Glosarium
Contents

Frontend Development › UI/UX for Engineers · also in Serving & Analytics

Product Analytics

Tracking how people use features to inform decisions.

Backend Engineer · Mid-levelFrontend Engineer · Mid-levelData Engineer · Mid-levelTech Entrepreneur · MVP · core

Also known as: product analytics, event tracking, user analytics

Product analytics instruments user behaviour — events (signed up, searched, purchased), funnels (signup → activation → retention), cohorts and properties — so product decisions rest on evidence rather than anecdote. Well-designed tracking answers “where do users struggle, and did our change help?” directly.

events: signup_completed {plan, source} → funnel: visit→signup→activate→retain
decision: ship, iterate, or revert — by the numbers

Value concentrates in a small, well-defined event taxonomy (consistent names, documented properties, versioned changes) wired to decisions — not in indiscriminate capture. Privacy constrains everything: consent, minimisation, retention limits and regional rules shape what’s collectable.

The classic mistakes:

  • Tracking everything. Indiscriminate events bloat costs, slow pages, and create noise nobody analyses — while expanding privacy exposure. Instrument decisions, not curiosity.
  • Inconsistent taxonomy. signup, sign_up, user-signed-up across teams makes analysis join-proof. One naming convention, documented, linted.
  • No consent handling. Collecting behavioural data without consent basis violates law in major jurisdictions and destroys trust. Consent-first instrumentation with graceful degradation.
  • Vanity metrics. Signups without activation, pageviews without outcomes — metrics that always look good and decide nothing. Tie events to decisions and outcomes.
  • Breaking changes unannounced. Renaming events silently breaks dashboards and experiments. Version the taxonomy; migrate consumers deliberately.
  • Sampling blindness. Aggressive sampling hides small-but-important segments (new features, rare errors). Sample rates must preserve decision-relevant slices.
  • Analysis without experiments. Observational data confounds relentlessly; causal claims need randomised tests (see A/B testing). Analytics finds candidates; experiments decide.

The practice: minimal decision-driven taxonomy, consistent naming, consent-respecting collection, funnels tied to outcomes, experiments for causality. Evidence infrastructure, maintained like production code.