GA4 funnel analysis
Builds the funnel from the events actually collected, measures completion and drop-off at each step, splits it by device and channel, shows what users who leave do next, and compares two periods.
| Item | Value |
|---|---|
| Skill name | ga4-funnel-analysis |
| Required MCP server | Pilot MCP, https://app.pilot.braike.com/mcp |
| Writes | None: the skill only reads |
| Download | ga4-funnel-analysis.zip |
MCP tools used
| Tool | Platform |
|---|---|
google_analytics_get_account_summaries | Google Analytics 4 |
google_analytics_get_metadata | Google Analytics 4 |
google_analytics_list_key_events | Google Analytics 4 |
google_analytics_run_report | Google Analytics 4 |
google_analytics_run_funnel_report | Google Analytics 4 |
google_analytics_check_compatibility | Google Analytics 4 |
Optional MCP servers
A file storage server to keep the result, a team messaging server to share it, a note-taking server to archive it. None is needed: the skill works with Pilot MCP alone.
Example requests
- "Where do users drop off in our checkout?"
- "Show our purchase funnel by channel."
What the skill returns
The funnel step by step, the weakest step, gaps by segment, what leavers do, the change and hypotheses to check on the site.
Install
Download ga4-funnel-analysis.zip, then follow the steps for your client: install a skill. With the Pilot CLI, one command is enough:
npx @braike/pilot skills install ga4-funnel-analysisThe SKILL.md file
See the file as your assistant reads it
---
name: ga4-funnel-analysis
description: "Analyzes a conversion funnel in Google Analytics 4 through the Pilot MCP server: builds the funnel from the events the property actually collects, measures completion and drop-off at each step, splits it by device, channel or country, shows what users do after leaving a step, and compares two periods, then names the step that loses most and why. Use when the user asks where users drop off, why the checkout converts poorly, or for a funnel by device or channel. Triggers: GA4 funnel, drop-off analysis, checkout funnel, conversion funnel. Déclencheurs : entonnoir GA4, analyse des abandons, tunnel de commande, entonnoir de conversion, où les visiteurs décrochent."
compatibility: "Requires the Pilot MCP server (https://app.pilot.braike.com/mcp) connected in the client and authorized on a workspace with a Google Analytics 4 property attached."
metadata:
pilot-mcp-server: "https://app.pilot.braike.com/mcp"
pilot-mcp-tools: "google_analytics_get_account_summaries google_analytics_get_metadata google_analytics_list_key_events google_analytics_run_report google_analytics_run_funnel_report google_analytics_check_compatibility"
pilot-mcp-read-only: "true"
---
# GA4 funnel analysis
Find the step where users drop, for whom, and what they do instead. Answer in the user's language.
## Requirements
- **Required MCP server:** Pilot MCP (`https://app.pilot.braike.com/mcp`), authorized on the workspace that holds the property.
- **Optional MCP servers:** a chart server to draw the funnel. None is needed.
- **Read-only:** never change anything in GA4.
## Tools
- `google_analytics_get_account_summaries`: GA4 properties reachable.
- `google_analytics_get_metadata`: dimensions and metrics valid for the property.
- `google_analytics_list_key_events`: key events, to anchor the last step.
- `google_analytics_run_report`: event counts, to build steps from events that exist.
- `google_analytics_run_funnel_report`: completion and drop-off per step, with optional breakdown and next action.
- `google_analytics_check_compatibility`: checks a breakdown dimension before using it.
## Steps
1. Call `google_analytics_get_account_summaries` and pick the property.
2. Ask for the funnel, or propose one: for a store, `session_start` or `view_item`, `add_to_cart`, `begin_checkout`, `add_payment_info`, `purchase`; for leads, landing page view, form start, form submit, the lead key event.
3. Call `google_analytics_run_report` with `dimensions: ["eventName"]` and `metrics: ["eventCount", "totalUsers"]` over the period, and `google_analytics_list_key_events`: keep only steps whose events exist, with their exact names.
4. Call `google_analytics_run_funnel_report` with `dateRanges` and a `funnel` whose `steps` each have a `name` and a `filterExpression` on the event name. Read completion and drop-off per step.
5. Call it again with `funnelBreakdown` on `deviceCategory`, then on `sessionDefaultChannelGroup`; check a breakdown with `google_analytics_check_compatibility` or `google_analytics_get_metadata` when in doubt.
6. Call it with `funnelNextAction` on `eventName` to see what users do after the step they leave.
7. Run the funnel on the period before to see whether a step got worse.
## Output
1. **Funnel**: users and completion rate per step, and drop-off between steps.
2. **Weakest step**: where most users are lost, with the figure.
3. **By segment**: the same by device and by channel, with the gaps.
4. **After leaving**: what users do instead at the weakest step.
5. **Change**: the funnel against the period before.
6. **Hypotheses and tests**: what could explain the drop, for a person to check on the site.
## Rules
- The funnel report is an alpha GA4 feature: if it fails, say so and fall back on event counts from `google_analytics_run_report`, as an approximation.
- Event names are case-sensitive: use them as returned.
- Name the period and say whether the funnel is open (users may enter at any step) or closed.
- A segment with few users is not judged.
## Examples
- "Where do users drop off in our checkout?"
- "Construis l'entonnoir d'achat GA4 par appareil."
- "Le tunnel de commande s'est-il dégradé ce mois-ci ?"GA4 traffic report
Gives users, sessions, engagement and key events for the period against the period before and last year, by channel, source, device and country, with the daily trend and annotations.
Landing page performance
Reads sessions, engagement, key events and revenue per landing page, by channel and device, spots pages that get paid traffic without converting, and ranks the pages to fix.