GA4 Explorations: 7 Techniques and When to Use Each in 2026

GA4 Explorations is the ad-hoc analysis surface in Google Analytics 4. Seven techniques, one canvas, no SQL. Free form is the workhorse, Funnel and Path answer different questions than people assume, and the whole thing samples at 10 million events per query, which is the moment you should be moving the analysis to BigQuery instead of fighting the limits.

I run independent GA4 audits across European businesses. Most analytics teams I meet treat Explorations as the answer to every reporting question. It isn't. Standard reports are faster for scanning, BigQuery is the right tool for repeatable or large-scale analysis, and Explorations sits in the middle: brilliant for one-off analyst questions, frustrating the moment your dataset crosses the sampling threshold or your dimension cardinality blows past 500.

This guide is the version I would hand to a junior analyst on day one. What each of the seven techniques is actually for, where each one breaks, and when to walk away from Explorations entirely.

What is GA4 Explorations?

GA4 Explorations is a separate analysis canvas inside Google Analytics 4, accessible from the left sidebar under "Explore." It lets you build custom analyses using techniques the standard reports do not offer: funnels, path flows, cohort retention, segment overlaps, and user-level inspection. Instead of editing a fixed report, you drag dimensions and metrics onto a tab, slice by segments, and save the result to your property or to your personal view.

The standard GA4 reports are fast and pre-aggregated. Explorations are slower, more flexible, and limited by sampling and cardinality. They are not a replacement for the standard reports. They are a separate tool for questions the standard reports cannot answer.

The seven exploration techniques and what each one is actually for

Every Exploration starts with picking a technique. You can change technique mid-stream by adding a new tab. Pick the technique that matches the question, not the one that looks most familiar.

Free form, the workhorse for ad-hoc questions

Free form is a pivot table. You drop dimensions on rows, columns, and segments, then add metrics. It is what 80% of Explorations should be, because most questions are pivot questions. "Sessions by source by device, last 30 days, only mobile users." That is a Free form.

Free form is also the technique that causes the most cardinality problems. Add three or four dimensions with thousands of unique values each, and you hit the cardinality limit fast. The (other) row appears, your conclusions get noisy, and most analysts never notice.

Funnel exploration, drop-off analysis with one critical gotcha

Funnel exploration shows step-by-step conversion through a sequence: visit, view product, add to cart, purchase. You define the steps as event names with optional filters, and GA4 plots the percentage of users who completed each step.

The critical gotcha: Funnel steps reference event names exactly. If your developer renamed purchase to purchase_event last sprint and your funnel still says purchase, the funnel shows 0% conversion at that step. The events are firing, DebugView shows them, the standard reports count them, but your funnel is empty. Verify event names in DebugView every time you build a funnel.

Path exploration, what users do before or after a key event

Path exploration visualizes the sequence of events or pages around a key node. "What do users do after add_to_cart?" or "What pages do users visit before purchase?" Useful for understanding flow and discovering unexpected drop-off paths.

The trap: Path exploration on raw page_location generates one node per unique URL, which means every query string variation is its own node. Your view becomes unreadable in seconds. Use page_title or page_path_plus_query_string with stripped parameters, or pre-group your pages with a calculated event before path-analyzing them.

Segment overlap, finding audience intersections

Segment overlap shows how two or three segments intersect with a Venn-diagram view. "How many users are in both purchasers_30d and newsletter_subscribers_30d?" Useful for audience strategy and lookalike seeding.

Practical use is narrow. Most analysts open it once a quarter and forget it exists. Worth knowing about for audience overlap questions, not for daily reporting.

User explorer, single-user session reconstruction

User explorer lets you see one specific user's session timeline: every event they fired, in order, with parameters. Useful for debugging weird user journeys and verifying tagging on a specific test user.

Privacy note: User explorer can identify individual users via client_id or user_id. EU teams should treat it as PII-adjacent. If your DPO has not signed off on user-level reporting, do not export or share screenshots that show the user identifier. Internal debug only.

Cohort exploration, retention by acquisition cohort

Cohort exploration groups users by acquisition date (or first-event date) and shows how each cohort behaves over time: returning sessions, conversions, revenue. The classic "week-1 retention by signup week" report.

Useful for SaaS and content businesses where retention matters. For e-commerce purchase-only flows, cohort retention is usually a vanity metric. Use it where retention is the actual KPI, not because it looks impressive.

User lifetime, LTV-style aggregates

User lifetime aggregates per-user lifetime metrics: total revenue, total sessions, days since first visit. The closest GA4 gets to native LTV reporting.

Limitation: GA4's data retention cap is 14 months on standard properties (38 months on GA360). User lifetime metrics are bounded by retention. If you need true LTV across 24+ months, the answer is BigQuery with the GA4 export, period.

How to build a Free Form exploration step by step

Free form is the most-used technique. Building one well is the foundation for every other Exploration. Here is the worked example I use to onboard analysts.

The question: which traffic sources drive the highest-value purchases on mobile, last 30 days?

1. Open GA4, then Explore from the left sidebar, then Blank exploration. Pick Free form as the technique on the right panel.
2. Set the date range to "Last 30 days" in the top-left controls.
3. In the Variables column, add dimensions: Session source, Session medium, Device category. Add metrics: Sessions, Conversions, Total revenue.
4. Drag Session source and Session medium onto Rows. Drag Device category onto Columns. Drag Sessions, Conversions, and Total revenue onto Values.
5. Add a Segment: in Variables, click + next to Segments, choose "User segment," set condition to Device category = mobile. Drag the segment onto Segment in Tab settings.
6. Sort the table by Total revenue descending.

You now have a pivot of revenue by source and medium, restricted to mobile, ordered by revenue. This is the workhorse of GA4 analysis. Most ad-hoc questions reduce to "build this, change one variable."

Two things to verify before trusting the output. First, check the sampling badge in the top right of the canvas. If it says "Sampling," your numbers are estimates from a 10 million event sample, not the full dataset. Second, scan the source list for an (other) row. If it appears, you have hit the cardinality limit and some sources are bucketed away from your analysis.

GA4 Explorations sampling, quotas, and the (other) row

Sampling and cardinality are the two truths most agency tutorials skip. Both quietly corrupt your conclusions if you do not check them.

When sampling kicks in

Standard GA4 properties sample any Exploration query that exceeds 10 million events in the date range. GA360 properties bump that to 1 billion. If you are running a 90-day exploration on a property with 200K events per day, you crossed the sampling threshold ten days into your range, and every metric is an estimate.

Sampling does not break Explorations. It estimates. The estimate is usually within a few percent of the true value, but for low-volume slices (a small source, a niche device, a rare event), the relative error can be 20% or higher. Treat sampled data as directional, not authoritative.

How to check: top-right of the Exploration canvas, look for the small green or orange shield icon. Green shield means unsampled. Orange (or any non-100% indicator) means sampled. Click it for the exact sample size. Most analysts never look there. Build the habit.

Cardinality limits and the (other) bucket

GA4 caps each dimension at roughly 500 unique daily values per report. When a dimension exceeds that, GA4 silently aggregates the long tail into a row labeled (other). Your top sources stay visible, your tail disappears into a single bucket.

The classic failure: you analyze traffic by Page location, every URL with a query string is unique, you blow past 500 in a day, half your data is in (other), including the high-converting referrer that you then accidentally demote because you cannot see it.

Fix the cardinality before the analysis: use Page path instead of Page location when query strings are not the question, group sources with a calculated event when there are many, or pre-aggregate in BigQuery if cardinality is structural.

How to detect both before they corrupt your conclusions

A 30-second pre-flight check on every Exploration:

• Look at the sampling shield. Sampled? Note it before reading numbers.
• Scan the table for (other). Present? You are missing tail data.
• Compare totals against the standard report for the same date range. If they differ by more than 10%, sampling or cardinality is hiding something.

Skip those checks and you will eventually present a wrong number to a stakeholder. I have seen it three times in audits this year.

When NOT to use GA4 Explorations

Explorations is the right tool for ad-hoc analyst questions. It is the wrong tool for several common requests.

• Date ranges over 90 days at scale. Sampling will hit, and you cannot turn it off. Use BigQuery with the GA4 export.
• Sharing with non-GA4 users. Explorations require GA4 access and live inside the Explore section. Stakeholders will not log in. Build the dashboard in Looker Studio or export the data.
• Repeatable scheduled analysis. Explorations are interactive. There is no scheduled refresh, no email digest, no embed. For weekly or monthly reporting that runs itself, use Looker Studio backed by BigQuery, or scheduled queries.
• Unsampled data requirements. If the answer needs to be exact (board reporting, audit defense, fraud investigation), Explorations is the wrong place. BigQuery with the GA4 export gives you raw, unsampled events.
• Unioned data across multiple GA4 properties. Explorations are scoped to one property. Multi-property analysis lives in BigQuery, with each property exporting to a separate dataset and unioned in SQL.

The pattern: Explorations is for "I have a question right now and want an answer in 5 minutes." Anything that needs to repeat, scale, share, or stay exact is BigQuery's job. The GA4 + BigQuery setup guide covers the migration when you are ready.

Five mistakes I see in client Explorations

These are the top five findings from my last fifty GA4 audits where Explorations were in active use.

1. Path Exploration on raw page_location

Every query-string variation becomes its own node. The Path view turns into spaghetti within five seconds. Conclusions become unreadable.

Fix: use page_title if your titles are clean, or build a calculated event that strips query strings before path-analyzing.

2. Funnel steps with typo'd event names

Funnel shows 0% conversion at step 3. Three weeks of panic. Cause: developer renamed purchase to order_complete and the funnel still references the old name.

Fix: paste each event name into DebugView before saving the funnel. Confirm the event fires under that exact name in real time.

3. Free form with too many dimensions

Six dimensions on one Free form, three of them high cardinality. The (other) row is half your data. Every conclusion drawn from this exploration is wrong, but the table looks complete.

Fix: maximum two high-cardinality dimensions per Free form. If you need more, run two separate Explorations or move to BigQuery.

4. Comparing two segments without sample size check

Segment A converts at 3.2%. Segment B converts at 4.1%. Conclusion: B is better. Reality: Segment B has 240 sessions, the difference is within statistical noise, and the recommendation triggers a six-figure budget shift on a fake signal.

Fix: include sample size in every segment comparison. If either segment has fewer than ~1,000 sessions, treat the difference as directional, not conclusive.

5. Treating User Explorer as user-level reporting

User Explorer is a debugging tool. Some teams use it as a CRM substitute, exporting individual user timelines for outreach. That is a privacy violation under GDPR for any non-consented or non-essential processing, and it bypasses your own consent setup.

Fix: User Explorer is for debugging tagging issues. If you need user-level customer analysis, that is a server-side data flow with proper consent grounds, not a GA4 export.

How to share, save, and export an exploration

Three save options when you click the icon in the top right:

• Save to property. Visible to anyone with edit access on the property. Good for shared dashboards.
• Save to personal. Visible only to you. Good for in-progress drafts and analyst-specific views.
• Export. PDF, CSV, or TSV. CSV/TSV exports cap at 50,000 rows per table.

The shared-edit gotcha: if two analysts edit the same property-saved Exploration simultaneously, the last save wins and the other's changes are gone, with no version history. For collaborative analysis, save to personal first, share read-only screenshots, finalize in property when you agree on the structure.

GA4 Explorations vs Looker Studio vs BigQuery

Tool	Best for	Sampling	Sharing	Scheduling
GA4 standard reports	Quick scanning, pre-built KPIs	None on standard reports	Within GA4	None
GA4 Explorations	Ad-hoc analyst questions	At 10M events / query	GA4 users only	None
Looker Studio	Stakeholder dashboards, scheduled reports	Inherits source	Public links, embed	Email digests
BigQuery	Large data, repeatable queries, exact numbers	None (raw events)	Anyone with access	Scheduled queries

Pick the lightest tool that answers the question. Explorations is rarely the wrong choice for an analyst, often the wrong choice for a stakeholder.

FAQ

What is the difference between GA4 Explorations and standard reports?

Standard reports are pre-aggregated, fixed-layout, and unsampled. They load fast and answer KPI-tracking questions ("how many users last week"). Explorations are interactive analysis canvases for techniques the standard reports do not offer: funnels, paths, cohorts, user-level inspection. Explorations are slower, sample at 10M events per query, and are limited to GA4 users.

Why does my GA4 exploration show "(other)" rows?

You have hit the cardinality limit, roughly 500 unique values per dimension per day. GA4 silently aggregates the long tail into the (other) row to keep the report performant. Reduce dimensions, switch to a lower-cardinality alternative (Page path instead of Page location), or move the analysis to BigQuery for unbucketed data.

Why is my Funnel Exploration empty even though I see the events in DebugView?

The funnel step references an event name that does not match what your tagging actually fires. Common causes: developer renamed the event in the last sprint, GTM trigger fires under a different name, or the funnel was copied from a different property. Open DebugView, fire the action, copy the exact event name, and update the funnel step. The funnel matches by literal string.

Can I export GA4 Explorations to BigQuery?

No, not directly. Explorations are a UI layer on top of GA4 data. To get the same data in BigQuery, enable the GA4 BigQuery export (free for standard properties up to 1M events per day) and rebuild the analysis in SQL. The setup is in my GA4 + BigQuery guide.

Why does my data sample after I add another dimension?

Sampling is per query, and adding dimensions increases the data volume the query has to process. If the new query crosses 10M events for your date range, sampling kicks in. Either narrow the date range, drop a dimension, or move the analysis to BigQuery if you need the new dimension and full data.

Are GA4 Explorations free?

Yes, included with any GA4 property at no extra cost, including the standard free tier. Sampling and cardinality limits are stricter on standard than on GA360. The 14-month data retention cap applies to all Explorations on standard properties.

Get your GA4 setup verified before you trust the analysis

Explorations are only as accurate as the data they sit on. If your GA4 conversion events are double-firing or your consent setup is dropping 30% of sessions, every Exploration in the property is wrong. Before you build the next funnel or cohort, verify the data foundation.

The Free GTM Audit catches the top tracking failures across your container in about 10 minutes. If you want ongoing GA4 accuracy monitoring, the GA4 Monitoring retainer runs a monthly check on event schemas, conversion definitions, and Consent Mode v2 signal flow, with a written report.