Amplitude (Product Analytics)

Amplitude's primary charts: (1) Segmentation — event frequency over time, filterable and grouped by any user or event property (e.g., "Daily unique users who performed 'document_created' in the last 90 days, grouped by plan tier"). (2) Funnel Analysis — conversion rates between a defined sequence of events (e.g., "Signup → Account Setup → First Document → Team Invitation" with drop-off % at each step). (3) Retention — cohort-based retention heatmaps showing the percentage of users who return to perform a defined action in each subsequent week or month. (4) Pathfinder — event sequence visualization showing the most common paths users take through the product after a starting event. (5) Experiment Results — statistical analysis of A/B test results with significance and confidence interval calculations. Amplitude Govern adds data quality management for enterprise event taxonomy control.

Product Ops establishes a set of canonical "always-on" Amplitude dashboards for each product squad: an adoption dashboard (feature-level weekly active users, activation rates), a retention dashboard (rolling cohort retention curves), and an engagement dashboard (depth of usage per active user). These standardize the metrics all squads reference, preventing teams from each defining the same metric differently. Product Ops also runs feature impact analyses after each release: compare the cohort of users who used the new feature vs. those who did not (using propensity-matched cohorts to control for self-selection), measuring the retention and revenue delta. This "did Feature X actually improve retention?" analysis is impossible without a tool like Amplitude and is the primary quality check on roadmap investments.

Amplitude data quality depends entirely on instrumentation quality. Common mistakes: tracking UI interactions (button clicks) rather than user outcomes (document created — why the button was clicked); inconsistent event naming across frontend and backend (web app sends "task_created," mobile app sends "taskCreated," backend sends "create_task" — Amplitude sees three separate events for the same user action); missing context properties on events (every event should carry user_id, account_id, plan, and session_id as minimum properties to support segmentation and cohort analysis); and tracking too many events without a plan (1,000 events with no documentation and no clear analytical use case). Product Ops owns the instrumentation specification: defining each event's name, trigger condition, required properties, and analytical purpose before any engineering work begins.