# Analyze Funnels

> Builds conversion funnels over ordered steps. Use for user journeys, drop-off, onboarding, checkout, or multi-step flows. Returns a funnel insight.

- Skill: `altertable-ai/analyze-funnels` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add altertable-ai/analyze-funnels`
- Raw SKILL.md: https://api.skillmd.com/api/skills/altertable-ai/analyze-funnels/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: altertable-ai (https://skillmd.com/u/altertable-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/altertable-ai/analyze-funnels

---


# Analyze Funnels

## Quick Start

To analyze a funnel:
1. Clarify the user journey the user wants to measure
2. Use `list_events` and `render_insight` to validate the funnel steps, window, and ordering
3. Query the funnel data and calculate drop-off rates per step
4. Identify the biggest bottleneck and present actionable findings

## When to Use This Skill

- User asks about conversion rates, drop-offs, or completion rates
- Analyzing a multi-step process (signup, checkout, onboarding, feature adoption)
- Comparing funnel performance across segments or time periods
- Identifying where users abandon a flow

## Core Workflow

### Step 1: Define the Funnel

Ask the user (or infer from context) what journey to measure. You need:
- **Steps**: Ordered list of events from entry to conversion goal
- **Conversion window**: Maximum time allowed to complete the funnel
  - Use short windows (30 min) for session flows like checkout
  - Use medium windows (24 hr) for day-bounded flows
  - Use long windows (7+ days) for consideration flows like onboarding
- **Ordering**: Strict (exact sequence required) or Any (steps in any order)

Default to **strict ordering** unless the user specifies otherwise.

### Step 2: Preview and Create the Funnel

Use the Altertable MCP server to:
1. Call `list_events` to confirm exact event names and recent volume
2. Call `render_insight` with `kind: funnel` to validate step definitions and check the data looks correct
3. Once validated, call `create_insight` with `kind: funnel` if the user wants a saved chart
4. Call `create_discovery` if the bottleneck or finding should be reviewed or notified
5. Use the rendered or saved insight results to retrieve per-step user counts

### Step 3: Calculate Metrics

For each step transition, compute:

| Metric | Formula |
|--------|---------|
| Step conversion rate | Users at step N+1 / Users at step N |
| Step drop-off rate | 1 - Step conversion rate |
| Overall conversion rate | Users at final step / Users at step 1 |

### Step 4: Identify Bottlenecks

Find the step transition with:
- The largest absolute user drop-off
- The largest percentage drop-off
- Any unexpected pattern (e.g., later steps dropping more than early steps)

### Step 5: Present Results

Present results as a step-by-step breakdown:
- Show each step with user count, conversion rate, and drop-off rate
- Highlight the primary bottleneck
- Provide a concise recommendation tied to the bottleneck (what to investigate or improve)

Format example:
```
Step 1: Page View         - 10,000 users
Step 2: Add to Cart       -  1,200 users (12.0% conversion, 88.0% drop-off) <-- biggest drop
Step 3: Checkout Started  -    800 users (66.7% conversion, 33.3% drop-off)
Step 4: Purchase Complete -    720 users (90.0% conversion, 10.0% drop-off)

Overall conversion: 7.2% (720 / 10,000)
Bottleneck: Step 1 to Step 2 -- 88% of users drop off before adding to cart.
```

## Segmented Analysis

When comparing funnels across segments (device, traffic source, user type):
- Always compare identical step definitions and time periods
- Call out which segment has the worst conversion and at which step
- Account for sample size -- small segments can produce misleading rates

## Common Pitfalls

- **Wrong conversion window**: Too short cuts off legitimate conversions; too long inflates rates with unrelated sessions. Match the window to the expected user behavior.
- **Too many steps**: Including minor intermediate events dilutes the analysis. Keep funnels to 3-7 meaningful steps.
- **Too few steps**: Jumping from entry to conversion hides where users actually drop off.
- **Ignoring ordering**: Using "any" ordering when the flow is inherently sequential produces misleading results.
- **Comparing mismatched periods**: Ensure segments or time comparisons use the same date ranges and funnel definitions.
- **Not previewing before creating an insight**: Always preview funnel results to verify step definitions are correct before saving.

## Reference Files

- [Funnel parameters](references/funnel-parameters.md)
- [Conversion metrics](references/conversion-metrics.md)