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

# Analytics

> Read request, automation, knowledge, and inventory metrics for authorized teams.

## What Analytics shows

**Analytics → Dashboard** brings request handling and automation outcomes into one view.
Open **Analytics** from the left sidebar. Your organization and team roles determine the data you can read.

The Dashboard uses one reporting period and team scope for every card. The existing
**Opportunities** page is separate; Dashboard filters do not change its scope or records.

## Read operations metrics

<Steps>
  <Step title="Open Analytics → Dashboard." />

  <Step title="Choose the reporting period.">
    Open the calendar button. Select **Last 30 days**, **This quarter**, or **Since launch**, or choose dates in the calendar.
    Custom dates include both selected days and use UTC. A custom period can span up to 365 days.
  </Step>

  <Step title="Choose the team scope.">
    **All teams** includes the active teams you can read. **My team** uses your current team.
    Use the team selector to choose one authorized team. Organization administrators can also select **Unassigned**.
  </Step>

  <Step title="Read the summary and metric cards.">
    The top-left summary uses the current report values. It is not a generated recommendation.
    Use **Refresh** beside the team and period selectors to load the same period and scope again.
  </Step>
</Steps>

### Dashboard metrics

| Metric | Meaning |
| - | - |
| **AI deflection** | Share of request items resolved by Flow without human resolution |
| **AI deflection by team** | The same share, grouped by team |
| **Requests handled** | Request volume normalized to one week |
| **Automation success rate** | Completed production runs divided by production runs with a terminal outcome |
| **Human time to resolve** | Median time from ticket creation to human resolution |
| **Requests by channel** | Request count and share for each source channel |
| **How Flow shares the load** | Requests grouped by when they began and who resolved them by report end |
| **Conversations created** | Requester conversations started in the period with a user message |
| **Answered by Intelligence** | Resolved conversations without tickets that received an answer citing connected knowledge |
| **Assets discovered** | Assets first observed through connected inventory in the period |
| **Assets handled automatically** | Unavailable until asset-level automation outcomes are recorded |

A request item is a ticket or a requester conversation that had not produced a counted ticket
by report end. A later ticket does not remove an earlier conversation from its reported period.

Channel chart titles show the channel name, request count, and share.
Titles use **request** for a count of one and **requests** for other counts.

Comparisons use the immediately preceding period of equal length. **Since launch** starts at
organization creation and has no comparison. Percentage-rate changes use percentage points.
Human resolution changes use duration; a shorter duration is favorable.

## Read the numbers together

These examples use made-up numbers for one reporting period and team.

| Example | What the dashboard shows | What it means |
| - | - | - |
| Flow resolves 12 of 20 request items. People resolve 5; 3 remain unresolved. | **AI deflection: 60%** | Flow resolved 12 out of 20. Human resolutions and unresolved requests remain in the total. |
| 9 production runs complete, 1 fails, and 2 are still running. | **Automation success rate: 90%** | 9 of the 10 finished runs completed. The 2 active runs do not yet affect the rate. |
| Three requests take people 10, 30, and 80 minutes to resolve. | **Human time to resolve: 30 minutes** | The median is the middle duration. It is elapsed time, not a measure of hands-on effort. |

AI deflection includes both tickets resolved by Flow and conversations that Flow completes without
creating a ticket.

A successful automation run and a resolved request measure different outcomes. A run can finish
one part of a request while a person still has work to do. Use **How Flow shares the load** to see
who resolved requests, then inspect individual runs in [Monitoring](/automations/monitor).

## Understand data limits

**-** means a rate has no observations, a duration has no resolved sample, or a metric lacks
supported evidence. It does not mean zero. Empty periods show zero counts while unavailable rates show **-**.

Conversations without tickets use outcome labels for resolution. Older ticket outcomes or knowledge answers
may lack retained evidence, so those counts can be incomplete.

Historical connected discoveries can lack the team recorded at discovery. Team-scoped totals exclude
those discoveries instead of guessing from current asset ownership. Organization-wide administrator
totals can include them. Creating an asset manually does not record a connected discovery.

The helpfulness buttons select a local response; they do not save feedback.

## Related topics

* [Monitor automations](/automations/monitor): inspect active runs, failures, blockers, and detailed run history
* [Automations](/setup/workflows): build and publish automations
* [Help desk](/setup/help-desk): manage the conversations behind request metrics


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.