The short answer
For CRO, Google Analytics 4 is mostly about three things:
- Events for each step of the journey you care about.
- Key events for the actions that matter most to the business.
- Funnel explorations to see where people drop off.
Everything else (dashboards, attribution, audiences) is useful, but these three answer the first CRO question: where do we lose people?
1. Track the steps as events
GA4 is built on events. Some are collected automatically. With enhanced measurement, GA4 can record page views, scrolls, outbound clicks, site search, video engagement, file downloads and form interactions without extra code (Google Analytics Help). Check they fire correctly on your site; automatic form events in particular can misfire.
For the journey itself, use Google’s recommended events where they exist (Google Analytics Help):
| Journey | Recommended events |
|---|---|
| Ecommerce | view_item, add_to_cart, begin_checkout, add_shipping_info, add_payment_info, purchase |
| Lead generation | generate_lead (plus your own form-step events) |
| Sign-up products | sign_up, login |
Recommended names unlock built-in reports and keep your data consistent. Add parameters such as value, currency and items to ecommerce events.
2. Mark the important ones as key events
Mark the events that represent business outcomes as key events: usually purchase, generate_lead or sign_up. Since 2024, GA4 uses “key events” for this, and “conversions” for key events used in Google Ads (Google Analytics Help).
Keep the list short. If every click is a key event, the reports stop meaning anything.
3. Build a funnel exploration
In Explore, choose Funnel exploration (Google Analytics Help):
- Add each step as a condition (for example the event
add_to_cart). - Choose closed (users must start at step 1) or open (users can enter at any step). Closed is usually clearer for checkout.
- Add a breakdown by device category.
- Read the biggest absolute drop first, then compare devices.
The limits to know
- Consent and blockers. Where analytics needs consent, and with ad blockers, GA4 only sees part of your traffic. Rates may be fine; totals will be lower than your backend. See consent and privacy.
- Thresholding and sampling can hide small numbers in explorations.
- Duplicate events (a purchase firing twice on a refreshed thank-you page) inflate key events.
- Check the backend. Compare GA4 purchases with real orders for a week. Write down the gap and keep it in mind.
Using GA4 with A/B tests
Send the test name and variation as an event parameter (or user property) from your testing tool, then compare variations in an exploration. Use GA4 to understand behaviour; make the test decision on the primary metric defined before launch.
What to do next
Put your events in a tracking plan so everyone uses the same names, then build your first funnel with the funnel analysis guide.