GA4 explorations crash course: funnels, paths, and cohorts

Explorations are where GA4 stops being a dashboard and becomes an analysis tool. Standard reports answer the questions Google anticipated; explorations answer yours — which steps of your funnel leak, what people do before converting, how cohorts behave over time. They are also constrained in ways that surprise people: they query event-level data, so your retention setting caps how far back they can see.

The techniques worth knowing

Free form — the flexible pivot table. Any dimensions as rows, any metrics as columns, with segments and filters. Where most analysis actually happens.

Funnel exploration — ordered steps with drop-off between each. The single most useful technique for ecommerce and lead-gen, and the one most worth setting up properly once.

Path exploration — what users did before or after a given event. Genuinely revealing, and easy to over-interpret.

Segment overlap — how three segments intersect. Occasionally the fastest way to answer a specific question.

Cohort exploration — retention and behaviour by acquisition cohort over time. Underused outside subscription businesses.

User lifetime — lifetime value and engagement per user. Useful if your user_id implementation is solid; misleading if it is not.

Funnels: get these three things right

Funnel exploration is where most of the value is, and three settings change the answer materially.

Open versus closed. A closed funnel requires users to enter at step one; an open funnel counts anyone entering at any step. Most people want closed for a purchase funnel — otherwise someone landing directly on checkout appears to have completed steps they skipped.

Direct versus indirect steps. Indirect allows other events between steps, which is realistic. Direct requires steps to be consecutive, which is almost never what you want and produces alarmingly low completion rates.

Elapsed time. Show it. Knowing that the median gap between add_to_cart and begin_checkout is four minutes tells you something the drop-off percentage alone does not.

A standard ecommerce funnel:

view_item → add_to_cart → begin_checkout → add_payment_info → purchase

Read the drop-off between each pair rather than the overall completion rate. The overall number is a summary; the step-to-step gaps are where the diagnosis is.

Two cautions. First, funnel steps depend on your events being correct — if add_to_cart fires on cart page views, your funnel is measuring something other than what it says. Second, a suspiciously large drop between two steps is more often a tracking gap than a UX catastrophe. Check the tracking before redesigning the page.

Path exploration, used carefully

Path exploration shows sequences forward from a starting point or backward from an endpoint. Backward from purchase is the more useful direction: what did converters actually do?

The trap is treating frequency as causation. The most common path to purchase on most sites includes the homepage, because most sessions include the homepage. That is not a finding.

What is useful: paths that differ between converters and non-converters. Build both and compare, rather than reading one in isolation.

Segments and the limits that catch people out

Segments are the main analytical lever — comparing converters against non-converters, mobile against desktop, new against returning.

Three limits worth knowing before you plan an analysis:

• Explorations query event-level data, so your data retention setting caps how far back they reach. On a default two-month property, a year-over-year exploration returns nothing for the older period. This is the single most common cause of "my exploration is empty".
• Sampling applies on large data sets. GA4 shows a sampling indicator; check it before drawing conclusions from a big query.
• Cardinality limits apply. High-cardinality dimensions collapse into "(other)" here as they do everywhere. Custom dimensions covers avoiding this.

If any of these bite regularly, that is the argument for the BigQuery export, where none of them apply.

Reading a funnel without fooling yourself

A funnel exploration produces a drop-off percentage at every step, and it is remarkably easy to draw the wrong conclusion from it.

A drop is not automatically a problem. Some steps are meant to lose people — a product page that loses 90% of viewers before add-to-cart is normal retail behaviour, not a broken page. Compare against your own history and against the step either side, not against an intuition about what a good number looks like.

Compare segments before you compare steps. The same funnel split by device usually reveals more than the aggregate. A checkout that converts at 4% on desktop and 0.6% on mobile is a mobile problem, and the blended 2.4% describes nobody.

Check the tracking before the UX. A step showing 95% drop-off is far more often a mis-wired event than a catastrophic interface. Verify the event fires correctly for a real journey — in the browser — before anyone schedules a redesign.

Watch for steps that cannot be skipped but appear to be. If more users reach step four than step three, your funnel is open when it should be closed, or an event is firing out of order.

A practical starting set

Build these four and you will cover most recurring questions:

1. Purchase or lead funnel, closed, indirect, with elapsed time shown.
2. Converters versus non-converters, free form, comparing engagement and content consumption.
3. Landing page performance, free form: sessions, engagement rate, conversion rate by landing page.
4. Path backward from your key event, to see what precedes conversion.

Save and share them rather than rebuilding each time — explorations are per-user by default until shared, which is why teams often have four people maintaining four copies of the same funnel.

FAQ

Why is my GA4 exploration showing no data?

Most often the date range extends beyond your event-level data retention window. Check Admin → Data settings → Data retention; if it is two months, older periods return nothing regardless of the range you select.

What is the difference between an open and a closed funnel?

A closed funnel only counts users who entered at step one. An open funnel counts users entering at any step. Closed is usually what you want for a purchase funnel.

Why do explorations and standard reports show different numbers?

Explorations query event-level data with their own scope and may be sampled; standard reports use pre-aggregated data. Differences are expected, especially at high volume or across long ranges.

Are GA4 explorations sampled?

They can be on large data sets. GA4 displays a sampling indicator — check it before drawing conclusions, and move to BigQuery if precision matters.

Can I share an exploration with my team?

Yes, but they are private by default. Share explicitly, or several people will independently rebuild the same analysis with slightly different definitions.

Explorations are only as good as the events beneath them — check yours with the free tracking audit on any URL.