# Conversation intelligence

> Call recordings and visit notes are evaluated automatically — a BANT / MEDDIC scorecard, coaching tips, keyword trackers and alerts, competitor, price, objection and loss reason analysis; a dashboard, a report type and the opportunity card.

Conversation intelligence measures what was discussed in sales conversations. It looks at two sources: the transcript of [call recordings](/en/guides/calls) and [visit notes](/en/guides/visits). Every conversation is evaluated in the background when it's saved (or when a note is edited); there's nothing to do by hand.

For each conversation:

- **Method scorecard** — for every criterion of the sales method you chose (BANT or MEDDIC): *Discussed*, *Partly* or *Not discussed*, with evidence (a quote from the conversation and its time) and a suggested question for each criterion that wasn't discussed.
- **Conversation metrics** (on call recordings) — talk share (your side), the number of questions you asked, your longest monologue, and whether the conversation ended with a clear next step.
- **Coaching tips** and a **conversation score** from 0 to 100.
- **Signals** — the keywords you track, competitor names, a competitor's price, objections (and whether they were answered), loss risk and the next step.

## Methods {#methods}

Pick a method in Settings › Sales process › **Conversation intelligence**; every conversation is evaluated with it.

| Method | Criteria | When |
|---|---|---|
| **BANT** | Budget · Authority · Need · Timing | First meetings and discovery; short sales cycles |
| **MEDDIC** | Metrics · Economic buyer · Decision criteria · Decision process · Identify pain · Champion | Large, multi-stakeholder, long sales |

A criterion counts as *Discussed* when its topic came up in the conversation (for Budget: budget, price, payment terms; a sentence with an amount makes it stronger). If it came up only once and indirectly, it's *Partly*. When you change the method, conversations are evaluated again.

## Conversation score {#score}

| Source | Calculation |
|---|---|
| Call recording (speakers known) | Method coverage 70% · talk balance 10% (full marks when your share is 35–65%) · questions 10% (full marks at 6 or more) · next step 10% |
| Visit note | Method coverage 80% · next step 20% (in the note or the *Next action* field) |

75 and above is green, 50–74 amber, below 50 red. Coverage is the share of criteria discussed (*Partly* counts as half).

## Coaching tips {#coaching}

Up to eight tips per conversation, most important first:

- Criteria that weren't discussed (with a suggested question) and a partly discussed criterion.
- Your talk share above 65%: let the customer talk more; below 25%: make your proposal clearer.
- Fewer than 4 questions in a conversation longer than 90 seconds; a single monologue longer than 2.5 minutes.
- No clear next step.
- A competitor came up: explain your difference with numbers; an objection left unanswered.

## Keyword trackers and alerts {#trackers}

A tracker is a list of words and phrases to look for in conversations (Turkish letters are folded, matches start at a word boundary: *fiyat* → *fiyatınız*). Built-in trackers:

| Category | Trackers |
|---|---|
| Competitor | **Competitors** — competitor names on company cards and opportunities are added automatically, plus phrases like *rakip firma* (another company) or *başka tedarikçi* (another supplier) |
| Objection | Price · No budget · Timing · Current supplier · Technical concern · Authority · No need |
| Loss risk | *iptal* (cancel), *ertelendi* (postponed), *başka firmayla* (with another company), *bütçe kesildi* (budget cut)… |
| Next step | *göndereceğim* (I'll send), *numune* (sample), *revize teklif* (revised quote), weekdays… |

You can add your own trackers (product names, a campaign, a complaint topic), change the words and switch built-in ones off. Each tracker has a **Notification** setting: *No notification*, *To the record owner* or *To the owner and admins*. The alert arrives as a **Conversation signals** notification (bell, email, phone — [Notifications](/en/guides/notifications)) the first time the tracker comes up in a conversation; it isn't sent to the person who logged the conversation, nor for conversations older than 3 days. On call recordings, objection and risk trackers look only at what the customer said.

**Back to built-in trackers** resets the trackers (trackers you added are removed); **Re-evaluate** scores every conversation again with the new settings.

## Competitor, price and objection {#signals}

- **Competitor:** a competitor name (from the tracker) mentioned in the conversation is recorded; unnamed phrases like *another supplier* count too.
- **Competitor's price:** a sentence with a price (*they sell it at €4.20 a kilo*) is linked to the competitor named in the same or the previous sentence; currency and unit (kg, l, piece, tonne…) are read. These prices reach the company card through the [Competitor price log](/en/guides/agents#prices) agent.
- **Objection:** each objection tracker is a separate objection. On call recordings, an objection followed by an answer from your side counts as *answered*, otherwise *unanswered*.

## Where you see it {#where}

| Place | Content |
|---|---|
| **Call recording page › Coaching** | Scorecard, evidence (click a time to play the recording from there), conversation metrics, tips, signals. **Re-evaluate** button |
| **Visit page** | Scorecard and signals for the note |
| **Opportunity card** | Method coverage across all the opportunity's conversations; **Mark in BANT** for criteria discussed in conversations but not ticked in the opportunity's BANT boxes (Qualify steps included) |
| **Conversation intelligence page** | The team's dashboard (below) |
| **Dashboards and report builder** | The *Conversation intelligence* report type and a ready-made dashboard |

## Conversation intelligence page {#page}

**Reports** in the menu › **Conversation intelligence** on the tab strip (next to Dashboard · Pipeline · Forecast · Dashboards · Reports; `/gorusme-zekasi`, *Conversation intelligence* in the command palette). At the top: period (Last 30 days · 90 days · 6 months · 12 months), source (All · Call recordings · Visit notes) and, for admins, a rep picker; sales users see only their own conversations.

| Tab | Content |
|---|---|
| **Overview** | Conversations evaluated · Average score · Method coverage · Without next step · Competitor mentioned; weekly score, coverage by criterion, keywords, number of conversations |
| **Coaching** | Average score by rep, talk share distribution (target 40–60%), coverage and the lowest-scoring conversations |
| **Keywords** | Conversations each tracker came up in, and the four most frequent trackers over time |
| **Competitors** | Conversations with competitors, outcome of opportunities where a competitor came up (won · lost · open), latest competitor prices |
| **Objections** | Conversations with objections and the outcome of opportunities where an objection came up |
| **Loss reasons** | Opportunities lost in the period; the **recorded loss reason** next to the **objection raised in conversations**, losses where another reason stands out in conversations (e.g. *Timing* recorded but a price objection was raised), competitor in losses, stage lost at |

Every number and chart slice can be clicked to open **the conversations / opportunities behind that number** on the right (number = list). **Play** and **Video** show the period as a short presentation ([Play and video](/en/guides/dashboards#motion)).

## Report type and ready-made dashboard {#reports}

In the report builder, the **Conversation intelligence** type: number of conversations, average score, method coverage, keywords, conversations with competitors, outcome of opportunities with competitors, objections, loss reason (recorded / raised in conversations), score by rep and talk share. The date range applies to the conversation date. **New dashboard › Conversation intelligence** sets up a dashboard of these reports. [Report types](/en/guides/dashboards#kinds)

## AI {#ai}

Rule-based evaluation runs on every plan and the conversation isn't sent anywhere. For call recordings, if AI is on (Settings › Conversation intelligence › *Also evaluate call recordings with AI*; a Call Recording licence is enough) the transcript is also sent to the model. Every finding the model returns is checked against **a quote that appears verbatim in the transcript**; an item whose quote isn't in the transcript is dropped or downgraded to *Partly*. The model isn't called again until the transcript changes.

## API {#api}

The recording detail ([`GET /calls/{id}`](/en/rest-api/calls/calls-get)) includes `call.intel`: `method`, `score`, `coverage`, `criteria` (`key`, `label`, `status` met · partial · missing, `evidence`, `t`, `question`), `coaching`, `talk_pct`, `questions`, `longest_sec`, `next_step`, `objections` (`handled`), `competitors`, `prices` (`value`, `currency`, `unit`, `competitor`), `risks`, `trackers`.

## Limits {#limits}

- Talk share, question count and whether an objection was answered are measured only on call recordings with known speakers; visit notes don't have them.
- Rule-based evaluation looks at words: *bütçemiz yok* (we have no budget) is an objection, but the same sentence also counts Budget as discussed.
- Conversations from up to the last 12 months are evaluated.