Reports and management

Where the numbers are

Both dashboards use the same calculation criteria, so the numbers match when the period is the same. Periods count whole calendar days including today: "last 7 days" starts at midnight six days ago and runs until now.

Times counted only during business hours

ChatISP average times only count what falls within the configured business hours. Both screens say so right above the numbers: Times count only the configured business hours.

Example: the internet service provider (ISP) works from 08:00 to 18:00. A customer writes at 19:50 and the conversation is closed at 08:10 the next day. On the wall clock that would be more than 12 hours; in ChatISP this conversation counts as 10 minutes, which was the time actually spent within working hours. Without this rule, nights and weekends would inflate everyone's average.

Rules

• The schedule of the conversation's department applies when it has one; otherwise, the system's general schedule. Schedules are set up in Departments and in Settings.
• With no schedule registered at all, the clock runs nonstop (24 hours a day, 7 days a week).
• A conversation that started and ended entirely outside business hours counts as 0 seconds.
• Overlapping time ranges on the same day are merged, so no minute is counted twice.
• In the duration of conversations, the periods in which the conversation was marked as awaiting reply from the customer are also deducted: time when "the ball is in the customer's court" does not count against the agent.
• There is no option to turn this off: it is the standard way to measure support times.

What stays on the wall clock

• The timeline of each conversation, in the chat, shows the real times.
• Waiting on customer (how long, on average, conversations stayed awaiting reply) is the real wait, on the wall clock.
• Longest wait, in the Overview, shows how long the oldest customer in the queue has been waiting, in real time.
• Agents' time on break.

Support overview

For users with the Supervise all conversations permission, the Conversations screen opens by default on the Support overview: "Numbers, results and team performance in one place." It is the first item in the screen's sidebar (Overview); clicking any queue goes back to the conversations, and a direct link to a conversation opens the chat without going through the overview.

Period and refresh

• Buttons at the top: Today, 7 days (default), 30 days and 90 days.
• Today runs from midnight until now, the chart switches to hourly (from 0h to the current hour) and the comparison is made against all of yesterday.
• The screen refreshes on its own every minute while it is visible. With the tab in the background or another queue open, it does not query the server.
• The charts adapt to the light or dark theme and to the screen size.

Support activity

The main block shows the total number of conversations received in the period, the sentence "X received · Y closed in the period" and an area chart with two lines: Received and Closed, per day (or per hour, in Today).

Next to the total is the change compared with the immediately preceding window of the same length (for example, the 7 days before the last 7), with an up or down arrow and the text vs previous period (or vs yesterday, in Today). If the previous window had no conversations, the change is not shown.

How conversations end

Donut chart of the conversations closed in the period, with the percentage solved by the AI in the center:

Overview: Right now

The cards in the Right now strip (with the Live badge) show a snapshot of the moment and do not depend on the chosen period:

Overview: In the period

The cards in the In the period strip follow the selected period button. Times follow the business hours rule.

When there is not enough data for an average time in the period, the card shows "-".

Overview: panels

When customers reach out

Heat map of 7 days of the week by 24 hours, with the number of conversations opened in the period for each day and hour combination. The stronger the color, the more activity. The Peak appears in the corner, for example "Peak: Mon at 9h". Use it to size the team's schedule.

Support flow

Funnel of conversations opened in the period, with the percentage of each stage over the total received:

Team

"Each agent's performance in the period", with up to 30 active agents who had conversations or messages in the period. At the top, three highlights:

• Most conversations: who closed the most conversations (in a tie, who handled the most).
• Fastest to pick up: lowest average time to pick up, among those who handled at least 3 conversations.
• Best rated: highest average score, among those who received at least 2 ratings.

The best value of Closed, Time to pick up and Score is highlighted in the table.

Departments, Closing reasons and Channels

• Departments: conversations opened in the period by department, with "wait time · N in queue": the average time to pick up in that department (on its own schedule) and how many conversations are in its queue right now.
• Closing reasons: how many conversations closed in the period received each reason.
• Channels: conversations opened in the period by connection, indicating whether it is WhatsApp (QR) or Official API.

Service orders

Only shown to users with the Manage service orders permission. Shows Open, Unassigned, Overdue and Solved today; each card and the Open board link lead to the Service orders board.

AI assistant

The percentage "of closed conversations solved by the AI", followed by Conversation turns (how many times the AI processed a reply in the period), Errors (turns that failed), Average latency (average AI response time) and Solved by the AI.

Reports screen

Open Reports in the menu (Reports permission): "Support volumes and results". At the top are the period selector (Last 7 days, Last 30 days, Last 90 days) and the Export CSV button. Changing the period reloads the whole screen and also adjusts the export.

Top indicators

Report charts and tables

Satisfaction survey

When the Send the survey when closing handled conversations option is turned on in Settings, the customer receives the configured question at the end of each conversation handled by a person (default: "How would you rate today's service? Reply with a score from 1 to 5."). The reply is only recognized as a score if it contains a number from 1 to 5, and a customer who replies with a score receives an automatic thank you message. Details in Settings.

Export CSV

The Export CSV button on the Reports screen downloads a spreadsheet with one row per conversation opened in the selected period. The file is named chatisp-conversas-30d-AAAA-MM-DD.csv (with the number of days in the period, followed by the date as year, month and day) and opens directly in Excel, LibreOffice or Google Sheets with correct accented characters.

With the system in Portuguese or Spanish, columns are separated by semicolons, which is the default for spreadsheets configured for Brazil; in English, by commas. Column names follow the system language.

AI consumption

On the Audit log screen, the AI consumption button opens the "Cost audit: what the AI consumed in each ticket and how much it cost". Access is the same as for the Audit log (Audit log permission), and the period can be 7, 30 or 90 days.

Period summary

• Cost in the period, in Brazilian reais, with the dollar exchange rate used and its date. With no exchange rate available, the amount appears in dollars with the notice No exchange rate available: amounts shown in US dollars (US$).
• Calls made to the AI, Input tokens and Output tokens.

Table by ticket

One row per conversation (ticket number and customer name), from most recent to oldest, showing what the AI was Used for and how many times (Customer service, Ticket summary, Service order summary, Translation, Menu titles, Menu interpretation, Agent text, Copilot), the calls, the tokens and the Cost. Usage that does not belong to any conversation (for example, the Copilot outside a conversation) is grouped as Outside tickets. The list loads more rows as you scroll.

Tokens are the exact numbers returned by the AI provider on each call, with no estimates. The cost is calculated from the official price table for each model.