Analytics and Product Intelligence: Measuring What Matters

Setting up event tracking, cohort analysis, and funnel reporting to make data-informed product decisions without drowning in metrics.

Product analytics is the discipline of measuring what users actually do with your product, as opposed to what you imagine they do. Qualitative research tells you why; analytics tells you what and how many. Together they form the evidence base for confident product decisions. The challenge is building an analytics system that is reliable enough to trust, granular enough to be useful, and simple enough to maintain. ## Defining Your Measurement Framework Before instrumentation, define what you're trying to measure and why. The most common framework is the North Star Metric—a single metric that best captures the core value your product delivers: - Spotify: minutes of music listened per user - Airbnb: nights booked - GitHub: daily active users who push code The North Star is supported by a tree of Input Metrics—leading indicators that drive the North Star. For a SaaS product: new user activation rate, feature adoption, support ticket volume, and NPS all feed into retention and expansion revenue. Avoid vanity metrics: page views, registered users, and app downloads look impressive but don't correlate with product health. Focus on metrics that reflect genuine user engagement. ## Event Schema Design The foundation of reliable analytics is a well-designed event schema. Poor schema design creates inconsistencies that make analysis impossible and trust in data impossible. **Naming conventions**: Use consistent, hierarchical naming. `object_action` or `screen_action` patterns work well: `article_viewed`, `checkout_started`, `checkout_completed`, `user_registered`. **Standard properties**: Every event should include: `user_id` (or anonymous ID), `timestamp`, `session_id`, `platform` (web/ios/android), `app_version`, `environment` (production/staging). **Event-specific properties**: Include properties that contextualize the event: `article_viewed` should include `article_id`, `article_category`, `article_author`, and `source` (how the user arrived at the article—search, recommendation, direct). Use a tracking plan to document your schema. Amplitude Data, Avo, or even a shared Notion doc ensures that events are consistently named and implemented across web and mobile. ```typescript // Typed event tracking with TypeScript type ArticleViewedEvent = { article_id: string; article_title: string; article_category: string; source: 'search' | 'recommendation' | 'direct' | 'social'; scroll_depth: number; }; function trackArticleViewed(props: ArticleViewedEvent) { analytics.track('article_viewed', { ...props, timestamp: Date.now(), session_id: getSessionId(), }); } ``` ## Funnel Analysis Funnels measure conversion between sequential steps of a user flow. A checkout funnel might be: product page → add to cart → checkout started → shipping entered → payment entered → order completed. Funnel analysis reveals where users drop off. A 60% drop between "add to cart" and "checkout started" tells you exactly where to investigate: is the cart experience confusing? Is shipping cost revealed for the first time at checkout? Are there technical errors? Segment funnels by user cohort, acquisition source, device type, and geography to understand whether drop-off is universal or affects specific segments. ## Cohort Analysis Cohort analysis groups users by when they joined (or completed an action) and tracks their behavior over time. The most important cohort is the retention cohort: of users who signed up in week X, what percentage was still active after 1 week, 2 weeks, 4 weeks, 8 weeks? A flat retention curve (where the percentage stops declining) indicates product-market fit—a portion of users have found genuine value and continue using the product. A curve that declines to zero indicates fundamental product-market fit problems regardless of growth rate. Analyze retention by cohort to detect the impact of product changes: did the March onboarding improvement produce better 30-day retention for that cohort compared to February? ## Privacy-Compliant Analytics GDPR, CCPA, and similar regulations require explicit consent for analytics tracking in many jurisdictions. Implement consent management: - Cookie consent banner before setting any analytics cookies - Respect `doNotTrack` browser header (controversial but shows privacy intent) - Anonymize user data in analytics by hashing user IDs before sending to third-party providers - Use server-side tracking (sending events from your backend to the analytics provider) to avoid browser-side blocking by ad blockers Consider privacy-first analytics tools (Plausible, Fathom, PostHog self-hosted) that collect aggregate data without user-level tracking for contexts where user-level analytics isn't required. ## The Analytics Stack For early-stage products, a simple stack works well: - **Event collection**: Segment (routes events to multiple destinations), Rudderstack (self-hostable), or direct SDK integration - **Product analytics**: Amplitude or Mixpanel for funnel and cohort analysis - **Dashboards**: Metabase or Grafana connected to your data warehouse At scale, a data warehouse becomes the central hub: - Raw events → event collection → data warehouse (BigQuery, Snowflake, Redshift) - Transformations with dbt to build clean, business-ready models - BI tool (Looker, Metabase) for self-serve analytics - Real-time monitoring on aggregated metrics