
GA4 explorations and funnel reports let you see where people drop off on the path to conversion. Instead of looking at total numbers, they show step by step where people are lost, and therefore where it is worth concentrating improvement.
The essentials in a definition
Explorations and funnels in GA4 let you analyse steps between visit and conversion, and see where people are lost. A well-defined funnel has clear steps (visit, product, cart, checkout, purchase), consistent events and comparable segments (device, source). For larger analyses, data can be exported to BigQuery. Base it on reliable events and key events.
What they are for
A funnel exploration defines steps (for example, product page visit, add to cart, checkout start, purchase) and shows how many users progress at each one. It lets you compare segments (device, source, campaign) and see differences. It helps decide whether the problem lies in the ad, the page or the checkout.
The funnel prioritises fixes.
How to build
Define steps with reliably measured events. Choose whether the funnel is open (can enter at any step) or closed. Choose the period and apply segments. Start with the main steps and add detail later. Adjust to the business: leads, subscriptions, bookings.
Clear steps, reliable measurements.

Explorations and funnels in GA4
- Define steps with reliable events
- Compare segments
- Identify the biggest drop
- Test the improvement and measure again
Read and interpret
Look for the biggest drop between two steps and investigate. Compare devices and sources: if mobile loses much more, the problem may be the experience. If a campaign has many visitors and few actions, the message may not match. Consider seasonality and insufficient data.
A big drop is a question, not an answer.

A funnel does not say why; it says where to look.
Act and measure again
Define improvement hypotheses for the critical step, test and measure again. Combine with session recordings, heatmaps and customer feedback, when available and with respect for privacy. Record results and update the funnel when the site changes.
Measure, improve, measure.
Decision table
| Step | What to measure | Problem clue |
|---|---|---|
| Product visit | Pass-through rate from entry | Weak landing pages |
| Add to cart | Pass-through from product to cart | Price, images, trust |
| Checkout start | Pass-through from cart | Unexpected shipping costs |
| Purchase | Pass-through from checkout | Payment methods, errors |
Common mistakes to avoid
- A funnel with poorly defined steps. Misleading results.
- Not segmenting. Averages hide problems by device or source.
- Drawing conclusions with low volume. Random variation looks like patterns.
How to measure the result
Track pass-through rate between steps by segment and its evolution after fixes.
Limits and exceptions
Explorations have sampling and retention limits, and depend on consent. A funnel shows where people are lost, not why. Complement with tests, recorded sessions and feedback.
Illustrative example
Illustrative example: a store builds a five-step funnel and segments by device. It finds that on mobile the drop between cart and checkout is much larger. It investigates the mobile form, fixes problems and measures pass-through rate again.
In practice: checkpoints
Start with a simple four-step funnel and compare by device and by source. Note the pass-through rate between steps and its evolution, and record hypotheses for the largest drops. Check whether each step’s events fire on all relevant pages, and use recordings and heatmaps to understand the reason. After each fix, measure again for a few weeks and keep the comparison. Add one segment at a time, such as new visitors or paid traffic, to avoid reports that are too complex. Share the funnel with the people who can change the pages, since insight only has value if somebody acts on it. Review the funnel together with customer support, who hear first about the problems visitors face.
Frequently asked questions
How many steps should it have?
As many as necessary, generally between four and six.
Open or closed funnel?
It depends. Closed shows the full sequence; open accepts entry at any step.
What after finding the drop?
Form hypotheses, test and measure again.
How many steps should I use?
Those matching real decisions, generally between four and six.
What if data are scarce?
Extend the period or combine segments, and avoid strong conclusions.
Define funnels with clear steps and reliable events, segment by device and source, and use drops to formulate fix hypotheses.
Sources and recommended reading
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.