Analytics implementation

Analytics Implementation

Pre-Processing (Auto-Context)

Project context, gathered before the skill runs. Values are injected inline below; in an environment that does not execute them (e.g. Gemini), run the shown commands instead.

• Portfolio: !sed -n '1,40p' PORTFOLIO.md 2>/dev/null || echo "(no PORTFOLIO.md)"
• Stack manifest: !head -40 package.json 2>/dev/null || head -40 build.gradle.kts 2>/dev/null || head -20 Podfile 2>/dev/null || echo "(none detected)"
• Recent commits: !git log --oneline -5 2>/dev/null || echo "(not a git repo)"
• Layout: !ls src/ app/ lib/ functions/ 2>/dev/null | head -25

Use this context to tailor all output to the actual project.

Instrument event tracking, funnels, and dashboards across mobile and web. Measure what matters, not everything.

Step 1: Classify the Analytics Need

Need	Output
Event taxonomy	Structured event naming + properties
Tracking plan	Full event spec for engineering implementation
Funnel analysis	Conversion funnel definition + instrumentation
Dashboard design	KPI dashboard layout + data sources
A/B test setup	Experiment design + tracking instrumentation
Attribution	Channel attribution model + UTM strategy

Step 2: Gather Context

1. Product — what app/platform are we instrumenting?
2. Analytics tool — Firebase Analytics, Mixpanel, Amplitude, PostHog, custom?
3. Key business metrics — what decisions will this data inform?
4. Current state — starting from zero or improving existing?
5. Privacy requirements — GDPR consent, CCPA opt-out, COPPA?

Step 3: Event Taxonomy Standards

Naming Convention

Format: object_action
Case:   snake_case
Tense:  past tense for completed actions

Examples:
  account_created        (not: create_account, userSignedUp)
  item_added_to_cart     (not: addToCart, cart_add)
  payment_completed      (not: pay, purchase_made)
  search_performed       (not: search, user_searched)
  screen_viewed          (not: pageView, open_screen)

Event Categories

Lifecycle:    app_opened, session_started, app_backgrounded
Auth:         account_created, login_completed, logout_completed
Navigation:   screen_viewed, tab_selected, deep_link_opened
Engagement:   feature_used, item_viewed, content_shared
Conversion:   trial_started, subscription_created, payment_completed
Error:        error_occurred, crash_detected, api_error

Property Standards

Every event includes:
  timestamp        — ISO 8601 (auto-captured by most SDKs)
  user_id          — authenticated user ID (if logged in)
  session_id       — auto-generated per session
  platform         — android / ios / web
  app_version      — semantic version

Event-specific properties:
  screen_viewed:    { screen_name: "home", previous_screen: "login" }
  item_viewed:      { item_id: "123", item_type: "product", source: "search" }
  payment_completed:{ amount: 29.99, currency: "USD", method: "card" }
  error_occurred:   { error_code: "AUTH_EXPIRED", screen: "checkout" }

Rules:
  - Property names: snake_case, no abbreviations
  - String values: lowercase (no "Home" vs "home" inconsistency)
  - Numbers: raw numbers, not strings ("amount": 29.99, not "29.99")
  - No PII in event properties (no email, name, phone, IP)

Step 4: Tracking Plan Template

## Tracking Plan — [Product Name]

| Event | Trigger | Properties | Priority |
|-------|---------|-----------|----------|
| screen_viewed | Every screen load | screen_name, previous_screen | P0 |
| account_created | After successful signup | method (email/google/apple) | P0 |
| feature_used | Core action completed | feature_name, duration_ms | P0 |
| payment_completed | Successful charge | amount, currency, plan_id | P0 |
| error_occurred | Unhandled error | error_code, screen, stack_hash | P0 |
| search_performed | Search submitted | query_length, results_count | P1 |
| item_shared | Share button tapped | item_id, share_method | P2 |

Priority: P0 = must ship with feature, P1 = next sprint, P2 = nice to have

Step 5: Funnel Definitions

Standard SaaS Funnel

1. Visitor         → screen_viewed (landing page)
2. Signed Up       → account_created
3. Activated       → [core_action]_completed (define per product)
4. Engaged         → session_started (day 2-7 return)
5. Converted       → subscription_created / payment_completed
6. Retained        → session_started (day 30+)
7. Referred        → referral_sent

E-Commerce Funnel

1. Browse          → item_viewed
2. Add to Cart     → item_added_to_cart
3. Begin Checkout  → checkout_started
4. Payment         → payment_completed
5. Repeat          → second payment_completed (within 90 days)

Define activation metric per product — the single action that predicts retention.

Step 6: Implementation Patterns

Firebase Analytics (Mobile)

// Android
analytics.logEvent("payment_completed") {
    param("amount", 29.99)
    param("currency", "USD")
    param("plan_id", "pro_monthly")
}

// iOS
Analytics.logEvent("payment_completed", parameters: [
    "amount": 29.99,
    "currency": "USD",
    "plan_id": "pro_monthly"
])

Web (Next.js)

// Wrapper function — swap provider without changing call sites
export function trackEvent(name: string, properties?: Record<string, unknown>) {
  // Firebase Analytics
  if (typeof window !== 'undefined' && window.gtag) {
    window.gtag('event', name, properties);
  }
  // Add other providers here (Mixpanel, Amplitude, etc.)
}

UTM Tracking

Required UTM params for all marketing links:
  utm_source     — where traffic comes from (google, linkedin, email)
  utm_medium     — marketing medium (cpc, social, email, referral)
  utm_campaign   — campaign name (spring_launch, black_friday)

Capture UTMs on first visit, persist in cookie/localStorage, attach to account_created event.

Step 7: Privacy & Consent

GDPR (EU users):
  - Show consent banner before any tracking
  - No analytics fired until consent given
  - Respect "decline all" — zero tracking, not degraded tracking
  - Provide data deletion on request

CCPA (California):
  - "Do Not Sell My Personal Information" link
  - Honor Global Privacy Control (GPC) signal

Implementation:
  - Consent state stored in cookie/localStorage
  - Analytics wrapper checks consent before firing
  - Firebase: Analytics.setAnalyticsCollectionEnabled(userConsented)

Automated Event Discovery

Before defining tracking plan, scan existing implementation:

1. Find existing events: Grep for tracking calls:
   • trackEvent|logEvent|analytics.track|posthog.capture|mixpanel.track
2. Find untracked interactions: Grep for click handlers, form submissions, and navigations without tracking
3. Count total events: Report how many unique event names exist

Code Generation (Required)

Generate analytics infrastructure using Write:

1. Event taxonomy: src/analytics/events.ts — TypeScript enum/const of all event names
2. Analytics wrapper: src/analytics/tracker.ts — type-safe wrapper with platform detection
3. Tracking plan: docs/tracking-plan.md — complete event documentation

Step 8: Dashboard Design

Executive Dashboard (weekly review):
  - Active users (DAU, WAU, MAU)
  - Activation rate (% of signups completing core action)
  - Conversion rate (% of active → paid)
  - Revenue (MRR, new MRR, churned MRR)
  - Retention curve (day 1, 7, 30)

Engineering Dashboard (daily monitoring):
  - Error rate by screen
  - API latency (p50, p95, p99)
  - Crash-free user rate
  - App load time
  - Feature adoption (% of users using new feature)