Product Analytics Pack

Ship a Product Analytics Workflow That Actually Tracks What Matters

You're building features, but your analytics are a black box. You ask "Why did retention drop?" and the dashboard says "Event X count: 42." But is that 42 the button click, the API call, or a retry from a flaky network? Your frontend team fires page_view on every route change, your backend fires it on every request, and your dashboard shows 10x traffic. Product analytics is supposed to be about using data from user interactions to drive development, optimization, and business strategies ^[5], but right now it's driving you crazy. You don't have a strategy; you have a graveyard of ad-hoc scripts, a Slack channel full of "can you check the count for Event Y?" requests, and a tracking plan that's just a Google Doc from six months ago.

The tracking plan zoo is real. You have signup, user_signup, and sign_up firing across different environments. You have user_id on some events and userId on others. Property scoping is a mess: plan_type shows up on page_view events where it makes no sense, bloating your storage and confusing your BI tool. When you try to run a funnel, you're guessing whether to filter on the event name or a property. Product management is riddled with buzzwords—growth loops, power user curves—and the sheer volume of data can paralyze decision-making if you don't have a clean taxonomy ^[3]. You're not analyzing product behavior; you're archaeology. You're digging through raw logs to reconstruct what the user actually did.

Every hour you spend fixing broken dashboards is an hour you aren't shipping. When your funnel analysis is broken because a property type changed from string to null, you're making product decisions blind. A misread funnel can cost you a feature launch or a marketing spend that drains your runway. If you can't measure retention accurately ^[4], you're guessing at churn. We've seen teams spend 15 hours a week on data wrangling instead of analysis. That's 600 hours a year. At a loaded rate of $150/hr, that's $90,000 burned on fixing tracking gaps that a schema could have caught instantly. Your stakeholders lose trust. "The dashboard is wrong" becomes the mantra. When product data doesn't reduce churn or boost feature adoption, the analytics team becomes a cost center, not a growth engine ^[7].

Imagine a fintech app launching a new onboarding flow. The PM wants to A/B test two variants ^[1]. They set up the experiment. Two weeks later, Variant B looks like it's converting 40% better. The team celebrates. Then the engineer digs into the raw logs and finds that Variant B's tracking script was firing on blur events instead of click, capturing accidental touches. The "win" was a false positive. The team rolled back the feature, lost a month of momentum, and now the PM doesn't trust the data team. This isn't hypothetical. Without a strict tracking plan and schema validation, your A/B tests are just expensive coin flips. You need a workflow where the tracking plan defines the contract before a single line of SDK code is written. If you're also setting up web analytics for marketing attribution, this pack handles the product-side events so your attribution model doesn't get poisoned by duplicate product interactions.

Once you install this pack, the chaos stops. validate-tracking.sh runs in CI. If a frontend dev tries to fire an event with a missing user_id or a property that violates the taxonomy, the build fails. No more "works on my machine" data. Your funnel queries in funnel-query.sql calculate conversion rates using window functions, handling drop-offs at each step without double-counting. You define cohorts in cohort-definition.json, and the retention matrix generator turns that into a clean CSV for your dashboard. You can finally answer "Which cohort has the highest LTV?" in minutes, not days. You integrate with PostHog or Amplitude using patterns from posthog-sdk-patterns.md, ensuring autocapture attributes don't bloat your storage. For the experiment side, pair this with our A/B Testing Framework Pack to handle statistical thresholds, guardrail metrics, and feature flags alongside your tracking data.

When you need to validate event payloads, think of this as the counterpart to a web analytics pack which handles the marketing side, ensuring your product data is as clean as your acquisition data. The event-schema.json enforces Draft 07 constraints, so your payloads are type-safe. The ab-test-config.yaml lets you define primary and guardrail metrics upfront, preventing metric hacking. You ship with confidence because the data is validated before it leaves your codebase.

What's in the Product Analytics Pack

• skill.md — Orchestrator skill that defines the product analytics workflow, maps analysis types to templates, and references all templates, scripts, validators, references, and examples.
• templates/tracking-plan.yaml — Production-grade tracking plan template enforcing Amplitude/PostHog best practices: event definitions, property schemas, taxonomy rules, and environment scoping.
• templates/event-schema.json — JSON Schema (Draft 07) for runtime validation of event payloads against the tracking plan, ensuring type safety and required property enforcement.
• templates/funnel-query.sql — Production SQL for funnel analysis (BigQuery/PostgreSQL dialect) calculating conversion rates across sequential product events with window functions.
• templates/cohort-definition.json — Structured JSON template for defining cohort segmentation logic, including seed events, retention actions, and time-window configurations.
• templates/ab-test-config.yaml — YAML configuration for A/B test setup, including experiment metadata, variant allocation, primary/guardrail metrics, and statistical thresholds.
• scripts/validate-tracking.sh — Executable bash validator that parses the tracking plan YAML, cross-references it against the JSON schema, checks for naming convention violations, and exits non-zero on failure.
• scripts/generate-retention-matrix.py — Executable Python script that ingests a CSV of user events, groups by cohort, and computes a retention matrix with configurable time windows.
• references/analytics-taxonomy.md — Canonical knowledge on event taxonomy design: naming conventions (verb_noun), property scoping, data governance, and preventing duplicate/broken metrics.
• references/posthog-sdk-patterns.md — Reference for PostHog SDK implementation patterns: autocapture attributes, feature flag tracking, manual capture, and LLM analytics integration endpoints.
• examples/worked-example.yaml — Complete, realistic e-commerce tracking plan example demonstrating taxonomy rules, property definitions, and environment-specific overrides.
• examples/retention-output.csv — Sample CSV output demonstrating the expected format and structure from the retention matrix generator for validation and testing.

Stop wrestling with broken tracking plans and guesswork. Upgrade to Pro to install the Product Analytics Pack. Ship with confidence.