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

# Funnels

> Put 2 to 10 steps in order and read arrival, conversion and drop-off at each, then see which step loses users, what causes it, and where dropped sessions go next.

Funnels put a conversion path in order and answer how many users reach each step, where they drop off, and why. It's the page **Behavior analytics** opens on.

## Steps

A funnel has 2 to 10 steps, executed in order from top to bottom. Select **Add step** and pick a step type:

| Step type | What it matches | What you fill in |
| - | - | - |
| **Page view** | A visit to a page | A page condition: pick a page from the list (equals that path), or match by path prefix (with subpaths) |
| **Element click** | A behavior event | Pick an existing event, or create one in place |
| **Custom event** | An action reported from code through `addAction` with that name | The event name; it must match exactly |

Steps can be **moved up**, **moved down**, **replaced** or deleted. When you pick an existing event for an **Element click** step, each event shows its hit count in the time range, which helps you judge whether the sample is large enough.

Common paths already seen on the site can seed a funnel: while a funnel is empty, the page lists frequent paths, and selecting one builds a multi-step funnel from it.

## Counting unit and conversion window

| Unit | Meaning |
| - | - |
| **Session** | The steps must be completed in order within one visit |
| **Visitor** | The steps can be completed across visits within the conversion window |

Choosing **Visitor** reveals **Conversion window**, set to **1 day** or **7 days**. The visitor unit requires sessions to carry a visitor identifier; without one the first step matches no visitors, and the page suggests switching to the session unit.

When the time range is shorter than the conversion window, some sessions haven't finished the window and the conversion rate is understated; the console says **The conversion window is longer than the time range**.

## Filters and breakdowns

The filter bar above the funnel covers six dimensions: **Environment**, **Version**, **Device type**, **Browser**, **Operating system** and **Country**.

**Break down by** splits the funnel into groups along the same dimensions, side by side. Each cell shows the group's arrival share and count at that step, and its colour shows how far it differs from the overall conversion rate. When a group's sample is too small the page marks it as indicative, and you can **Show only this group** to look closer.

## Reading the result

A set of summary metrics sits above the step table:

| Metric | Meaning |
| - | - |
| **Overall conversion** | The share of sessions (or visitors) that complete every step, comparable with **the previous period** |
| **Entered** | How many reached the first step |
| **Converted** | How many completed every step |
| **Time to convert P50 / P90** | The median and 90th-percentile time from the first step to this one |
| **Largest drop-off** | The step that lost the most units; select it to open that step's drop-off analysis |
| **Conversion trend** | The daily conversion rate. Vertical lines mark when a version first reported, which can lag the actual release for a gradual rollout; dashed days are those whose conversion window is still open, where the rate is understated |

When few sessions reach the first step, the page warns that the conversion rate can swing and is only indicative.

## Drop-off analysis

Select **View drop-off** on any step to see:

* **Contributing factors**: compare the conversion rate at this step for sessions with and without one factor — JavaScript errors, slow requests, slow loads, rage clicks, dead clicks. A difference is a correlation, not a cause.
* **Where sessions went**: the pages dropped sessions visited next, with the rest folded into "Other".
* **Example sessions**: open a specific session's replay.
* **AI interpretation**: hand the step's factors, exits and example sessions to AI SRE for the most likely cause and the next thing to check.

To compare the clicks of dropped and converted sessions on a page, see [Heatmaps](https://docs.flashduty.com/en/rum/behavior-analysis/heatmap.md).

## Saving and sharing

A funnel is saved under the application. An unnamed funnel shows **Untitled funnel**, and an edited one is marked **Unsaved**. Once saved you can copy a link to the current funnel; deleting a funnel doesn't affect the events it references.

**Create a funnel with AI** opens an AI SRE conversation where describing the start and end steps in plain language builds the funnel, and the result is marked **Funnel created by AI**. **Generate a comparison report** hands the funnel data of two breakdown groups to AI SRE for a written comparison.

## When a query fails

If **Query failed** appears, shorten the time range and retry. In a broken-down view you can **Cancel the breakdown and retry** to reach a conclusion at the overall level first.


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