[GA4] Explore playbook

In this article:

Which pages on my site/app drive the most views?

In Universal Analytics, we’re used to seeing the Pages report detailed out by the dimension of “Page Path” and metrics such as “Pageviews”, “Unique Pageviews”, “Bounce Rate”, and more. In Google Analytics 4 properties, the pages report looks different, but you can still easily recreate what you are looking for as an exploration in Explore.

First, you will need to enable an additional dimension via the + icon for adding dimensions on the Tab Settings pane in Explore. Specifically, you’ll add the “Page Path + Query String” dimension. If you also want to check your data by page title (in UA, the Pages report uses “Page Path”), you can enable an additional dimension of “Page Title & Screen Name”.

Next, you’ll need to add the metric for “Views” so that you can bring it into your exploration. To do so, click the + icon for the Metrics section in the Variables pane, and search for the “Views” metric to add.

To build the pages exploration, you’ll then need to remove the default metrics and dimensions applied in the rows and columns settings and add “Page path + query string” as the row dimension and “Views” as the column metric. For comparison purposes, you can also add “Active Users” as a column metric to give you a sense of popularity of a page by a unique user.

For the Google Merchandise store, if you remove the home page (/) and the basket page, you’ll quickly check that the most popular product pages are the Clearance section and the Men’s Apparel section of the website.

Which landing pages perform the best or drive the most conversions?

Understanding where users first land on your app or website can help you to optimize the user experience and your marketing efforts. In Universal Analytics this is an out-of-the-box report in the reporting UI, however, in Google Analytics 4 properties it does not yet exist. We can recreate this as an exploration from scratch in Explore.

There are a couple of important things you will need to build this exploration: The "Landing page + query string" dimension and relevant metrics such as "Views", "Sessions", and "Users". You will need to enable the additional dimensions & metrics via the + icon for adding dimensions:

Once you have these dimensions and metrics available for use, you will start with adding the "Landing page + query string" dimension as a row, and relevant metrics such as "Views", "Sessions", and "Active Users" as values. This will give you an exploration of your landing pages by these metrics.

By adding in an additional metric of "Purchases", you can view which landing pages were viewed when there was a purchase:

Now you can check which landing pages were viewed when there was a purchase, and can compare to the totals overall to better understand landing page effectiveness.

What are people searching for on my site/app? Where could I improve my site content or navigation?

Another common report that you may want to create in Explore is a Site Search exploration. There are a couple steps to do this.

First, you must have either the Enhanced Measurement event or a custom event enabled to collect your site search data. If you’re using Enhanced Measurement, this event will be called “view_search_results”. Next you’ll need to ensure that you’ve enabled the parameter for “search_term” as a Custom Dimension in your property. Once you do, it will be available to use in Explore.

Once you’ve got the right data available, you can build your exploration. This will consist of a couple pieces. First, you’ll need to enable the dimension for “search_term” to be available to use in your Exploration:

Once this is enabled, you can add it to your exploration under the “Rows” setting. You will then observe an output similar to this:

This exploration is showing a lot of (not set) values because it is looking across all events. To reduce the noise, you’ll need to build a filter for “Event name” exactly matches “view_search_results” in order to only show data for that event:

Once your filter is applied, you will have an updated exploration showing you the number of searches for each term listed, hence, a site search term exploration:

Knowing what users are searching for on your website is a great way to help you optimize your content for higher user engagement and user satisfaction. For example, if you notice a spike in a particular search term, you may want to consider adding more content around that term to help your site users more easily find what they are looking for or answer their questions.

How do my landing pages perform over time? Do users land on various pages at the same rate?

For every exploration, you have several visualization options. One that may be useful is the line chart to observe how data is trending over time. If I want to check my landing pages trend over time, all I need to do is change the visualization type by clicking on one of the 6 visualization options under the Tab Settings pane.

Changing the visual to a line chart will then create a graphic of the top 10 landing pages trended over time.

If you hover over the chart, you will notice the data points listed for each day.

You may also notice anomaly data highlighted by an empty circle. If you hover your cursor over the empty circle, the anomaly information will show. In the below example, based on the previous data collected, Google Analytics was anticipating around 1,000 active users to visit the home page on February 2, however, the observed number of active users was 2,200 which was 116% higher than expected.

Funnels are a great tool to easily visualize how users are progressing through a set of steps you imagine them taking. Common funnel use cases might be a shopping behavior flow or checkout behavior flow for ecommerce businesses. For game developers, you might want to check how users are progressing through levels in your game. Let’s look at specific examples of each of these.

How are users progressing through the shopping funnel?

For shopping behavior, you might want to analyze how different segments of users are progressing through a shopping behavior funnel of viewing an item, adding to cart, and purchasing. To get started, you can build this funnel from scratch, or, you can use one of the convenient templates available through the Google Merchandise Store GA4 demo account. To choose this, go to the Exploration hub, scroll down, and select “Shopping Behavior Funnel” from the listed “Demo” explorations.

When this exploration opens, it will be read-only, but you can click to “Make a copy” in the upper right hand corner to make a copy of this demo exploration that will then be owned by you. Once you have a copy, you can edit and adjust any way you want.

The Google Merchandise Store sells Google branded gear, and 2 categories of gear that are popular on the site are items branded with Android and items branded with Youtube. Comparing these 2 item categories could be useful in understanding shopping behavior. To do so, you’ll want to create 2 new segments, one for each category. In this case, you can create a segment with the condition with the event “select_item” and the parameter that “item_name” contains either “Android” or “Youtube”.

These segments will then show users who progressed through the shopping behavior funnel who at least clicked on either an Android branded product or a Youtube branded product. Interestingly, it appears that users who at least clicked on an Android branded product were nearly 2 times more likely to purchase something than those who interacted with a Youtube branded product. Note this doesn’t mean they purchased a branded product, just that they clicked on it. If you were interested in only purchases of specific branded gear, you would use the purchase event instead, like below:

Once these segments have been applied, your Shopping Behavior funnel comparison will look like this:

A new feature to funnels in GA4 is the ability to look at this funnel trended over time. To do so, change the visualization type to “Trended funnel” on the Tab Settings pane. This will open up a trended funnel view show all steps trended over time. You can scroll over them to check counts by day for each step and segment.

You can also view just one step at a time trended in order to more easily visualize changes over time. In this example, you can notice that as of March 10th there was a spike in purchases that wasn’t apparent when looking at the previous view of the data. Interestingly, the difference in purchases from those viewing Android vs Youtube appears to happen almost entirely after this spike, indicating that there may have been a change in the website layout or marketing efforts making Android products more prominent from that time forward.

How are users progressing through levels in my game?

This is a common question a developer may have about their game or app. You can analyze this question with a funnel.

First, you’ll need to build steps specific to level progress. In this example, you can use the “level_up” event and build out 5 steps, for 5 levels in the Flood-it! GA4 demo account.

Applying these steps will then show you how users progressed through each game level.

It looks like users who get past level 1 are more likely to continue playing. This information could help you A/B test scenarios where you introduce tips or other methods to help users complete level 1 in order to reduce churn.

If you trend the above funnel over time, you’ll also notice an uptick across all steps starting around March 19, indicating that something may have changed with promotion or visibility of the game at that point.

You can also segment this funnel to check if different acquisition channels impact overall level progression. In this example, you can look at Direct vs Paid traffic acquisition. You’ll notice that users who were acquired via paid campaigns actually persist longer (lower abandonment rates) than those acquired directly, a sign that your paid advertising may be working as intended.

Another very useful feature of the Funnel exploration is that you can create a segment or an audience from the drop off. In the below example, you’ll notice that abandonment rate begins to increase around level 4. You may want to create a segment of those who dropped off at Level 4 (did not make it to level 5) and use that as an audience to extend your reach or to send a push notification to to encourage users to return to finish the level and continue on in the game.

To create that segment, right click on the step you are interested in and choose to create a segment. The segment you’ll want to use here will also include an exclusion criteria for level 5 to ensure that you are targeting the right audience.

Finally, one last thing you may want to add in to your funnel exploration is the new metric of “Elapsed Time”. You can do this by toggling the button for “Show Elapsed Time” on the Settings pane. Doing so will add a new metric column to your table:

Seeing how long it takes, on average, to go from step to step can be a great indicator of your customer success, or in this case, how easy or difficult a level may be to progress through. This could also be a good case for creating an audience to action on, such as sending a push notification or a remarketing message. For example, it looks like it takes 3 hrs 53 min on average to complete level 2. You could create an audience of the abandonment of this by right clicking and choosing that option:

This would then be a great audience to use to target some sort of encouragement towards so that they don’t end up churning from the game or app.

A great new feature of Explore in GA4 is the Pathing exploration. You can now choose a specific event or page/screen you want to path forwards or backwards from to observe how users are progressing through your site or app. Understanding user flow through a site has been a long time feature request, and with the new Pathing explorations in GA4 it’s finally possible.

What drove a user to add a product to cart?

This is a question many ecommerce sites might ask and can now be investigated using the backward pathing feature.

To start, open a new Pathing exploration and hit “Start Over” in the upper right hand corner. You will then be able to select an endpoint. Choose “Event name”

You’ll then choose the event you want to path backwards from in the slide out menu. In this case, since we want to check what led users to add to cart, we’ll choose “add to cart”. If you don’t notice the event name you are interested in, you can hit “Load more” or just use the search bar at the top to search for it.

My exploration will load with a couple of steps populated already based on event name. If I want to instead path by page name, I can change that per step. In the below example, I’ve changed to “Page title and screen class” and I can observe that a decent number of people add to cart from the Sale page. I might want to then look more into this by digging into the Sale page to check if particular items were of higher interest to help inform my marketing strategy.

What led to users removing my app?

For app developers, a big concern is churn (users removing the app). Backwards pathing can be a great resource to help find out what led to customers churning from an app. To start, you’ll create a new backwards path by hitting the “Start over” button at the top of a Path exploration. You’ll then choose the event name of “app_remove” as your starting point.

If you path back a couple of steps, you can start to observe activities that may have led to churn. In this case, we can observe that roughly 13% of churn that occurred had seen an ad impression within 2 steps of removing the app. This is a pretty large percentage indicating that perhaps the way that you show ads may need to be optimized in order to reduce the negative impact on users and therefore reduce churn.

Are there interesting intersections between various segments of my data?

The Segment Overlap technique is a great way to easily visualize how different segments of users interact with each other. For example, I could visualize the overlap of desktop users vs mobile users who subscribed to a site’s newsletter. This could more easily show me where a majority of newsletter signups come from, rather than breaking down a table exploration to get this info.

To do this, you’ll first need to add the segments you want to analyze to the variables pane. Click the + icon to add new segments (doing so will open the segment builder).

You can create the Newsletter Sign Up event by selecting an event segment and choosing the appropriate event to filter on, in this case, “sign_up”

Once you’ve built the segments you want to visualize, you’ll need to apply them by double clicking them to be added to the Tab Settings pane. You can add up to 3 segments at a time to visualize their overlap via a Venn diagram.

In this example, I’ve added segments for “Desktop traffic”, “Mobile traffic”, and “Newsletter Sign Up” to check where the majority of newsletter signups come from.

As you can notice from both the Venn diagram and row 6 of the table below, the vast majority of newsletter sign ups come from desktop users. This may indicate that either the mobile experience isn’t tailored towards this objective, or that there may be a UX issue preventing users from easily signing up for the newsletter. This is a takeaway I may want to look further into with my dev and UX teams.

A very useful feature of the Segment Overlap technique is the ability to create a new segment from an overlap. For example, if I want to create a new segment that includes desktop traffic and newsletter sign ups, I can right click on the visual slice of data I’m interested in, either on the Venn diagram or via the table, and select the option to create a segment.

Doing this will open up the segment builder with the conditions pre-selected based on the segment set you’ve chosen. It will assign the descriptive name of the combined conditions as the segment name, but you can change this if desired.

You can even choose to create an audience from this overlap by checking the box in the upper right hand corner, allowing you to share this audience with other Google Marketing Platform products, such as Google Ads, to help increase your reach.

Do users behave differently depending on when they first visited my site?

The Cohort exploration is a useful tool to understand how different groups of your site’s users behave, based on when and how they entered a cohort. Changing the metric or calculation of what you’re analyzing in a cohort exploration can make it even more useful.

When you first open a new cohort exploration in Explore, you’ll notice an exploration generated for you based on cohort inclusion of first_touch (aka how someone qualified for the cohort - in this case, the first time they visited the site or app), return criteria of any event (meaning they came back and did anything on your site or app), a metric of active users, and a calculation type of standard. This results in a basic cohort exploration showing you how many users you are attracting back to your site each week over the course of 5 weeks, and if that varies depending on when they first arrived on your site or app.

If you change the metric type of this exploration from “sum” to “per cohort user”, you can get a sense of the return percentages by week. This is likely more useful than the sum for this particular view because it gives you a sense of how each cohort compares for returning users.

For an ecommerce site, you may also find value in changing the Value from a metric of Active Users to a metric of “Purchase Revenue” to not only view users but the revenue value each cohort is bringing in. You can look at this as the sum for the cohort, or on a per cohort user basis. In the below example, you’ll notice that the first 2 cohorts generated very little revenue each week, but starting with the third cohort, revenue starts to grow.

If you change the calculation type from Standard to Cumulative, you can observe the overall impact of each cohort on the business revenue.

In this example, the data suggests that starting around the week of March 7th, there was likely an increase in marketing efforts to drive high converting users to the Google Merchandise store.

Which referral sources are generating the most valuable users?

Knowing this information can help inform your marketing team which referral sources they may want to try to cultivate as partners. In this case, value indicates users are less likely to churn, have higher expected lifetime revenue and higher historical LTV.

To build this exploration, you’ll first want to bring in the metrics and dimensions needed to explore on things like Churn, LTV, and lifetime revenue. We’ll use source and medium as our dimensions, and for metrics, we’ll pull in a few additional by adding them from the “User Lifetime” section using the + icon on the metrics section of the variables pane.

Once you apply these metrics and dimensions, you’ll notice that many rows don’t have values for churn probability, because those users do not qualify for the churn prediction modeling. You can create a filter where churn probability > 0 to exclude all users without a computed churn probability. You’ll then have a clean looking exploration like this:

Then you can sort your data by churn probability and check if there is any campaign with a sizable amount of users and a low expected churn. Those campaigns are likely to generate more engaged users than campaigns with a higher predicted churn value.

Similarly you can look at predicted purchases, historical LTV, and predictive engagement to identify the value of the acquired users beyond the single session.

Taking a look at the screenshot below, in this case, there are a few rows of non-google referrals to the Google Merchandise store which have low users, but have lower churn, higher engagement duration, higher transactions, and good average LTV. These referrals appear to be good value to the store, and so these are the referral sources we’d recommend the marketing team reach out to in order to cultivate more of a relationship with.