sunny.
Back to PM Skills
Analysis & Diagnostics

Funnel Diagnosis

Diagnose conversion funnel problems and generate data-backed improvement hypotheses.

# Drop into ~/.claude/skills/funnel-diagnosis/
curl -L https://github.com/sunnyyang-hicks/pm-skills-for-claude/raw/main/funnel-diagnosis/SKILL.md \
  -o ~/.claude/skills/funnel-diagnosis/SKILL.md

Overview

You diagnose where and why users drop off in multi-step flows, then generate testable hypotheses for improvement.

Before You Start

Ask the user:

  1. What funnel? — Signup, onboarding, checkout, upgrade, feature adoption.
  2. Current data — Step-by-step conversion rates, or raw numbers.
  3. Time period — Recent snapshot or trend over time.
  4. Segments — Any known differences by user type, channel, device?

Diagnosis Process

Step 1: Map the Funnel

Step 1: [action] → [X users] (100%)
    ↓ [Y% drop-off]
Step 2: [action] → [X users] ([Z]%)
    ↓ [Y% drop-off]
Step 3: [action] → [X users] ([Z]%)
    ↓ [Y% drop-off]
Step 4: [action] → [X users] ([Z]%)

Step 2: Identify the Biggest Drop

Calculate both absolute and relative drop-off per step. The biggest optimization opportunity isn't always the biggest percentage drop — it's the step where the most recoverable users are lost.

| Step | Users In | Users Out | Drop-off | Recoverable? | |------|---------|----------|----------|--------------| | [step] | [N] | [N] | [%] | High/Med/Low |

Recoverable = users who show intent but don't complete. Low recoverability = users who were never the right audience.

Step 3: Generate Hypotheses

For each major drop-off, systematically consider:

Motivation problems: Users don't want to complete this step.

  • Value proposition unclear at this point
  • Asked for too much commitment too early
  • Competing alternatives are easier

Ability problems: Users can't complete this step.

  • UX is confusing or broken
  • Too many fields / too much information required
  • Technical barriers (load time, browser compatibility, mobile issues)

Trigger problems: Users lose momentum.

  • No clear CTA or next step
  • Distraction or interruption
  • Too much time between steps

Step 4: Prioritize Hypotheses

| Hypothesis | Evidence | Impact if True | Ease to Test | Priority | |-----------|----------|---------------|-------------|----------| | [hypothesis] | [what supports this] | [potential lift] | [effort] | [1-N] |

Step 5: Recommend Actions

For each top hypothesis:

  • Quick test: Fastest way to validate (5-user test, copy change, A/B test)
  • Full solution: What to build if the hypothesis is confirmed
  • Expected impact: Modeled lift based on current data

Output

# Funnel Diagnosis — [Funnel Name]

## Current State
[Funnel visualization with conversion rates]

## Key Finding
[The single biggest opportunity, in one sentence]

## Step-by-Step Analysis
### [Step with biggest drop-off]
**Drop-off:** [X%] ([N] users lost)
**Hypotheses:**
1. [Hypothesis] — Evidence: [data] — Priority: High
2. [Hypothesis] — Evidence: [data] — Priority: Medium

[Repeat for each significant drop-off]

## Recommended Action Plan
1. [First thing to test/fix] — Expected impact: [X%] — Effort: [S/M/L]
2. [Second] — Expected impact: [X%] — Effort: [S/M/L]
3. [Third] — Expected impact: [X%] — Effort: [S/M/L]

## Benchmarks
[How does this funnel compare to industry standards?]

Save as FUNNEL-DIAGNOSIS-[name]-[date].md.