Module 2.2: Analyze Data

Reference Guide

• Time to Complete: 30-40 minutes
• Prerequisites: Completed Module 2.1 or familiar with @ mentions and Composer

Start this module in Cursor: Run /start-2-2 to begin the interactive lesson.

📖 Overview

Learn how to use Cursor for the complete data-driven PM workflow: discovering problems with funnel and survey data, estimating impact with ROI models, and analyzing A/B test results.

Key takeaway: What takes 4-6 hours in spreadsheets takes 30 minutes with Cursor. Iteration is free - keep asking questions until you find the insight.

Data-Driven PM Workflow

Cursor handles the full cycle:

Phase 1: Discovery

1. Analyze funnel data to find drop-off points
2. Extract themes from survey responses
3. Synthesize findings into problem analysis

Phase 2: Impact Estimation

1. Use framework to build ROI model
2. Create three scenarios (pessimistic, realistic, optimistic)
3. Calculate expected lift and revenue impact

Phase 3: Experiment Analysis

1. Compare control vs treatment topline
2. Segment by key dimensions (company size, user type)
3. Check quality metrics (retention, engagement)
4. Review leading indicators (template usage, invites)
5. Create executive readout with recommendation

Working with CSV Files

The Cursor Advantage

CSVs won’t preview visually in Cursor, but that’s actually fine:

• Cursor reads CSV files and presents data in formatted tables
• No formulas, no pivot tables needed
• Just @ mention the file and ask for analysis

Want to see raw data? Open in Excel, Google Sheets, or VS Code. But for analysis, stay in Cursor.

Basic Pattern

@ mention your-data.csv and analyze it to find [what you're looking for]

Cursor will:

• Read all rows
• Calculate metrics
• Format results as tables
• Identify patterns
• Present insights

Common PM Analytics Tasks

The Power of Segmentation

When topline results look disappointing, dig deeper:

Iteration is Free

With spreadsheets:

• Each new segment = rebuild pivot tables (20-30 min)
• Filters require formula updates
• One view at a time

With Cursor:

• Each new segment = one more command (instant)
• “Not satisfied? Try another view”
• Keep iterating until you find the insight

Example: Revealing Hidden Wins

Topline: 45% → 48% activation (+2.6pp)

• Looks underwhelming
• Barely significant (p=0.04)

One command: “Segment by company size”

Result:

• Small teams (target market): 45% → 56% (+11.4pp, p < 0.001) ✓
• Enterprise: 45% → 42% (-3.5pp) ✗

New strategy: Ship to small teams only

This is why Cursor shines: effortless iteration reveals insights that spreadsheets hide.

Three-Scenario Impact Models

The Framework

Impact = Users Affected x Current Rate x Expected Lift x Value per Action

Always Model Three Scenarios

Why Three?

• Shows range of outcomes
• Builds confidence with leadership
• Accounts for uncertainty
• Even pessimistic shows ROI → strong case

Quality Over Quantity

Don’t stop at topline activation - check if activated users are better quality:

Metrics to Check

Retention:

• Week 1 active days (3+ = retained)
• 7-day return rate

Engagement:

• Tasks completed in Week 1
• Feature usage depth

Leading indicators:

• Template/example usage (predicts stickiness)
• Invite rate (predicts viral growth)
• Time to first value

Pattern

Filter to activated users only and compare [quality metric] between control and treatment

Quick Example: TaskFlow Activation

Scenario: 45% activation stuck for 6 months

Discovery (5 min):

• Funnel data: 60% drop between task creation and completion
• Survey themes: “Didn’t know what to create” (35%), “needed examples” (28%)

Impact Estimation (10 min):

• Three ROI scenarios: 2.6x to 17.6x over 3 years
• Solution: Guided onboarding with sample project

Experiment Analysis (15 min):

• Topline: +2.6pp (underwhelming)
• Segmented: Small teams +11.4pp, enterprise -3.5pp
• Quality: Retention +18pp, engagement 2.3x
• Recommendation: Ship to small teams only

Total time: 30 minutes vs 4-6 hours in spreadsheets

🐛 Troubleshooting

”Cursor won’t open my CSV file”

Fix: CSV files don’t preview visually in Cursor. That’s expected. Use @ mention to analyze the data - Cursor will read it and present results in tables.

”Results don’t match my Excel calculations”

Fix: Ask Cursor to show its calculation logic. Likely counting/filtering differently. Specify exactly what to include/exclude.

”I need a different view of the data”

Fix: Just ask! “Now segment by X” or “Filter to users who Y” or “Show me metric Z instead.” Iteration is instant with Cursor.

”How do I know if results are statistically significant?”

Fix: Ask Cursor to calculate significance. With 2,000+ users per cohort, lifts >3-5pp are usually significant. Request p-values for clarity.

📚 Resources

Official:

• Cursor Documentation - Official guides
• Cursor Forum - Community support

Related Modules:

• Module 2.1: Write a PRD - Learn @ mentions and Composer
• Module 2.3: Product Strategy - Apply frameworks with Cursor

🚀 What’s Next?

Module 2.2 complete:

• ✅ Analyzed funnel, survey, and experiment data in minutes
• ✅ Built three-scenario ROI models
• ✅ Segmented data to reveal hidden insights
• ✅ Created executive-ready documents

Next: Module 2.3 - Product Strategy

Develop a complete AI product strategy for TaskFlow using strategic frameworks, make hard choices with devil’s advocate challenges, and create an executive presentation. Less about execution, more about WHERE to compete and HOW to win.

Go to Module 2.3 →