backoff if your endpoint errors or times out
Identifier	`call_id` — always present, always matches the `call.id` you sent

Your endpoint must be publicly reachable over HTTPS.

If your endpoint restricts inbound traffic by IP

We can send these callbacks from a fixed whitelisted IP address:

13.201.157.246

This is not enabled by default. Contact us with the workspace and the callback URL, and we will switch it on for you. Whitelisting the IP on your firewall alone is not sufficient — the routing has to be enabled on our side too.

Payload

{
  "call_id": "call_98231",
  "customer_id": "CRM-100294",
  "agent_id": "AG-4471",
  "type": "CALL_SUMMARY",
  "intent": "INTERESTED",
  "intent_justification": "The customer confirmed budget availability and asked for a demo next week, indicating active evaluation. No pricing objection was raised.",
  "intent_score": 74,
  "quality_score": 82,
  "summary": [
    {
      "chapter_name": "Introduction and Requirement Gathering",
      "chapter_summary": "The agent introduced themselves and confirmed the customer had submitted an enquiry. The customer explained they are looking for a solution for a 40-member sales team."
    },
    {
      "chapter_name": "Pricing and Next Steps",
      "chapter_summary": "Pricing tiers were discussed along with the annual discount. The customer asked for a demo the following week and agreed to loop in their finance lead."
    }
  ],
  "bant": {
    "needs": "The customer needs call quality visibility across a 40-member sales team and wants to reduce manual call auditing effort.",
    "objection": "Concerned about how quickly the team can be onboarded before the quarter ends.",
    "concern_categories": "Implementation Timeline",
    "objection_handling": "The agent explained that onboarding takes under a week and offered to run the setup session on a Saturday to avoid disrupting the team."
  },
  "qualification": [
    { "label": "Needs", "answer": "Call quality visibility for a 40-member sales team." },
    { "label": "Budget", "answer": "Confirmed budget of approximately 3 lakh per annum." },
    { "label": "Timeline", "answer": "Wants to go live before the end of the quarter." }
  ],
  "quality": [
    {
      "parameter_name": "Setting Context",
      "justification": "The agent clearly stated the purpose of the call and confirmed the agenda with the customer before proceeding.",
      "suggestion": "Continue opening calls by stating the purpose and confirming the agenda.",
      "inference_score": 10,
      "parameter_score": 10,
      "status": "Done",
      "is_fatal": false
    },
    {
      "parameter_name": "Competitor Differentiation",
      "justification": "The customer mentioned they are also evaluating another vendor, but the agent did not explain how Zipteams differs.",
      "suggestion": "When a competitor is mentioned, walk through the two or three capabilities they do not offer before moving on to pricing.",
      "inference_score": 0,
      "parameter_score": 10,
      "status": "Needs Improvement",
      "is_fatal": true
    }
  ],
  "meta": {
    "internal_row_id": "8842",
    "campaign": "July Retargeting"
  }
}

Field reference

Top level

Field	Type	Always present	Description
`call_id`	string	Yes	The `call.id` you sent in the call-sync request. Use this to join the callback back to your own record.
`customer_id`	string	No	The `customer.id` you sent, if you sent one.
`agent_id`	string	No	The `agent.id` you sent.
`type`	string	Yes	Always `"CALL_SUMMARY"` for this callback. Use it to distinguish this payload from the Customer Summary Callback if both post to the same endpoint.
`intent`	string	No	The customer's overall intent, e.g. `INTERESTED`, `NOT_INTERESTED`, `NEUTRAL`, `NOT_QUALIFIED`, `NOT_AVAILABLE`.
`intent_justification`	string	No	Why that intent was assigned, in plain language. Useful to show agents and managers rather than just the label.
`intent_score`	number	No	Buying-propensity score from 0 to 100, rounded to a whole number. Higher means the prospect looks more like the profile of customers who historically converted.
`quality_score`	number	No	Overall call-quality score from 0 to 100, aggregated across all the parameters in the `quality` array. Higher is better. Use it for performance tracking; use the `quality` array for the per-parameter detail behind it.
`summary`	array	No	Chapter-by-chapter summary of the conversation. See below.
`bant`	object	No	Budget / Authority / Need / Timeline extraction. See below.
`qualification`	array	No	Answers to your configured qualification questions. Only questions actually discussed on the call are included. See below.
`quality`	array	No	Per-parameter pitch-quality scoring, for coaching. See below.
`meta`	object	No	The `metadata` object you sent in the call-sync request, returned verbatim. Empty object if you sent nothing.

`summary` array

Each element is one chapter of the conversation, in order.

Field	Type	Description
`chapter_name`	string	Title of this part of the conversation.
`chapter_summary`	string	What was discussed in this part.

`bant` object

Field	Type	Description
`needs`	string	What the customer said they need.
`objection`	string	Objections or concerns the customer raised.
`concern_categories`	string	Category of the concern, e.g. `Pricing`, `Functionality`, `Implementation Timeline`.
`objection_handling`	string	How the agent responded to those objections.

Any of these can be empty if the topic did not come up on the call.

`qualification` array

Field	Type	Description
`label`	string	The qualification question / topic, e.g. `Needs`, `Budget`, `Timeline`.
`answer`	string	What was established on the call for that topic.

Only discussed topics appear

Qualification topics that were not covered in the conversation are omitted from the array entirely — they are not returned with a null answer. Do not assume a fixed set of labels or a fixed array length.

`quality` array

One element per pitch-quality parameter configured for your workspace. This is the data to use for agent coaching and QA dashboards.

Field	Type	Description
`parameter_name`	string	The parameter being assessed, e.g. `Setting Context`, `Competitor Differentiation`.
`justification`	string	Evidence from the call for the score given.
`suggestion`	string	Actionable coaching feedback for the agent.
`inference_score`	number	The score the agent actually got on this parameter.
`parameter_score`	number	The maximum score available for this parameter. Use it as the denominator — parameters are not equally weighted.
`status`	string	Human-readable outcome derived from the two scores. One of `Done`, `Partially Done`, `Needs Improvement`, `Not Applicable`.
`is_fatal`	boolean	`true` means this parameter is critical for your business and needs attention regardless of the overall score. Surface these first.

How status is derived, if you want to reproduce it:

Condition	`status`
The parameter was not discussed / no data found on the call	`Not Applicable`
`inference_score` equals `parameter_score`	`Done`
`inference_score` is above 0 but below `parameter_score`	`Partially Done`
`inference_score` is 0	`Needs Improvement`

Receiving the callback

A minimal receiver — acknowledge immediately, process afterwards:

// Express example
app.post('/zipteams/call-analysis', (req, res) => {
  // 1. Acknowledge first. Zipteams retries on non-200 responses.
  res.sendStatus(200);

  // 2. Then process asynchronously.
  const payload = req.body;

  if (payload.type !== 'CALL_SUMMARY') return;   // ignore other callback types

  const yourRowId = payload.meta?.internal_row_id;  // whatever you put in metadata
  const callId = payload.call_id;                   // always present

  queue.push({ callId, yourRowId, payload });
});

Testing your endpoint

Before going live, verify your receiver handles the payload shape by posting the sample above to it yourself:

YOUR_CALLBACK_URL="https://yourcompany.com/zipteams/call-analysis"

curl -X POST "$YOUR_CALLBACK_URL" \
  -H "Content-Type: application/json" \
  -d '{
  "call_id": "call_98231",
  "customer_id": "CRM-100294",
  "agent_id": "AG-4471",
  "type": "CALL_SUMMARY",
  "intent": "INTERESTED",
  "intent_justification": "The customer confirmed budget availability and asked for a demo next week.",
  "intent_score": 74,
  "quality_score": 82,
  "summary": [
    { "chapter_name": "Introduction", "chapter_summary": "The agent introduced themselves and confirmed the enquiry." }
  ],
  "bant": {
    "needs": "Call quality visibility for a 40-member sales team.",
    "objection": "Concerned about onboarding time.",
    "concern_categories": "Implementation Timeline",
    "objection_handling": "The agent explained onboarding takes under a week."
  },
  "qualification": [
    { "label": "Budget", "answer": "Confirmed approximately 3 lakh per annum." }
  ],
  "quality": [
    {
      "parameter_name": "Setting Context",
      "justification": "The agent clearly stated the purpose of the call.",
      "suggestion": "Keep doing this.",
      "inference_score": 10,
      "parameter_score": 10,
      "status": "Done",
      "is_fatal": false
    }
  ],
  "meta": { "internal_row_id": "8842", "campaign": "July Retargeting" }
}'

Integration notes

Always key on call_id. It is the only field guaranteed to be present, and it always matches the call.id you sent.
Treat every analysis field as optional. Depending on the recording quality, its length and what was actually said, individual fields can be empty or absent. Never assume a field exists.
Respond 200 fast. Do your processing after acknowledging. A slow endpoint looks like a failure and triggers retries.
Be idempotent. Because failed deliveries are retried, your endpoint may see the same call_id more than once. Upsert rather than insert.
Use meta for correlation. Anything you put in metadata at sync time comes back here untouched. It is the cleanest way to tie the analysis to your own primary key without storing a mapping table.
Check type if you also enable Section 4 and both post to the same URL. CALL_SUMMARY is per call; CUSTOMER_SUMMARY is per contact.

Section 1 — Call Sync API — where callback_url and metadata are set
Section 4 — Customer Summary Callback — the contact-level equivalent

What you receive​

How to enable it​

Delivery details​

Payload​

Field reference​

Top level​

summary array​

bant object​

qualification array​

quality array​

Receiving the callback​

Testing your endpoint​

Integration notes​

Related​