Product Analytics Audit

Goal

Audit product analytics and instrumentation quality for existing user/product flows: event taxonomy, firing points, properties, privacy, deduplication, funnel usefulness, tests, and docs. Keep only issues supported by repository evidence; do not recommend adding a new vendor unless the repo already has an explicit direction.

Inspect

Analytics wrappers, telemetry clients, event constants/types, product metrics/events maps when present, server/client tracking calls, consent/privacy gates, route/action handlers, feature flags, auth/session/tenant context, onboarding/payment/core flows, docs, tests, dashboards/config if committed, and observability conventions.

Issue classes

• Event taxonomy: duplicated event names, inconsistent properties, no source of truth, unstable IDs, missing tenant/plan/context where safe and necessary.
• Firing correctness: events fire before success, miss failure/cancel states, double-fire on rerender/retry, omit server-confirmed outcomes, or cannot distinguish user intent from system effect.
• Funnel coverage: critical steps in signup/onboarding/billing/core workflow are unmeasurable with existing patterns.
• Privacy and consent: PII/secrets logged, excessive payloads, analytics before consent where the repo models consent, unsafe diagnostic data, or cross-tenant leakage risk.
• Regression resistance: no tests/types/docs for critical events, stale event docs, fragile wrappers, or instrumentation coupled to UI implementation details.
• Operational signal: important product failures lack safe correlation with logs/metrics where the repo already has patterns.

Finding quality

• Include expected event semantics, privacy boundary, acceptance criteria, and validation direction.

Trace and hand off

Choose the product action/state the event claims to measure and follow its producer, payload identity/grain, delivery/retry, transformation and consuming metric. Compare a normal success with failure/cancel/duplicate or consent conditions relevant to that claim. Pass the first semantic mismatch and expected raw-to-consumer observation into Improve Product Analytics; a dashboard value cannot establish correct event timing.