Analytics — Analytics Implementation

When to Activate

• User invokes /godmode:analytics
• User says "add analytics", "track events", "set up tracking"
• User says "create a funnel", "A/B test", "implement experiments"
• User says "add Segment", "set up Amplitude", "configure Mixpanel", "add PostHog"
• User says "privacy-friendly analytics", "GDPR-compliant tracking", "cookieless analytics"
• When building new features that need usage measurement
• When /godmode:plan identifies analytics instrumentation tasks

Workflow

Step 1: Analytics Strategy Discovery

Understand what needs to be measured and why:

ANALYTICS DISCOVERY:
Project: <name and purpose>
Goals:
  - <business goal 1 — e.g., increase conversion rate>
  - <business goal 2 — e.g., reduce churn>
  - <business goal 3 — e.g., improve feature adoption>
Key questions to answer:
  - <question 1 — e.g., "Where do users drop off in onboarding?">
  - <question 2 — e.g., "Which features correlate with retention?">
  - <question 3 — e.g., "What is our activation rate?">
Platform: <web | mobile | both | server-side>
Framework: <React | Next.js | Vue | React Native | iOS | Android | backend>
Privacy requirements: <GDPR | CCPA | HIPAA | none | strict — no third-party>
Existing analytics: <none | GA4 | Segment | Amplitude | Mixpanel | custom>
Budget: <free tier | startup plan | enterprise>

If the user hasn't specified, ask: "What do you want to learn about your users? What decisions will this data inform?"

Step 2: Analytics Platform Selection

Choose the right analytics stack:

PLATFORM SELECTION:
| Platform | Best For | Privacy Model |
|--|--|--|
| Segment | Data routing hub, | Third-party, consent needed |
|  | multi-destination CDP | GDPR tools available |
| Amplitude | Product analytics, | Third-party, consent needed |
|  | behavioral cohorts, | SOC 2, GDPR compliant |
|  | journey mapping |  |
| Mixpanel | Event analytics, | Third-party, consent needed |
|  | funnel analysis, | EU data residency available |
|  | retention tracking |  |

Step 3: Event Taxonomy Design

Design a structured, consistent event naming system:

EVENT TAXONOMY:
Naming convention: <Object Action> (e.g., "Button Clicked", "Page Viewed")

Format: <object>_<action> (snake_case) or <Object> <Action> (Title Case)
Selected: <format>

EVENT CATALOG:
| Event Name | Trigger | Properties |
  LIFECYCLE EVENTS
| User Signed Up | Server | method, referral_source, |
|  |  | plan |
| User Logged In | Server | method, mfa_used |

Step 4: Event Property Standards

Define property types and validation:

PROPERTY STANDARDS:

GLOBAL PROPERTIES (sent with every event):
  user_id: string (anonymous ID or authenticated user ID)
  session_id: string (unique per session)
  timestamp: ISO 8601 datetime
  platform: "web" | "ios" | "android"
  app_version: string (semver)
  device_type: "desktop" | "tablet" | "mobile"
  browser: string (web only)
  os: string
  locale: string (BCP 47 — e.g., "en-US")
  experiment_ids: string[] (active A/B test assignments)

USER PROPERTIES (set once, updated on change):

Step 5: Implementation

Implement analytics tracking in the codebase:

Provider Setup

Configure the selected provider SDK (Segment, Amplitude, PostHog, etc.) with environment-keyed initialization. Build a unified abstraction layer (AnalyticsProvider interface with track, identify, page methods) so you swap providers without touching component code.

Step 6: Funnel Analysis Setup

Design and instrument conversion funnels:

FUNNEL DESIGN:
Funnel name: <name — e.g., "Onboarding Funnel">
Goal: <what conversion means — e.g., "User completes onboarding">

FUNNEL STEPS:
| Step | Name | Event | Expected % |
|--|--|--|--|
| 1 | Visit landing page | Page Viewed | 100% |
| 2 | Click sign up | CTA Clicked | 30-40% |
| 3 | Complete registration | User Signed Up | 60-70% |
| 4 | Start onboarding flow | Onboarding Started | 80-90% |
| 5 | Complete onboarding | Onboarding Done | 50-60% |
| 6 | First core action | Feature Used | 40-50% |
| 7 | Activation (aha moment) | Activation Done | 30-40% |

Step 7: A/B Test Instrumentation

Design and instrument experiments:

EXPERIMENT DESIGN:
Name: <experiment name>
Hypothesis: <if we change X, then Y will improve by Z%>
Primary metric: <metric to optimize — e.g., conversion rate>
Secondary metrics: <guardrail metrics to monitor — e.g., session duration, error rate>
Minimum detectable effect: <smallest meaningful change — e.g., 5% relative improvement>
Required sample size: <calculated based on MDE, baseline, significance level>
Duration: <estimated experiment duration>
Significance level: alpha = 0.05
Power: 1 - beta = 0.80

VARIANTS:
| Variant | Description | Traffic % |

Experiment Implementation

// experiments/ab-test.ts
interface Experiment {
  id: string;
  name: string;
  variants: { id: string; weight: number }[];
  isActive: boolean;

Step 8: Analytics Data Modeling

DATA MODEL (3 core tables):
  EVENTS: event_id (PK), event_name (indexed), user_id (indexed), session_id,
          timestamp (partitioned by day), properties (JSONB), context (JSONB)
  USERS:  user_id (PK), traits (JSONB), first_seen, last_seen, event_count
  SESSIONS: session_id (PK), user_id (indexed), started_at, ended_at,
            duration_sec, entry_page, exit_page, device_type, utm_*

COMMON QUERIES:
  - DAU/WAU/MAU: COUNT(DISTINCT user_id) WHERE timestamp >= <period>
  - Retention: cohort analysis grouping by first_seen week
  - Funnel: sequential event matching with time constraints
  - Feature adoption: COUNT(DISTINCT user_id) WHERE event_name = '<feature>'

Step 9: Privacy & Consent

Implement privacy-compliant analytics:

PRIVACY IMPLEMENTATION:
| Requirement | Implementation |
|--|--|
| Consent management | Cookie banner with granular |
|  | opt-in/opt-out per category |
| No tracking before consent | Analytics SDK loads only after |
|  | user grants consent |
| Data minimization | Track only necessary events, |
|  | no PII in properties |
| User data deletion | API endpoint to delete all data |
|  | for a user_id (GDPR Art. 17) |

Step 10: Validation & Delivery

Validate the analytics implementation:

ANALYTICS VALIDATION:
| Check | Status |
|--|--|
| All events in taxonomy are instrumented | PASS | FAIL |
| Event names follow naming convention | PASS | FAIL |
| Properties match documented schema | PASS | FAIL |
| No PII in any event properties | PASS | FAIL |
| Consent gate works (no tracking before consent) | PASS | FAIL |
| Funnels capture all steps correctly | PASS | FAIL |
| A/B test assignment is deterministic and sticky | PASS | FAIL |
| Data appears in analytics dashboard | PASS | FAIL |
| DNT/opt-out disables all tracking | PASS | FAIL |
| User deletion API works (GDPR compliance) | PASS | FAIL |
| Events fire on correct triggers (not duplicated) | PASS | FAIL |

ANALYTICS IMPLEMENTATION COMPLETE:

Artifacts:
- Analytics config: src/analytics/config.ts
- Event taxonomy: docs/analytics/event-taxonomy.md
- Tracking module: src/analytics/index.ts
- Provider implementations: src/analytics/providers/<provider>.ts
- Consent manager: src/consent/manager.ts
- Funnel definitions: docs/analytics/funnels.md
- A/B test configs: src/experiments/<experiment>.ts
- Data model: docs/analytics/data-model.md

Platform: <platform(s)>
Events tracked: <N> events across <M> categories
Funnels: <N> funnels defined

Commit: "analytics: <platform> — <N> events, <M> funnels, <privacy model>"

Key Behaviors

Never ask to continue. Loop autonomously until all events are instrumented and validated.

# Validate analytics implementation
npm run test:analytics
npx ts-node scripts/analytics-audit.ts --check-pii --check-taxonomy

IF event count > 100: audit for redundancy, merge similar events. WHEN PII detected in event properties: remove immediately, purge from provider. IF consent gate bypassed: block deployment, treat as P0 bug.

1. Taxonomy first, tracking second. Design catalog before code.
2. Privacy by default. No tracking before consent. No PII.
3. Measure what matters. Track business questions only.
4. Consistent naming saves hours. Enforce naming convention.
5. A/B tests need rigor. Calculate sample size before launch.
6. Abstraction layer. Unified interface over vendor SDK.
7. Debug before shipping. Verify events in staging first. On failure: revert with git reset --hard HEAD~1.

Flags & Options

Flag	Description
(none)	Full analytics design and implementation workflow
--platform <name>	Force platform: segment, amplitude, mixpanel, posthog, plausible, umami, ga4
--taxonomy	Design event taxonomy only (no implementation)

Auto-Detection

AUTO-DETECT:
1. SDKs: grep for '@segment/analytics', 'amplitude', 'mixpanel', 'posthog', 'plausible', 'umami', gtag.js
2. Framework: React/Next.js (app vs pages router), Vue/Nuxt, Mobile SDKs
3. Existing events: grep for '.track(', '.capture(', 'analytics.track', 'gtag('
4. Consent: grep for 'cookie-consent', 'cookiebot', 'onetrust'
5. Data warehouse: BigQuery, Snowflake, Redshift configs
6. Privacy: GDPR/CCPA references, EU deployment regions
7. Auto-configure: recommend platform or audit existing for gaps

Output Format

ANALYTICS IMPLEMENTATION REPORT:
  Platform: <Segment | Amplitude | PostHog | etc>
  Events tracked: <N> across <M> categories
  Funnels defined: <N>
  Experiments: <N> A/B tests instrumented
  Privacy model: <GDPR compliant | cookieless | consent-based>
  PII audit: CLEAN | <N> violations found
  Verdict: PASS | NEEDS REVISION

TSV Logging

timestamp	skill	action	platform	events	funnels	privacy_model	status

Success Criteria

Complete when ALL true:

1. Event taxonomy designed before tracking code
2. All events follow naming convention (0 violations)
3. 0 PII in event properties (verified with audit)
4. Consent gate blocks tracking until granted
5. DNT/opt-out disables all tracking
6. Funnel steps fire in correct order
7. A/B assignments deterministic and sticky
8. Analytics SDK loads async (< 50ms impact on LCP)
9. Debug/dev events filtered from production

Error Recovery

Failure	Action
Events not appearing	Check console errors, verify API key, use analytics debugger, check ad blockers/CSP.

Keep/Discard

KEEP if: improvement verified. DISCARD if: regression or no change. Revert discards immediately.

Stop Conditions

Stop when: target reached, budget exhausted, or >5 consecutive discards.