GA4 funnel exploration is the report you open when traffic looks fine but revenue does not. It answers one question well: of the people who reached step one, how many made it to step two, and how long did it take them? Everything else in the Explore section is secondary to that.
Most teams build a funnel once, see a scary 91% drop between steps, and quietly stop using it. That number is usually a configuration artifact, not a user behavior problem. Getting a funnel to tell the truth takes about fifteen minutes of deliberate setup, and the payoff is a drop-off point you can actually fix this week.
Where to find it and what it replaces
Open Explore in the left navigation, then pick the Funnel exploration template. You get a blank canvas with a Variables column (dimensions, metrics, segments) and a Settings column (steps, breakdown, visualization). Drag what you need into the tab settings; nothing appears until you configure steps.
If you came from Universal Analytics, this replaces both the Goal Flow report and the Enhanced Ecommerce shopping behavior funnel — and it is far better than either. UA funnels were locked to a goal definition and could not be changed retroactively. A GA4 funnel exploration is built at query time against your stored event data, so you can rebuild a funnel from six months ago without having tagged anything in advance. That single difference is the strongest argument for the event model.
Open funnels versus closed funnels
This toggle causes more confusion than anything else in the report, and it sits quietly in the top-right of the settings panel as “Make open funnel.”
- Closed funnel (default): users only enter at step one. Someone who lands directly on your checkout page from an email never appears in the funnel at all, even though they completed steps two and three.
- Open funnel: users can enter at any step. Your step-one count stays the same, but later steps swell with people who skipped the earlier ones, so conversion rates between steps can exceed what a strict reading suggests.
Use a closed funnel when you are auditing a linear, controlled path — a multi-step form, an onboarding wizard, a single-page checkout. Use an open funnel when entry is genuinely distributed, like a product catalog where organic search drops people onto category pages, product pages, and the cart with equal frequency. Run both. If closed and open produce wildly different pictures, that gap is itself the finding: your users are not walking the path you designed.
Building a funnel that reflects reality
Steps are the whole game. GA4 funnels support up to ten steps, and each step can combine multiple conditions with AND/OR logic. A few rules that keep results honest:
- Scope each step tightly. A step defined as
page_viewwith no parameter matches every page on the site. Add a condition on page path or page location so step one means the page you actually mean. - Match on parameters, not page titles. Titles change with A/B tests and CMS edits; paths are stabler. If you use titles, use “contains” rather than “exactly matches.”
- Set the “is indirectly followed by” option where appropriate. Each step after the first lets you choose direct or indirect. Direct means the very next event must be the step event — brutally strict on a real site, where an intermediate
scrolloruser_engagementevent will break the chain. Indirect is the sane default. - Apply a time constraint only when the journey is genuinely short. The “within” setting on each step is useful for a checkout completed in one sitting, and misleading for a B2B trial that takes eleven days.
Then set the funnel scope. Session-scoped funnels ask whether the whole path happened in one visit. User-scoped funnels allow it to happen across visits within the date range. For SEO-driven journeys, user scope is almost always the right call — organic visitors research, leave, and come back.
Breakdowns, elapsed time, and next action
Three settings turn a flat chart into a diagnosis.
Breakdown splits every step by a dimension — device category, session default channel group, landing page, country. This is where most real insights live. A 40% drop at step two that turns out to be 62% on mobile and 18% on desktop is not a messaging problem; it is a layout problem. Keep breakdowns to five or six rows so the chart stays readable.
Show elapsed time adds average duration between steps. A step that converts well but takes four minutes is a friction point even though the completion rate looks healthy.
Next action shows the most common event users fired immediately after each step, including the ones who dropped out. If the top next action after your pricing page is a second page_view of the same page, people are hunting for something and not finding it.
Turning a drop-off into a segment
Right-click any node in the funnel chart. You can create a segment from the users who completed that step, or the ones who abandoned it, and then apply that segment across other explorations. Build the abandoner segment once and reuse it everywhere: which landing pages fed them, which channels, which devices, what content they read before giving up.
You can also publish that segment as an audience for Google Ads remarketing, though audiences only collect members from the moment you create them — they are not retroactive. Create the audience early even if you are not ready to advertise yet.
The limits worth knowing before you trust a number
Explorations are not the same engine as the standard reports, and the differences matter:
- Sampling. Standard reports are unsampled. Explorations query raw event data and will sample on large properties or long date ranges. Watch the icon at the top-right of the tab; if it says the report is based on a percentage of sessions, shorten the date range before you draw conclusions.
- Cardinality. High-cardinality breakdown dimensions get bundled into “(other),” which quietly hides the rows you care about.
- Data retention. Explorations can only reach back as far as your event-level retention setting. The default is two months. Change it to fourteen months in the data retention settings today — it costs nothing and it is not retroactive.
- Consent and modeling. With consent mode or heavy ad blocking, step-one counts are under-reported unevenly across steps, which distorts the ratios rather than shrinking them proportionally.
- Key events, not “conversions.” Google renamed conversions to key events in 2024. A funnel’s last step is usually a key event; the wording in the UI changed, the mechanics did not.
Connecting funnels to SEO work
Funnels are where organic traffic proves its worth or fails to. The pattern that pays: break the funnel down by landing page, filter to organic sessions, and compare step-one-to-step-two rates across your top 20 entry pages. Pages that pull traffic but convert 3x worse than the site average are usually intent mismatches — the page ranks for a research query and pushes a purchase decision too early.
One caveat that trips up every SEO: GA4 does not report the search queries behind organic sessions. That data lives in Search Console. Linking the two surfaces a limited set of Search Console fields inside GA4, but you cannot blend queries against arbitrary GA4 event data, and you will not get query-level funnel steps. Diagnose intent from Search Console, then diagnose behavior in the funnel, and hold both in your head.
If you would rather not stitch that together by hand every month, SEO Rocket connects Search Console and GA4 as ground truth alongside third-party rank estimates, saves a snapshot on each refresh with movement deltas, and writes the plain-English version of what changed. It does not build custom dashboards or replace Explore — for step-level funnel forensics, GA4’s own report is still the tool. Start with one funnel on your highest-intent path, fix the worst step, and rebuild it in a month.
