Event parameters are the key-value pairs you attach to GA4 events to capture context, like page_title on a page_view or item_id on a purchase. GA4 supports up to 25 event-scoped parameters per event, and you must register them as custom dimensions in the admin to use them in reports beyond the Realtime view.
The GA4 MCP server is an implementation of the Model Context Protocol that exposes Google Analytics 4 data to LLM clients like Claude Desktop. It lets you query your GA4 property in natural language, with the LLM translating prompts into Data API calls and returning the results as structured context.
The measurement ID is the unique identifier for a GA4 data stream, formatted as G-XXXXXXXXXX. You find it in the GA4 admin under Data Streams, and it's the value you pass to the gtag config command or to the GA4 Configuration tag in GTM.
The Measurement Protocol is an HTTP API that lets you send events to GA4 directly from a server, without going through gtag.js or GTM. You authenticate with an api_secret created in the GA4 admin, then POST event payloads as JSON to the collection endpoint with your measurement_id and client_id.
A segment in GA4 is a filtered subset of your data scoped to users, sessions, or events, used inside Explorations to compare cohorts side by side. Unlike Universal Analytics, GA4 segments only exist within the Exploration where they were created and cannot be applied to standard reports.
user_id is a Google Analytics 4 parameter that ties events to a stable identifier you control, like a hashed email or database ID. Setting it lets GA4 stitch sessions and devices for the same person into a single user, which improves cross-device attribution and audience accuracy.
Migrating from Universal Analytics to GA4 means recreating your tracking from scratch in a new property, since GA4's event-based data model has no direct mapping to UA's category, action, and label structure. The standard path is to set up a parallel GA4 stream, replay your existing events as GA4 custom events, and rebuild reports as Explorations or BigQuery models.