Analyze Product Usage and Drive Growth

This skill covers the full product usage analytics stack, from event tracking architecture to advanced analysis and reporting. It implements structured event tracking with a defined taxonomy spanning acquisition, activation, engagement, retention, and monetization events:

// Standard event structure
const trackEvent = {
  event: 'feature_used',
  properties: {
    feature_name: 'dashboard_filter',
    user_id: 'user_123',
    session_id: 'session_456',
    timestamp: Date.now(),
    context: {
      page: '/dashboard',
      user_segment: 'power_user',
      plan_type: 'premium'
    }
  }
};

// Product-specific events taxonomy
const eventTaxonomy = {
  acquisition: ['signup_started', 'signup_completed', 'trial_started'],
  activation: ['first_login', 'profile_completed', 'first_action'],
  engagement: ['feature_used', 'content_viewed', 'search_performed'],
  retention: ['return_visit', 'weekly_active', 'monthly_active'],
  monetization: ['upgrade_clicked', 'payment_completed', 'subscription_renewed']
};

Metrics are structured hierarchically from strategic to tactical: North Star metrics (the primary business outcome, like Weekly Active Users or Revenue per User), primary metrics that directly drive the North Star (Feature Adoption Rate, Time to Value), secondary supporting metrics (Session Duration, Page Views), and guardrail metrics for quality assurance (Error Rate, Load Time).

Advanced patterns include a SQL-based cohort retention analysis (grouping users by first-seen month and tracking active users per subsequent period), a feature adoption scoring function that combines usage count, days used, and session spread into a 0-100 score, and a stickiness calculation (daily active users over monthly active users). It adapts RFM analysis (recency, frequency, monetary) to product usage - recency of last activity, session frequency, and engagement depth - segmenting users into quintile-based scores. For reporting it applies multi-touch attribution to the feature-discovery journey, weighting first and last touches at 40% each and splitting the remainder across middle touches, and builds predictive usage scoring - engineering features like days since last login, session duration, and feature-adoption breadth/depth into a weighted user health score for churn risk and expansion opportunity.

Use it when you need to go beyond basic pageview tracking into a structured event taxonomy, cohort-based retention analysis, feature adoption scoring, and predictive health scoring tied to business outcomes. Its implementation practices are built for scale and accuracy - schema validation before ingestion, deduplication by session/timestamp/event type, consistent user-based sampling, and explicit privacy compliance (retention policies and consent tracking) - so it's meant for a real production analytics pipeline, not a one-off spreadsheet analysis.

Input is raw product event data. Output spans event schemas and taxonomies, SQL cohort/attribution queries, Python scoring functions (feature adoption, RFM segmentation, predictive health score), and a reporting layer with automated anomaly detection, comparative context (versus prior period, cohort, or segment), drill-down views, and action recommendations based on usage patterns.

It combines a JavaScript-based event tracking layer, SQL for cohort and attribution analysis, and Python (pandas, numpy, scikit-learn's RandomForestClassifier) for adoption scoring and predictive health-score modeling, plus data-partitioning and pre-aggregation strategies (DAU/WAU/MAU tables) for dashboard performance.

Product analysts and data teams building or maturing a product usage analytics stack - from event taxonomy design through cohort retention, feature adoption scoring, and predictive churn/expansion signals.