> ## Documentation Index
> Fetch the complete documentation index at: https://docs.armin.cx/llms.txt
> Use this file to discover all available pages before exploring further.

# How metrics are calculated

> Understand the definitions, timing, date ranges, and exclusions behind dashboard metrics.

Use this reference when a dashboard number does not match what you expect.

## Date ranges and trends

Widget dashboards currently filter by date only. Choose today, the last 7, 14, or 30 days,
or a custom range of up to 90 days. Every widget query is limited to 90 days.

The trend compares the current range with the previous range of equal length. Its direction
shows whether the value increased or decreased.

## When values appear

Most ticket-level facts are calculated asynchronously after a ticket moves to **Resolved**
or **Closed**, so they can take a short time to appear. Reopening the ticket clears those
facts. They are recalculated from the ticket history after the next resolution.

Some metrics are calculated live when the widget runs.

## Response and resolution time

### First response time (FRT)

FRT is the business-hours wait until the first human response. Auto-replies, AI messages,
workflow messages, and internal notes do not count as a human response. If AI replies before
the first human, the ticket is omitted from FRT.

The FRT widget shows the median and falls back to the average when a median is unavailable.

### Average resolution time (ART)

ART is the business-hours time from ticket creation to resolution or close. Time after a
human agent reply is excluded until the customer responds. ART is used mainly in the agent
leaderboard.

### Business hours

FRT and ART use your organization's weekly schedule, special days, and timezone. If no
business hours are configured, they use a 24/7 schedule.

## Support metrics

| Metric family       | What it shows                                                                       |
| ------------------- | ----------------------------------------------------------------------------------- |
| **Ticket volume**   | New, open, and resolved tickets in the selected period                              |
| **SLA performance** | Compliance with first-response or resolution targets                                |
| **Topic analysis**  | Ticket volume grouped by detected topic                                             |
| **Tag analytics**   | Ticket volume grouped by tag                                                        |
| **Ticket fields**   | Ticket volume grouped by a selected ticket field                                    |
| **Support revenue** | Attributed order revenue in the selected period; this usually requires Shopify data |

## AI impact

| Metric              | How it is calculated                                             |
| ------------------- | ---------------------------------------------------------------- |
| **Automation rate** | Average AI share of tracked actions across eligible tickets      |
| **Time saved**      | Estimated time per action multiplied by the number of AI actions |
| **Money saved**     | Estimated saved time valued at the organization's hourly rate    |

Time and money saved are estimates, not measured payroll savings. Automation rate measures
AI's share of work, not the share of tickets that included AI.

For each non-imported ticket with at least one tracked action, Chatarmin CX divides
AI-attributed actions by all tracked actions on that ticket. It then averages those ticket
percentages. By default, tracked impact actions are customer replies plus integration and
MCP actions.

## CSAT

CSAT uses survey scores from 1 to 5. The score is the mean of submitted responses. Response
rate is **responded ÷ sent**. CSAT widgets attribute survey throughput to the period when the
survey was sent.

## Agent leaderboard

The leaderboard can compare own and helped resolutions, median response and resolution
times, AI drafts, CSAT, and message activity where the viewer has permission.

* **Own resolution:** the resolving agent was also the assignee.
* **Helped resolution:** the resolving agent was different from the assignee.

The latest resolution cycle determines resolution attribution after a ticket reopens.

## AI drafts

Draft metrics include suggestions offered, suggestions used, and acceptance rate. Suggestion
fidelity estimates how much of the AI draft overlaps with the sent message. It uses
approximate word overlap, so it does not mean the exact text was kept.

## Exclusions

Most dashboard calculations exclude deleted tickets, side conversations, and imported
tickets.

Related: [Dashboard](/dashboard/intro).