Events, persons, and why it isn't GA4

Everything in PostHog hangs off events (a thing that happened) attached to a person (who it happened to). "User signed up", "project created", "page viewed" — each is an event with properties, tied to a person (capture docs). Get those two things right and every chart in the product just works. The instrumentation chapter is about getting them right.

This sounds obvious, but it's the exact place where web analytics and product analytics part ways, and the reason "we already have Google Analytics" isn't an answer. GA4 does technically have an event model now, but the product is built to answer marketing questions — sessions, channels, campaign attribution — aggregated, sampled at scale, and reported around anonymous traffic. What it can't honestly answer is the question every chapter of this course turns on: what did this specific user do next, and do users who did X retain better than users who didn't? User-level funnels, N-day retention cohorts, and jump-to-the- replay-of-that-drop-off all require events keyed to durable persons — which is the whole design of a product analytics tool (PostHog's side-by-side with GA4 is candid about what each is for).

For a growth team, events are also the team's shared language. "Activated" stops being a vibe and becomes project_created fired within seven days of signup. When everyone names the same events the same way, the dashboard, the lifecycle journeys, and the ad conversions all speak to each other for free. The event taxonomy you set up in the instrumentation chapter is, quietly, the most important architectural decision in the whole course.