GA4 Explorations: Complete Guide to All 5 Exploration Types.

Google Analytics 4 replaced the old Universal Analytics Custom Reports with a new tool called Explorations — and the upgrade is real. Where Custom Reports were essentially pivot tables bolted onto a fixed data model, Explorations give you a proper analysis workspace: segments, funnels, path diagrams, cohort retention, and user-level event streams. The trade-off is that Explorations are meaningfully more complex to set up and interpret. Many practitioners open the Explore section once, feel confused, and retreat to standard reports for the rest of their careers.

This guide covers all five exploration types, the settings that directly affect what data you can see in them, and the mistakes that cause analysts to pull wrong conclusions. By the end you should know which technique to reach for when a standard report stops answering your question.

What Are GA4 Explorations?

Explorations live in the left navigation under "Explore." They are a separate analysis environment from the standard reports you find under "Reports." Standard reports are pre-built, executive-friendly summaries that aggregate data for the entire property. Explorations are analyst-facing: you build the query yourself, choosing dimensions, metrics, filters, and segments. This makes them more powerful but also means you need to know what you are looking for before you open one.

There are several constraints worth understanding upfront. First, Explorations are subject to the property's user data retention setting. If your property is on the default two-month retention window, Explorations will only show two months of data. The standard reports are not affected by this setting — they draw from aggregated tables — but Explorations run against the raw event data, which obeys the retention limit. Change this to 14 months in Admin → Data Settings → Data Retention before you need the historical data, not after. Second, while Explorations are not subject to the standard report sampling that applies to large properties running Data API queries, they can be affected by thresholding. Thresholding is different from sampling: Google removes entire rows from the result when those rows could be used to identify individual users, which happens more often when Google Signals is enabled for demographic data collection. The two mechanisms are easy to conflate but have different causes and different fixes.

The 5 Exploration Types

Free-Form Exploration

Free-form is the default exploration type and the closest equivalent to the old Custom Reports. It gives you a drag-and-drop pivot table where you build rows, columns, values, and filters from the left panel. You can switch the visualization between table view and various chart types.

The most important feature free-form adds over standard reports is segments. You can create up to four segments and place them as column headers, letting you directly compare behavior between distinct groups — for example, paid traffic vs. organic vs. direct on the same rows. Without segments, most free-form explorations will produce output that looks nearly identical to what you could pull from a standard report.

Key things to know before building a free-form exploration:

• You are limited to 10 dimensions and 10 metrics per query. This is not a soft recommendation — the interface stops accepting additional fields at those limits.
• Dimension scope matters significantly. A user-scoped dimension combined with an event-scoped metric produces a different result than an event-scoped breakdown of the same metric. The scope is listed next to each dimension name in the selector.
• The date range you set inside the exploration overrides the property-level date range. Segment definitions do not inherit date ranges from the exploration — they evaluate across the full session/user lifetime depending on how you define them.
• "(other)" rows appear when the result set has more unique dimension values than GA4 can return. This is not an error — it is real data that GA4 has bucketed. See the section on common mistakes for how to handle it.

Funnel Exploration

Funnel exploration visualizes drop-off between a sequence of steps you define. Each step is an event or set of conditions, and the visualization shows how many users completed each step and what percentage continued to the next one.

The most consequential decision when building a funnel is whether to use a closed funnel or an open funnel. In a closed funnel, a user must complete the steps in the exact order you specify and cannot skip steps. If they reach step 3 without having triggered step 2, they are excluded from the funnel entirely. In an open funnel, users can enter at any step — a user who arrives directly at step 3 will count at step 3 even if they never hit steps 1 and 2. Checkout funnels should almost always be closed; content engagement funnels often make more sense as open.

Funnel explorations have a useful feature standard reports lack: you can enable elapsed time between steps. This tells you not just how many users dropped off, but how long it typically took users to move from one step to the next. For SaaS onboarding funnels, time-to-step is often as revealing as drop-off rate.

One important limitation: funnel explorations are user-scoped by default, not session-scoped. A user who completes step 1 in one session and step 2 in a later session will still count as progressing through the funnel. This is usually the right behavior for conversion funnels, but it can produce surprising numbers if you are analyzing a flow you expect to be completed in a single visit.

Path Exploration

Path exploration generates a tree graph showing the sequence of events or pages users navigate through. You can run it in two directions: a forward path starts from a specific event and shows what users did next, and a backward path ends at a specific event and shows what users were doing immediately before it.

This technique is best suited for behavioral questions: "After users trigger add_to_cart, what do they do next — do they go to checkout or back to the product listing?" and "What were users doing right before they hit the purchase confirmation page?" These questions are very difficult to answer with standard reports because standard reports aggregate across all users rather than following individual journeys.

Path exploration works best when you constrain it to a single starting or ending event rather than trying to build a full site-wide path. The tree branches very quickly, and nodes with fewer than a certain number of users get collapsed into an "(other)" category. For high-volume properties, this is manageable; for lower-traffic sites, the tree can collapse significantly at depth.

• Use the node breakdown selector (click any node) to drill into a specific branch rather than trying to read the full tree at once.
• You can switch between event name and page path as the node value, which changes the granularity of the analysis significantly.
• Session-scoped path exploration is available but resets at session boundaries — useful when you want to understand within-session behavior only.

Segment Overlap

Segment overlap renders a Venn diagram of up to three segments, showing how many users belong to each segment, and how many belong to the intersections between them. This is a fast way to understand audience composition: if you have a segment for "users from paid search," a segment for "users who added to cart," and a segment for "users from mobile," segment overlap tells you how these groups relate to each other.

The technique is most useful for audience strategy work — identifying groups that are commercially valuable but may not be well-served by your current targeting. If a significant portion of your high-intent users (added to cart, started checkout) are also in your mobile segment but your mobile experience has known performance issues, segment overlap surfaces that conflict quickly.

The limitation is that segment overlap is purely descriptive. It tells you that two groups overlap but not why, and it does not reveal behavioral sequences within the overlap. For the "why," combine segment overlap findings with a free-form exploration or funnel exploration filtered to the intersection segment.

User Exploration

User exploration lets you inspect the full event history for individual users. You start with a list of users — either pulled from a segment or identified by User ID — and click through to see every event a specific user triggered, in chronological order, with timestamps and associated parameters.

This is primarily a debugging and anomaly investigation tool. When a reported conversion looks wrong, or when you need to verify that a specific event is firing with the correct parameters for real users in production, user exploration gives you the raw timeline without needing to go through DebugView. It is also useful when investigating the journeys of users who completed a rare or valuable action — understanding the path of your first ten purchases in detail can reveal patterns that aggregate analysis misses.

User exploration requires either a User ID implementation or Google Signals to be enabled. Without one of these, the "users" shown are based on device-level client IDs, which means a single person on multiple devices appears as multiple users. This limits the usefulness of user exploration for cross-device analysis on properties that have not implemented User ID.

Cohort Exploration

Cohort exploration groups users by their acquisition date and tracks their behavior over subsequent time periods. The default view shows cohorts by week, with each cohort column representing the users who were first acquired in that week and each row representing a subsequent period (week 0, week 1, week 2, and so on).

The primary use case is retention analysis: what percentage of users who were first acquired in a given week returned and engaged in the following weeks? A well-performing product should show cohorts that retain a meaningful percentage of users over time. A high immediate drop-off after week 0 is a signal that acquisition quality is low or that the initial product experience is failing to hook users.

Cohort exploration becomes more informative when you layer it against acquisition channels. Creating a segment for users acquired from organic search and comparing their cohort retention against users acquired from paid social will often reveal significant differences in user quality that aggregate acquisition reports obscure. Cohorts are also the right tool for measuring the long-term effect of product changes: if you shipped a major onboarding improvement in a specific week, cohorts acquired before and after that week should show different retention curves.

Key Settings That Affect Explorations

Four property-level settings materially affect what you can see and how reliable the data is in Explorations.

Data retention. As noted above, this must be set to 14 months in Admin → Data Settings → Data Retention. The default is two months. This setting only affects Explorations and User Explorer — standard reports aggregate into permanent tables and are not affected. The change applies going forward only; you cannot recover data that was already purged.

Google Signals. Enabling Google Signals activates demographic and interest data collection, cross-device tracking, and remarketing capabilities. The trade-off is thresholding: when Google Signals is enabled, GA4 removes rows from exploration results when a dimension combination has too few users to protect individual privacy. Thresholded rows show as blank or are omitted entirely — you will see a yellow warning at the top of the exploration indicating that data has been thresholded. For properties where data completeness in Explorations is more important than cross-device attribution, disabling Google Signals at the property level or the data stream level removes the thresholding. This is a real trade-off with no perfect answer; the right choice depends on how much you rely on the demographic and cross-device data.

Sampling vs. thresholding. These are two different mechanisms that both reduce data completeness, and confusing them leads to wrong fixes. Sampling means GA4 did not process all events — it took a representative subset. Sampling in Explorations is less common than in the old Universal Analytics, but it still occurs on very large properties running queries with high date ranges or many dimensions. You will see a green shield icon (unsampled) or yellow/red shield indicating sampling when it occurs. Thresholding means all events were processed but specific rows were removed from the output for privacy reasons. The fix for sampling is to narrow your date range or reduce dimension cardinality. The fix for thresholding is to disable Google Signals or aggregate to a coarser dimension that yields larger row counts.

Date range limits. Explorations can query up to the length of your data retention window. If you set a date range beyond what the retention window covers, GA4 will query but return incomplete data without always surfacing a prominent warning. Always cross-check your date range against your retention setting before drawing conclusions about long-term trends in Explorations.

Common Exploration Mistakes

Using the Wrong Dimension Scope

GA4 dimensions have scope: event-scoped dimensions exist once per event, user-scoped dimensions describe properties of the user across their entire history, and session-scoped dimensions describe properties of the session. When you combine dimensions of different scopes in a single exploration, the results are sometimes meaningful and sometimes misleading depending on what question you are trying to answer.

A concrete example: "country" is a session-scoped dimension (it is determined per session based on the user's location). "First user medium" is a user-scoped dimension (it reflects how the user was first acquired, regardless of how they arrived in any specific session). If you put both in the same free-form exploration and add a metric like "purchase revenue," you are asking GA4 to reconcile a session attribute with a user attribute against a session-level metric. The result will be a single row per unique combination, but the session count behind that row is doing something non-obvious. Understanding the scope of every dimension you use is not optional — it changes what the numbers actually mean.

Not Using Segments

Explorations without segments frequently produce output that standard reports could have generated faster. The value of Explorations comes primarily from the ability to isolate and compare behavior across defined groups. An exploration that asks "what pages did users visit?" without a segment to define which users is essentially the Pages and Screens standard report with extra steps.

Before building an exploration, write down the comparison you are trying to make. "How does checkout completion differ between users who saw the product video vs. users who did not?" is an exploration question. "What was our most visited page last month?" is a standard report question. If you cannot articulate a comparison, a standard report is probably faster.

The 10-Dimension / 10-Metric Limit

Each free-form exploration query is limited to 10 dimensions and 10 metrics. This limit applies to the fields in the left panel variables section, not just what you have placed on the canvas. If you hit the limit and need to add more dimensions, you need to remove existing ones from the variables panel first.

In practice, hitting this limit is a signal that the exploration is trying to answer too many questions at once. A more effective approach is to build multiple focused explorations with fewer dimensions each, rather than one large exploration that becomes difficult to interpret.

Treating "(other)" as Noise

The "(other)" row in an exploration represents real data that GA4 could not display because the result set exceeded the display limit for that dimension. It is not missing data or an error — it is your users, compressed. In a free-form table, the "(other)" row can account for a substantial percentage of your total metrics.

The correct response to a large "(other)" row is not to ignore it but to reduce dimension cardinality so that more of the actual values fit within the display limit. Apply filters to narrow the dimension value range, use a less granular dimension (for example, landing page group instead of full page path), or split the analysis into separate explorations covering different slices of the dimension. Treating "(other)" as noise leads to analyses that systematically undercount specific groups of users — often the long-tail segments that are disproportionately important for specific business questions.

Frequently Asked Questions

What is the difference between GA4 Explorations and standard reports?

Standard reports are pre-built aggregate summaries designed for executive dashboards and quick status checks. They cannot be customized beyond applying standard filters and comparison date ranges. Explorations are a flexible analysis workspace where you define the query yourself: which dimensions, which metrics, which segments, which visualization type. Explorations run against the raw event data (subject to data retention settings), while standard reports draw from aggregated tables that persist independently of the retention window. For routine monitoring, standard reports are faster. For custom analysis — funnel measurement, segment comparison, path analysis, cohort retention — Explorations are the right tool.

How long does data stay in GA4 Explorations?

Data in Explorations is subject to the property's user data retention setting, which you configure in Admin → Data Settings → Data Retention. The default retention period is two months. The maximum is 14 months. This setting must be changed before you need the historical data — there is no way to recover data that has already been purged by the retention policy. Note that standard reports are not affected by this setting; they aggregate into permanent tables that persist regardless of the retention window. Only Explorations and User Explorer are subject to the data retention limit.

Why is my GA4 Exploration showing (other)?

The "(other)" row appears when the dimension you have selected has more unique values than GA4 can return in the result set. GA4 shows the highest-traffic dimension values and groups everything else into "(other)." This is real data — users whose dimension values did not make it into the top results. To reduce the size of the "(other)" row, apply dimension filters to narrow the scope of the analysis, use a less granular dimension (for example, grouping pages by section rather than full URL path), or segment the analysis to focus on a subset of users where fewer dimension values are relevant.

What is a segment in GA4 Explorations?

A segment is a defined subset of users, sessions, or events that you create to filter or compare data. In Explorations, segments are created in the Segments section of the left panel and can then be applied as filters or placed as comparison columns in the visualization. User segments include all users who match the defined criteria across their entire history. Session segments include only the sessions that match the criteria. Event segments include only the specific events that match. Segments are the primary way to get value out of Explorations — comparing behavior between two or more segments is the analysis pattern that Explorations are built for.

How do I share a GA4 Exploration with my team?

Explorations can be shared with other users who have access to the same GA4 property by using the share icon in the top-right corner of the exploration. Shared explorations give the recipient a copy of your exploration template — they can view and modify it, but their changes do not affect your original. You can also export exploration data directly to Google Sheets or as a CSV using the export option, which is useful for sharing with stakeholders who do not have GA4 access. Note that Explorations cannot be embedded in dashboards or Looker Studio directly — if dashboard sharing is the goal, rebuild the query in Looker Studio using the GA4 connector instead.

Why does my GA4 Funnel Exploration show different numbers than my standard reports?

There are several common causes. First, funnel explorations are user-scoped by default, while standard reports aggregate by session or event. A user who triggered a funnel step in a previous session still counts in the funnel, but may not appear in a session-level standard report breakdown. Second, standard reports may apply different sampling thresholds or date range logic. Third, if Google Signals is enabled on your property, thresholding may be removing rows from the exploration that appear in standard reports (standard reports are less affected by thresholding). Finally, the event definitions in your funnel steps may not exactly match the event definitions underlying the standard report metrics — verify that the event names and parameter values in your funnel steps match what is actually being tracked.

Can GA4 Explorations be exported?

Yes. Any exploration can be exported as a CSV or to Google Sheets using the export button (the download icon) in the top bar. The export produces the data as currently visible in the active exploration tab — the currently selected visualization type determines what gets exported. Note that the export respects the same thresholding and data limits that apply to the on-screen view; it does not give you access to data that is not shown due to thresholding. For large data exports without thresholding, the BigQuery export (configured in Admin → Integrations) is the more reliable path because it exports raw unsampled event data before any exploration-level privacy filtering is applied.

What is the difference between open and closed funnel in GA4?

In a closed funnel, users must complete each step in the exact order you specify and cannot skip steps. A user who reaches step 3 without completing step 2 is excluded from the funnel entirely. This gives you a strict measure of the intended conversion path. In an open funnel, users can enter at any step and can skip steps — a user who arrives directly at step 3 counts at step 3 regardless of whether they triggered steps 1 and 2. Open funnels generally show higher step completion numbers because they include users who entered the flow at a later point. Use closed funnels for purchase and checkout analysis where step sequence is meaningful; use open funnels for content engagement flows where users may start at different points.