Product analytics
Analytics that measures what people do inside your product, action by action, instead of which pages they loaded.
See it
What it is
Web analytics counts pages: visits, sessions, bounce rate. Product analytics counts behavior: who created a project, who invited a teammate, who came back on day 30. The unit of measurement moves from the pageview to the named event attached to a persistent identity, which is the only way to answer 'do the people who use this feature stick around?' That identity does not have to be an account: an anonymous device id works fine and is how most of these tools start, then gets stitched to a user id once someone logs in.
The tools in this shape (Mixpanel, Amplitude, PostHog, Heap) all give you the same four views: funnels, retention curves, cohorts, and segments, sliced by event and user properties. Reach for it once your product is something people do things inside rather than something they read. A marketing site is fine on page analytics. A dashboard, an editor, or an app is not.
Gotcha: product analytics is only as good as the events you send. A curated set of hand-written semantic events ('project_created' with the properties you meant) almost always produces cleaner analysis than anything you get for free. Autocapture, where the SDK records clicks, pageviews, and form interactions without you writing them (Heap built its whole product on it, and PostHog, Amplitude and Mixpanel all offer some form of it), is a legitimate way to collect data and it means you can query events you forgot to instrument. The tradeoff is governance: selector-based events break when the markup changes, names come from the DOM instead of from you, volume and cost climb, and somebody has to curate the useful ones into stable definitions. Most teams end up with both. Two more surprises: pricing is usually per event, so chatty instrumentation gets expensive fast, and browser-side SDKs get blocked, which is why a lot of teams end up sending key events server side.
Ask AI for it
Set up product analytics for my app (describe it below). First list the 10 to 15 events I should instrument to answer these questions: who activates, who becomes habitual, who churns, and which features correlate with retention. For each event give the name, the properties, and where in the code it fires. Then define the user properties to set on identify. Finally spec the 4 starting reports: a signup-to-activation funnel, a weekly retention curve by signup cohort, a power-user segment, and a feature adoption breakdown. Tell me which events should fire server side instead of in the browser and why.