TL;DR
- Explorations offer seven techniques: free form, funnel, path, segment overlap, user explorer, cohort and user lifetime.
- Standard properties are sampled when one query exceeds 10 million events; Analytics 360 allows up to 1 billion.
- Each user can create up to 200 explorations per property, and a property can hold up to 500 shared ones.
- Analysts can share an exploration; Viewers can open it read-only and must duplicate it to edit.
GA4 Explorations is the part of Google Analytics 4 where you build your own analysis instead of reading a fixed report. You pick the dimensions, metrics, segments and a technique (free form, funnel, path, segment overlap, cohort, user explorer or user lifetime), and GA4 queries event-level data to draw it. Use Explorations when a standard report cannot answer your question, and keep the standard reports for routine monitoring.
This guide covers what each technique does, how sampling and data thresholds change the numbers you see, when an exploration beats a standard report, and what the sharing limits mean for a team. The mechanics below come from Google Analytics Help; where the help pages do not give a limit, the article says so rather than guessing.
What GA4 Explorations actually are
Standard reports in GA4 are prebuilt tables and charts with a fixed layout. Explorations sit in a separate area of the interface, under Explore in the left navigation, and start from a blank canvas. You import dimensions and metrics into the Variables panel on the left, drag them into Tab Settings, and the canvas on the right draws the result in the visualization you choose. Google's help says Explorations give access to data and analytical techniques that are not available in the standard reports.
Each exploration can hold several tabs, so one file can show a funnel on one tab and a path analysis on another, all using the same segments and date range as a starting point. Because explorations query event-level and user-level data, they can answer questions that a summarized report never stores an answer for, such as "which pages did people view in the three steps before they submitted the contact form?"
Two account limits are worth knowing before you build a large library. Each user can create up to 200 individual explorations per property, and a property can hold up to 500 shared explorations. Those caps matter more for agencies and larger teams than for a single owner, because abandoned experiments pile up quickly.
The exploration techniques and what each one answers
Free-form exploration
Free form is the crosstab. You can show up to 5 dimensions as rows, 2 as columns and 10 metrics as values, and switch the visualization between a table, bar chart, line chart, donut chart, scatter plot or geo map. You can apply up to 4 segments at once, which makes it the default starting point for questions such as "how does revenue differ by device category and screen resolution?" Line charts also offer anomaly detection, where you can change the training period and the sensitivity threshold.
Funnel exploration
A funnel exploration shows the steps users take to complete a task and where they drop out. You can define up to 10 steps using events or dimension values, and each step can be "directly followed by" or "indirectly followed by" the one before it, optionally with a time limit. The choice between an open funnel (users may enter at any step) and a closed funnel (users must start at step one) changes the story completely, so decide it before you read any numbers. You can also show elapsed time between steps, compare up to 4 segments, break the funnel down by a dimension such as device category, and view the top 5 next actions after a step. A user is counted only for their first pass through the funnel in the date range, and a finished funnel can be saved as a funnel report for faster access later.
Path exploration
Path exploration draws the journeys between screens or events. It uses either a starting point or an ending point, never both. A forward path starts at an event or screen and shows what users did immediately after it. A backward path starts from an ending point, such as a purchase or a form submission, and shows what happened immediately before. The node type decides what each box represents: event name, page title, page path, screen name or screen class. By default each step shows the top 5 nodes; choosing More adds up to 20, and anything beyond that is grouped into an Others node. Node values can be event count or total users.
Segment overlap
Segment overlap shows how user segments relate to each other, so you can see which users belong to more than one group. It suits questions like "do the users who viewed pricing also come from the newsletter audience?" You can compare up to 3 segments at once, and you can create a new segment from any overlap you find.
User explorer
User explorer lists individual users within a segment and lets you open one user's activity. Google's help positions it for troubleshooting a specific user flow: it lists users by Device ID and shows each user's summary metrics and a timeline of their activity. Treat it as a debugging and pattern-spotting tool, not a way to build a customer file.
Cohort exploration
Cohort exploration groups users who meet an inclusion criterion (acquisition date, an event, a transaction or a conversion) and tracks how many of them meet a return criterion over daily, weekly or monthly periods. It answers retention-type questions: do people who arrived in the same week come back, and for how long?
User lifetime
User lifetime looks at behavior and value across a user's whole relationship with the property, not only one session. It is the technique for questions such as which acquisition channel produces users who generate more value over time.
Sampling in GA4 Explorations
Explorations run queries over raw event and user data, and every query has a quota. For standard Google Analytics properties the quota is 10 million events per query. For Analytics 360 properties the limit is up to 1 billion events, with an initial default of 100 million events per query for faster results and a "more detailed results" option for higher accuracy. When a query goes over the quota, GA4 analyzes a portion of the data and scales it up to give directionally accurate results.
You can tell it is happening because the data quality icon at the top of the exploration reports the sample size as a percentage. If you see a figure below 100 percent, the numbers are estimates. The same help page notes that filtering a large data set, for example by country, makes sampling more likely, and that this can occur in standard reports and in Explore.
The practical ways to reduce sampling follow from the rule: shorten the date range, remove unnecessary filters and segments, or split a long period into several shorter explorations and combine the results yourself. For recurring, high-volume analysis, exporting event data to BigQuery moves the work out of the sampled interface entirely, although it requires SQL and carries its own costs.
Data thresholds
Data thresholds are a separate mechanism from sampling. GA4 hides some rows when showing them could let someone infer the identity or sensitive details of individual users. Google's documentation says thresholds can apply to reports or explorations that include demographic data, to audiences defined by demographic attributes, and to search query information when user counts are too low. They are system-defined, you cannot adjust them, and they apply to standard reports as well as explorations.
When thresholding is active, the data quality indicator says that GA4 applied thresholding to one or more cards and will only display data that meets the minimum aggregation thresholds. Google suggests widening the date range so more users accumulate, and exporting to BigQuery as an alternative. The same help page warns that Google signals data is excluded from the BigQuery export, so event counts per user can differ from the interface.
The two problems look similar on screen, so keep them apart. Sampling gives you every row but with estimated values. Thresholding removes rows. A missing row is not evidence that nothing happened; it may mean that too few users matched.
Explorations versus standard reports
Standard reports are the better tool when the question is routine and the layout is fixed: how many users arrived this week, which channels sent them, what the top pages were. They are faster to read and easier to hand to a stakeholder. Explorations are the better tool when you need a custom combination of dimensions, a sequence of steps, a path between pages, or a view of individual users.
A workable rule is to start in a standard report and move to an exploration when you notice that you are exporting the same table again and again to rearrange it in a spreadsheet. If the same exploration becomes part of weekly reporting, think about whether a saved funnel report, a dashboard in another tool or a BigQuery table is a steadier home for it, because explorations depend on the retention window described next.
Google's help says properties retain 2 months of data by default for explorations, and you can change this under Admin, Data collection and modification, Data Retention. Standard properties can choose 2 or 14 months; longer periods are available only in Analytics 360. If your retention setting is short, an exploration over a longer date range will return data for only the retained part. Check the setting in Admin before promising a client a year-over-year path analysis.
Sharing and export limits
Explorations are private to the person who built them until they are shared. Users with at least the Analyst role can share an exploration. Anyone with the Viewer role on the property can open a shared exploration but cannot edit it; to change anything they must duplicate it and work on their own copy. Owner and viewer therefore see the same configuration, but a viewer's experiments never overwrite the original.
For taking results elsewhere, Explorations can export to Google Sheets, TSV, CSV, PDF and PDF with all tabs. For Sheets, TSV and CSV, all the data available in the selected visualization is exported, while the PDF options export what is currently displayed. Remember that the exported file is a snapshot: it will not update when the data does.
| Technique | Question it answers | Confirmed settings |
|---|---|---|
| Free form | How do metrics break down across dimensions? | 5 row dimensions, 2 column dimensions, 10 metrics, 4 segments |
| Funnel | Where do users drop out of a sequence? | Up to 10 steps, open or closed, 4 segment comparisons |
| Path | What happens before or after a given event? | Start or end point, top 5 nodes by default, up to 20 with More |
| Sharing | Who can see what? | Analyst shares, Viewer reads only, duplicate to edit |
The table above lists only settings that Google Analytics Help states explicitly. Everything else, such as cohort granularity options, is better read inside the interface for your own property, because the options change as Google updates the product.
What this means for Thai businesses
If your site has Thai and English page versions and collects leads through forms or LINE links, a path exploration using page path nodes can show whether visitors on the Thai pages reach the contact form more directly than those on the English pages, and a funnel can compare mobile and desktop drop-off. These are questions about your own data, so the answers differ for every site.
The bigger risk is a broken setup, not a missing feature. Explorations are only as good as the events feeding them. If key actions such as form submissions, phone clicks or checkout steps are not tracked as events, a funnel has nothing to count. Before spending time on advanced techniques, confirm that the events you care about fire once and carry sensible names. A GA4 migration and setup service is the usual route when a site moved over from Universal Analytics and nobody has audited the events since.
Tracking quality also connects to search work. Landing page and path questions are far easier to answer when your pages are clearly structured, which is where an SEO audit helps, and where results from SEO in Thailand campaigns can be measured by organic landing pages in a free-form table. If your data sits in several tools, data integration and consolidation is the step that makes a cross-source view possible, because GA4 Explorations only sees what GA4 collected.
Frequently asked questions (FAQ)
What is the difference between GA4 Explorations and standard reports?
Standard reports have a fixed layout for routine questions, while Explorations let you choose dimensions, metrics, segments and a technique yourself. Explorations also query event-level data, so they can show sequences and individual users, but they are subject to sampling quotas and the retention window.
How do I know if my exploration is sampled?
Check the data quality icon at the top of the exploration; it reports the sample size as a percentage when sampling applies. A standard property is sampled when a single query goes beyond 10 million events, so shortening the date range or removing filters is the first fix.
Can I share a GA4 exploration with a client?
Yes, if you have at least the Analyst role on the property and the client has access to it. People with the Viewer role can open a shared exploration but cannot edit it, and they need to duplicate it to make changes.
Why are some rows missing from my exploration?
The likely cause is data thresholding, which hides data when too few users match, especially with demographic data or low-volume search queries. Widening the date range often helps, and the data quality indicator tells you when thresholding was applied.
Which exploration technique should I start with?
Start with free form for most questions, then move to a funnel for conversion drop-off or a path exploration when you need to see what happened before or after a key event. Use cohort and user lifetime only after the basic events are tracked correctly.
Next step
If your explorations return empty funnels, odd path nodes or numbers you cannot reconcile, the cause is usually in the event setup rather than in the technique. Relevant Audience can review the tracking and rebuild the analyses on a clean base; start with the GA4 service page linked above.
