How to Build a Funnel Exploration Report in GA4
Practice4 minInnovialabs editorial team
In short
In GA4 you build a funnel in Explore with the Funnel exploration template: define each step as an event with conditions, choose an open or closed funnel, add a device or channel breakdown and look for the step with the biggest drop-off.
The short answer
Funnel exploration is a report in Explore that shows how many users went through a sequence of steps and where they left. Unlike standard reports, you define the steps yourself: any event, with parameter conditions if needed.
Before building it, make sure the events you need are actually collected. Check this in the Events report or in DebugView.
Step 1. Plan the funnel on paper
First describe the user's path in words, then map each step to an event.
Example for a services website:
| Step | Event | Condition |
|---|---|---|
| Service page view | page_view | page_location contains /services/ |
| Form start | form_start | — |
| Lead submitted | generate_lead | — |
Example for an online store: view_item → add_to_cart → begin_checkout → purchase.
A good funnel has 3-6 steps. A funnel that is too detailed blurs the picture.
Step 2. Create the exploration
- Open Explore and pick the Funnel exploration template.
- In the settings panel, click the pencil next to Steps.
- For each step, enter a name and choose an event. Add a parameter condition if needed, such as page_location or item_category.
- Choose how steps connect:
- Indirectly followed by: other actions may happen between steps (usually the right choice);
- Directly followed by: the next step must happen immediately.
- Optionally set a time constraint between steps, for example "within 30 minutes".
- Click Apply.
Step 3. Open or closed funnel
| Type | How it works | When to use |
|---|---|---|
| Closed (default) | A user counts only if they started at the first step | A strict path: checkout, sign-up |
| Open (Make open funnel) | A user can enter at any step | A path with several entry points, such as a form placed on different pages |
If many people enter an open funnel at step two or three, you have other entry points worth considering in your analysis.
Step 4. Add a breakdown
Drag a dimension into Breakdown and the funnel splits into rows:
- Device category: compare mobile and desktop. A big gap at one step often points to a layout or form problem on phones.
- Session default channel group or source / medium: see which traffic reaches the lead and which drops off early.
- Country, browser, landing page: as needed.
For trends over time, switch to Trended funnel: it shows how step completion changes day by day and helps you see the effect of changes.
You can also add segments and compare, for example, new and returning users.
Step 5. Find the drop-off points
Look at the Abandonments and Abandonment rate columns between steps. Look for a step where:
- drop-off is clearly higher than at neighboring steps;
- drop-off differs a lot between devices or channels.
Useful features:
- Next action: which events users triggered right after leaving a step.
- Elapsed time: how much time passes between steps.
- Right-click a step → create a segment from users who abandoned and study them in other explorations or use them for remarketing.
Step 6. Turn drop-offs into fixes
A drop-off number alone fixes nothing. Write a hypothesis and an action:
- Mobile drop-off at the form step → test the form on a phone, cut fields, enlarge buttons.
- Paid traffic leaves at step one → compare the ad copy with the landing page.
- Next action is often page_view on the pricing page → people look for information missing at that step; add it.
- Long time between steps → simplify the path or add a reminder.
After the changes, compare the funnel for periods before and after, keeping seasonality and ad changes in mind.
Common mistakes
- Steps built on events that are not collected everywhere or are duplicated.
- Using Directly followed by where other actions naturally happen between steps.
- Drawing conclusions from small samples or data flagged as sampled.
- Analyzing without a breakdown: the average hides a problem on a specific device.