# Bot conversation analytics: which numbers an owner should watch

What Onswer analytics shows: new conversations, leads, agreements to buy, response time, follow-ups, channels and ad tags, reasons for lost customers and plan usage.

Updated: November 5, 2026 · https://onswer.ai/docs/en/analytics/dashboard · Onswer

Analytics answers the owner's questions: how many customers write, how many of them leave a contact, how many agree and why the rest leave. Test chats and test calls from the dashboard are not included in the numbers.

> **Note:** Detailed analytics is included in the Starter plan and above. On the free plan an example marked "Example" is shown instead.

## Overall analytics

The "Overall analytics" button gathers all your bots in one place for a chosen period (for example 30 days):

| Metric | What it means |
|---|---|
| New conversations | How many new customers wrote |
| Customer messages | Incoming messages |
| Bot replies | The bot's outgoing answers |
| Leads | Customers who left a contact |
| Agreed | A purchase, a meeting or a call |
| Calls | The number of calls and bookings from them |
| Conversion | The share of agreed customers and call bookings to the number of conversations |

Below are a chart of messages by day and a comparison of the bots. The "Plan" block shows how many messages and call minutes are used from the limit.

## One bot's analytics

In the bot's card, the "Analytics" tab:

- unique conversations, messages, new users;
- **average response time**;
- how many **follow-ups** were sent and how many customers replied after them, see [follow-ups](/docs/en/sales/follow-up);
- a split by channel: WhatsApp, Telegram, web chat, phone;
- for the web chat, **ad tags (UTM)**, to see which ads bring customers;
- **"Why we lose customers":** customers who went quiet over a day ago, left no request and did not pay. The reason is determined by rules, so it is approximate: price, not answered in time, handed to a manager and the customer vanished, the service did not fit;
- **"Why they buy and leave":** results from customer cards that the AI fills in (by the [lead fields](/docs/en/bot/lead-fields)).

## What to do with the numbers

- Few leads with many conversations: check whether the bot asks for a contact ([instructions](/docs/en/bot/instructions)).
- Slow answers: see whether the bot waits for you to answer "Needs a human".
- Many losses on price: add arguments and installments to the [knowledge base](/docs/en/bot/knowledge-base) and go through the [weekly review](/docs/en/analytics/weekly-review).
