AP3
All topics

Evidence / defining the questions before the events

Product Data & Analytics

Design events and measures that make product behavior visible without turning people into a dashboard.

Why this matters

Product judgment improves when observation, instrumentation, and qualitative context can disagree in useful ways. The goal is not more tracking; it is enough trustworthy evidence to understand activation, return, failure, and feature adoption.

Concepts to learn

  • Event vocabulary and property design
  • Funnels, activation, retention, and adoption
  • Product and operational metrics
  • Data quality and event-version governance
  • Privacy-conscious instrumentation
  • Turning evidence into a product decision

Apply It in Casset

Casset is the working environment for this skill—not the curriculum itself.

  • Extend the canonical `app_*` event vocabulary instead of creating a competing `mini_app_*` schema; distinguish open, mount completion, mount failure, and playback outcomes.
  • Add a deliberately scoped `profile_viewed` event and define the privacy-safe properties needed to study return and mobile failure.
  • Verify which events persist only in production, then document the difference between accepted telemetry, trustworthy data, and an actual product insight.
  • Study which Casset Apps open, whether playback begins, where an App fails to mount, and which flows are abandoned—then make one product decision.

Suggested project

One trustworthy product funnel

Instrument, validate, and interpret one Profile World journey from view to App open to successful playback.

Open project notebook

Evidence of completion

These are completion conditions, not claimed achievements. Each stays planned until it links to a real artifact or verification.

  • implementationA versioned event schema using the existing analytics bridgeplanned
  • testsValidation tests for event names, properties, and sanitizationplanned
  • analyticsOne funnel or cohort result with data-quality caveatsplanned
  • observationA product decision that combines telemetry with direct observationplanned

Open questions

  • What is the smallest event vocabulary that can still explain a failed experience?
  • How should operational failure data and product-behavior data meet without becoming the same thing?

Related notes

Books, talks, and useful things

  • book
    Lean AnalyticsAlistair Croll & Benjamin Yoskovitz
  • article
    How to Measure Anything

    Keep the question closer than the chart.

Reflection journal

Instrumentation is a product model written in verbs. If the verbs are vague, the analysis will be too.