# Chat, appointment and performance analytics

> Measure how many chats your agent handles and escalates, how many appointments get booked and how fast each advisor responds, with date filters and CSV export.

Source: https://docs.watsi.ai/en/insights/analytics

**Analytics** shows you how your WhatsApp operation is doing: how many chats came in, what share the AI answered, how many were escalated to a person, how many appointments were booked and how fast each advisor responds. It has four views — Conversations, Appointments, Performance and Conversation usage — with date filters and a comparison against the previous period.

## Open Analytics

### Go to "Analytics"
    It is in the side menu and opens the **Conversations** view.
### Switch views
    Use the selector next to the title to move to **Appointments**, **Performance** or
    **Conversation usage**.
### Pick the period
    Use "7 days", "30 days" or "90 days", a shortcut ("Today", "Yesterday", "Last week", "Last
    month", "Last quarter") or "Custom". Weeks start on Monday.

> **Note**
  By default only **owners and admins** see Analytics. You can give it to other members with a
  custom role.

## Conversations

Shows the **last 7 days** by default. Each card is compared "vs previous period".

| Card                 | What it measures                                                                   |
| -------------------- | ---------------------------------------------------------------------------------- |
| **Total chats**      | Conversations in the period.                                                       |
| **AI response rate** | Share of answers resolved by the AI out of all chats in the period.                |
| **Escalation rate**  | Share of chats that required human intervention.                                   |
| **AI latency**       | How long the AI takes to generate each reply. It is not your team's response time. |
| **Advisor response** | Median response time of your advisors, in business minutes.                        |

Below the cards you can switch between three sections:

- **List** — the period's chats, with search ("Search by name or phone…") and filters by Status, Channel, Origin, Advisor and Stage.
- **Funnel** — the "Chat funnel": how many **Entered**, were **Profiled**, **Booked**, **Attended** and how many were a **No-show**. With two or more agents you can filter by agent.
- **Volume** — "Chat volume" per day, grouped in your workspace's time zone.

### Export to CSV

In the list, "Download CSV" opens "Download options" so you can pick columns (customer, phone, topic, channel, origin, status, assigned advisor, stage, messages and last activity). Each download includes up to **5,000 rows**.

## Appointments

Also shows **7 days** by default, but it looks **forward**: from today through the next days of the period, compared with the same number of days before today.

| Card                   | What it measures                                                               |
| ---------------------- | ------------------------------------------------------------------------------ |
| **Total appointments** | Appointments in the period.                                                    |
| **Confirmed**          | Appointments scheduled or confirmed.                                           |
| **Attendance rate**    | Of the appointments already marked attended / not attended, how many attended. |
| **Cancelled**          | Appointments cancelled or declined.                                            |

The **Volume** section shows "Appointment volume" per day.

## Performance

The "Team performance" view shows **30 days** by default and measures your advisors: response time, SLA, reassignments and appointments. Filter by advisor with "All advisors".

- **Advisor response** — from the moment the chat is assigned until the first human action (a reply to the customer or an internal note). AI messages do not count.
- **SLA compliance** — the share of assignments where the advisor acted before their SLA ran out.
- **Reassignments** — how many times a chat moved from one advisor to another (the first assignment does not count).
- **Appointments booked by AI**.

Further down you get **At risk now** (assigned chats with no response, close to breaching), **Performance by advisor** (p50 and p90 response, % within SLA, received, lost, AI appointments, own appointments, attended and notes), **Response time distribution**, **SLA compliance**, **Reassignment source**, **Who booked the appointment** and **Human follow-up**.

> **Tip**
  Times are measured in **business hours** following the schedule in Settings: a lead arriving at
  21:00 and answered at 09:05 counts as 5 minutes, not 12 hours. With no schedule configured,
  natural time (24/7) is used.

> **Note**
  If your role only lets you see your own figures, you get the notice "Your role only allows viewing
  your own figures." and the filter stays fixed on you.

## Conversation usage

Shows how many conversations your workspace used in the period. Your balance, plan and top-ups are managed in **Usage**, from your account's settings menu.

## Related

- [Assistant](/insights/assistant)
    Ask about your metrics in plain language.
- [Escalation to a human](/agent/escalation)
    When the agent hands a chat to your team.
- [Sales funnel](/agent/funnel)
    Define the stages each prospect moves through.
- [Assignment and profiling](/agent/assignment-and-profiling)
    How chats are shared out among your advisors.

## FAQ

### What does "AI latency" measure?

How long the AI takes to generate each reply, measured per message. It is not your team's response time; that is in Analytics → Performance, as "Advisor response".

### How is an advisor's response time measured?

From the moment the chat is assigned to them until their first human action in it, whether a reply to the customer or an internal note. It is counted in business hours following the schedule configured in Settings; AI messages never count.

### Who can see Analytics?

By default, owners and admins. A member does not see it unless you give them a custom role with that permission.