analytics（分析）

Analytics & Instrumentation

Track decisions, not dashboards. This skill helps you set up analytics that answer the questions that matter — without drowning in data.

Core Principles

• Every metric should connect to an action you'd take differently based on the result.
• Instrument events at build time, not as an afterthought. It's 10x easier.
• More data ≠ better decisions. Focus on 5-10 key metrics per stage.
• Data quality matters more than data quantity. One clean funnel beats fifty broken ones.

Choosing an Analytics Tool

You don't need to write code to track events. Pick a tool, add the snippet, and configure in the UI.

Recommended by Stage

Recommendation for solo founders: Start with PostHog (free tier). It covers product analytics, funnels, session recordings, and feature flags — all things you'll need.

For basic traffic analytics (where visitors come from), add Plausible or Google Analytics 4 (free).

Setup

Tell AI:

Set up PostHog analytics:
1. Create a PostHog project and get the API key
2. Add the PostHog JavaScript snippet to our app (in the <head> or layout component)
3. Verify it's tracking page views by checking the PostHog dashboard
4. Configure: exclude admin/test accounts from tracking

What to Track (By Stage)

Pre-Launch (Just These 3)

1. Landing page visits (traffic source)
2. Waitlist/signup conversions
3. Where visitors drop off

0-$1K MRR (Add These)

1. Signup → activation rate (completed key action)
2. Time to activate (hours from signup to key action)
3. D7 retention (% returning after 1 week)
4. Free-to-paid conversion

$1K-$10K MRR (Add These)

1. Feature adoption (% of users using each feature weekly)
2. Cohort retention curves (track each month's signups over time)
3. Expansion revenue triggers (what behavior precedes upgrades)
4. Churn rate and reasons

Event Naming Conventions

Use a consistent schema so events make sense to anyone reading them.

Format: [Object] [Action] in past tense, Title Case.

Rules:

• Object first, then action (makes grouping in tools easier)
• Past tense (the event already happened)
• Be specific: "Signup Form Submitted" not "Form Submitted"
• Never include personal data in the event name

Event Properties

Every event should include context:

"Project Created"
  template_used: "blank" | "marketing" | "engineering"
  source: "dashboard" | "onboarding" | "api"

"Subscription Upgraded"
  from_plan: "free"
  to_plan: "pro"
  billing_period: "annual"
  trigger: "usage_limit" | "feature_gate" | "self_serve"

Tell AI:

Set up event tracking for our core funnel:
- Track these events: [list your key events from the table above]
- Include these properties with each event: [relevant properties]
- Set user-level properties on identify: plan, signup_date, role
- Verify events are firing correctly in PostHog/Mixpanel

Core Metrics Framework

The 5 Metrics That Matter

Setting Up Your Dashboard

Tell AI:

Create an analytics dashboard with these 5 charts:
1. Daily signups (line chart, last 30 days, by source)
2. Activation rate (% of signups who [completed key action], last 30 days)
3. D7 retention (% of each week's signups who returned 7 days later)
4. Free-to-paid conversion rate (last 30 days)
5. MRR (line chart, all time)

This should be the first thing I see when I open analytics.

Funnel Instrumentation

Define your core funnel and track every step:

Typical SaaS funnel:
  1. Landing Page Viewed
  2. Signup Form Viewed
  3. Signup Form Submitted
  4. Onboarding Started
  5. [Key Activation Action] Completed
  6. Second Session (returned next day/week)
  7. Upgrade Page Viewed
  8. Subscription Created

The biggest drop-off between steps = your biggest opportunity. Fix that first.

Tell AI:

Instrument our conversion funnel:
- Track these events in order: [list your funnel steps]
- Build a funnel visualization showing drop-off between each step
- Set up an alert if any step's conversion drops below [X]% week-over-week

Privacy and Compliance

Add these before you launch:

• Add analytics tool reference to your Privacy Policy
• Add cookie consent banner if required (depends on tool and geography)
• Enable "anonymize IP" in your analytics tool
• Exclude admin/test accounts from tracking
• Never track PII (email, name) in event properties without consent
• Allow users to opt out of tracking
• Respect Do Not Track (DNT) browser settings

Tell AI:

Add privacy compliance for analytics:
- Add a cookie consent banner that loads analytics only after consent
- Exclude users with DNT enabled
- Add an "opt out of tracking" toggle in account settings
- Ensure no PII is sent in event properties

Common Mistakes

If You Can't Set Up Analytics Right Now

Track these manually in a spreadsheet until you can:

Even rough manual tracking beats flying blind. Graduate to a real tool as soon as you can.