Should you implement GA4 now?


TLDR: Yes. Hop on the GA4 bandwagon. Google’s new platform is ready to collect useful insights and the sooner you start gathering data, the more you can enjoy the advantages of Google’s Machine Learning expertise to enhance your datasets. 

Since some features of Universal Analytics are still missing, I suggest you keep your existing analytics properties in place for now, while you start familiarizing yourself with GA4.

Below, I’ll help you understand the main traits of GA4 so that you can estimate the value of having the new property and how to get the most from it. 

I’ll cover;


What problems does GA4 solve?

Google’s announcement about GA4 argues that “with major shifts in consumer behavior and privacy-driven changes to longtime industry standards, current approaches to analytics aren’t keeping pace” and cites a survey from Forrester Consulting to support marketers’ complaint that “existing solutions make it difficult to get a complete view of the customer and derive insights from their data”.

This pain point seems justified if we observe how frequently users are oscillating between platforms and devices making it more difficult to evaluate the performance of marketing campaigns and manage KPIs.  

GA4 has been conceived to tackle these challenges and in Google’s own words: 

“GA4 has machine learning at its core to automatically surface helpful insights and give you a complete understanding of your customers across devices and platforms. It’s privacy-centric by design, so you can rely on Analytics even as industry changes like restrictions on cookies and identifiers create gaps in your data. The new Google Analytics will give you the essential insights you need to be ready for what’s next.”

To achieve this goal, the new GA4 property has been rethought-out, creating a unified structure across platforms and several features that are completely new or have been available only in the paid enterprise versions are now accessible.

What’s new in GA4?

The short answer is a lot. From how the data is collected, stored and used, to reporting features. 

Since a full review of all the changes and developments would require way more than a blog post, here is a pick of the most interesting new aspects that deserve a special mention in relation to how GA4 tackles the pain points described above.

The cross-platform, unified measurement model

GA4 is conceived to reduce fragmentation and facilitate data being combined across different platforms (website, mobile app), thus offering a better overview of the complete customer journey. The underlying unified data model, which standardizes the collection of essential metrics across platforms, is what makes this possible. 

This approach is inherited from Firebase and departs schematically from the session-hit data model of Universal Analytics (UA). GA4 introduces a flexible structure built upon simple events triggered by user interaction, which register a series of automatically collected or custom parameters. Sessions are still counted but they are not the cornerstones of data segmentation anymore, which leaves free space for marketers to analyze their data in ways more relevant to business KPIs.

While the event hits in UA present a rigorous format, using predefined attributes, GA4’s central event model has a freely customizable structure accepting user-scoped (user properties) as well as event-scoped parameters.

A comparison between UA and GA4 data models


Page_view event as visualized in GA4 debug mode


Why is an event-based model better than sessions and hits?

Firstly, we can observe that for different mediums the traditional sessions and hits measurement is hard to apply. Take apps. They have significantly different behavior compared to websites - so there are several cases when an analytics model using traditional website behavior conventions might not make much sense when applied to user journeys that involve them. 

You can see that a flexible event structure is generally more suitable to describe the variable nature of apps (mobile or web) where the functionality is centered around user actions rather than sessions and screen views.

Secondly, you can separate activity by the event rather than just a page the user sees. For example, the majority of transactional sites use page view hits of the “Thank You” page to measure purchases. In reality, two different events are happening: a page view, and a purchase, which would be better described by separate events with detailed parameters of their own.

All hit types used previously in UA can now be described in terms of events. Event types belong to two major categories:

Automatically collected events

  • Built-in automatic events e.g first_visit and session_start for the web.
  • Enhanced measurement events. This is a feature that can be customized in the admin UI to enable tracking of the following events, automatically; first visit, user engagement, page/screen views, 90% scroll, outbound clicks, site search, video engagement, file downloads).
 You can find the full list of automatic events (built-in or enhanced) here


Custom events

You should check out Google’s recommended events which apply to the majority of business verticals. Their usage is strongly recommended by Google, since they may be used for deeper machine-learning-based sector comparisons. 

If no automatic or recommended event type satisfies your collection goal, you can create fully custom ones. When it comes to implementing custom events, it’s standard to use a tagging solution like Google Tag Manager (GTM). Once you set up your new GA4 property and your first data stream–a data source from a website or app–you can create a connection from GTM using the new GA4 configuration tag. To use this configuration tag and add any fields or user properties, you need the GA4 “Measurement ID” of that particular data stream. 


GTM GA4 Configuration tag


You can create custom events from scratch using GTM’s Event tag, referencing either the above configuration tag or GA4’s built-in event creation feature from inside the admin UI. This lets you use any existing event as a basis, for which you can set up parameter conditions to fire a new event when the requirements are met.

GTM GA4 event tag


GA4 event creation feature


The creation of conversion goals is just as easy as toggling on options next to the event itself in GA’s UI.


However, before jumping into creating events you should always take your time to plan exactly how you will model your business KPIs in terms of the new data model. 

Moving from sessions to events

As you might have guessed there is no predefined universal upgrade path from previous UA ‘hits’ to the new GA4 event-based structure. 

When approaching your new GA4 setup it’s important therefore to think about;

  • Naming conventions. 
  • How you will use predefined events and the new event creation feature 
  • Limiting the number of completely custom events by utilizing Google’s recommended custom events.

The efficiency of different data architectural choices depends therefore on the particular measurement requirements and set up choices you make. 

A final recommendation from me on this point; it’s always good to document any new events, parameters, and conversions. Then run through as many test cases as possible to make sure the configuration delivers on your measurement requirements.  

Default reporting identity

When a visitor lands on or interacts with one of your touchpoints (website or app), analytics platforms need to identify the person to be able to decide whether it’s a new or a returning user and consequently whether to stitch together single interactions across touchpoints, if necessary. 

Universal Analytics restricts this identification process to a cookie-based Client ID by default. This means it represents a single device-browser combination, that is inherently unsuitable to ‘follow’ users across all potential touchpoints they might interact with. 

UA does offer some enhancement options on top of the above, such as the activation of a User ID view, gathering interaction data for ‘logged-in or known’ users across devices. GA4 streamlines this operation, however, and packages all the available identity data into a unified, fallback style identification logic, thus generating reports with more precise metrics. 

In practice, GA4’s default logic looks for User ID first when identifying the visitor. Then the optional Google Signals data is used, which is data from visitors who are signed into their Google account, and have turned on ‘Ad Personalization.’ If neither of these are present, GA4 falls back to Client ID (either the Analytics cookie from websites or the app Instance ID for apps). This reduces measurement fragmentation out-of-the-box and offers a better view of a user’s real journey, irrespective of whether they provided their contact information or where they made transactions. 

The integration of Google Signals into GA4 reports represents a big opportunity in itself since it delivers a holistic picture of some users’ behavior. Even if it’s only a relatively small dataset, it opens up the opportunity to use advanced machine-learning algorithms in the future to fill the potential data gaps and improve predictions. 

Nevertheless, the selection of default reporting identity is at your discretion. You can switch between the options at any time without making any permanent impact on data collection or processing.


Machine Learning

GA4 automatically enriches your dataset with machine learning predictions if the minimum requirements are met. The requirements are;

  • Over seven days at least 1,000 returning users triggered the relevant predictive condition during the previous 28 days, and that at least 1,000 users did not trigger the relevant predictive condition.
  • Model quality must be sustained over a period of time to be eligible - though Google doesn’t quantify this. 
  • To use purchase probability and churn probability, your property has to send the purchase (recommended for collection) and/or in_app_purchase (collected automatically) events.
  • When you collect the purchase event, you need to also collect the value and currency parameters for that event. 

The current predictive metrics are; 

Source: Google

 

The predicted metrics can be used in the exploration reports or to build audiences. Such predictive audiences, which are defined by using one or more of the above metrics can then be used in advertising to create remarketing audiences, or in re-engagement campaigns.

Google also suggests a couple of audiences you could set up, such as ‘users who are likely to purchase in the next 7 days’ or ‘users likely to churn in the upcoming week.’ 

 I’ve already touched on the advantages of Google Signals, but perhaps the biggest advancement with GA4 is the application of machine learning to fill in the data gaps arising from different privacy regulations or cross-device, cross-platform user activity. The process is known as conversion modeling and given our distributed, cookie-less future such models will be able to complete the fragmented picture of a business’s performance.

Refined privacy configuration options

Currently, there is no real consensus between browsers, platforms, and legislatures over what privacy rules should universally apply. With changing cookie policies, GDPR, and other specific regulations, analysts need to keep themselves updated about where data gaps caused by policy differences may arise. And what they can do to ensure seamless compliance. To aid them in this task, GA4 allows more granular control over users’ privacy compared to its predecessors.

Among the features, you can automatically enable IP anonymization, update data processing agreements, and use more GDPR friendly data storage periods (2 or 14 months). You can control your Ad personalization exclusions for single conversions or audiences as well as for specific geographies, and you also have more options to refine your data deletion requests. 

The enhanced setup options don’t remove the responsibility on marketers to keep an eye out for the upcoming policy changes but go a long way to help you set up a compliant analytics property.

Customer-centric UI

Marketers cited the difficulty involved in pulling together the pieces that build up a user journey and deriving clear insights from the data. Therefore, in addition to the above improvements, GA4 restructured the whole reporting UI, which now revolves around the customer and their lifetime with the business. 


Also, GA4’s new data model means all standard reports are unsampled. Sampling is present only in the Exploration module starting from 10 Million events per query.

BigQuery integration and Exploration module

Good marketers always look for actionable data in their Analytics accounts. There are useful and not so useful metrics, so instead, you need to focus on the conditions under which your KPIs are impacted. It’s easy to get lost among a large number of standard reports and average values but the truth about why or how your business succeeds is in the detail.

Which device or browser converts more? If there are multiple ways for the user to get to a CTA and convert, which one of these paths takes the largest slice of the pie? To get a valuable answer you need to continue down this chain of questioning and dig as deep into your dataset as possible. Once you perfect your setting and let the data flow in, GA4 delivers on its promise to serve you tools that help explain the full user journey for any segment you can think of. 

A great bonus here is BigQuery. In Universal Analytics only the paid-for “360” version had built-in integration with BigQuery. GA4 gives you it as standard. With this huge data storage option, you can build all sorts of custom reports, combining data from different sources over arbitrary periods, build integrations or run predictive analysis algorithms of your own.

You also get the new Exploration Module, free of charge which is an upgraded version of the Analysis Beta function available in UA 360. It’s an easy-to-use and fully customizable report creation canvas that features the following report types:

Free-form exploration: You can analyze the interesting datasets in a familiar crosstab layout or pick one of several visualization techniques such as scatter plot, pie chart, or geo maps.

Funnels: Create custom steps and examine the contribution or churn rate of each one, to find areas needing optimization. Build retroactive and segmentable funnels.

Path analysis: Visualize the paths users take on your site or app. See how they react to transaction errors or find suboptimal funnel configurations.

Segment overlap: Check out the common aspects of different user groups, and identify new target audiences.

Cohort analysis: Analyze user segments with common attributes.

User explorer: Drill into the data and see the behavior or performance of single segments or users. 

User lifetime: Check out user value through their lifetime.

Exploration Module is intended to be the go-to place to execute ad-hoc queries, data drill-downs, or create custom segments that can be converted into audiences to be used in standard reports or ads integrations.

Is GA4 ready? Will it replace Universal Analytics?

In reality, GA4 is still far from complete and not mature enough to replace previous solutions.

It’s a full rebuild of the old logic and it simply takes time to ensure a seamless transition. Even if Google constantly updates the product, considering user feedback, it’s still missing some critical features that marketers will find hard to live without. Some of these are (at the time of writing this post):

  • Missing Multi-Channel Funnels
  • No custom channel groupings
  • Missing integrations with other Google products such as Search Console, Adsense, Optimize.
  • Somewhat limited filtering of traffic on a property level
  • Difficult eCommerce setup in GTM
  • No product list attribution
  • Product-scoped event parameters don’t appear in reporting
  • No data-stream level access management

You can keep up to date with ongoing releases on Google’s announcements board but as of yet, it’s still unclear when GA4 will be able to fill all the gaps. 

Google is determined to do what it takes and make GA4 the Analytics product of the future. Proof of this commitment is the fact that GA4 is now the default property type for new accounts and new product releases come out almost every month.

Other factors to consider during a transition from UA to GA4 is the steep learning curve and implementation headaches. Since the structure is so much different, it will take time for analysts to familiarize themselves and learn how to take advantage of GA4’s features and quirks. The lack of detailed documentation about non-standard configuration requirements could also potentially represent an obstacle in this early phase.

Verdict

The custom reporting options, machine learning features, and the promise of eventually having an all-in-one tool to tackle cross-device, cross-platform fragmentation in the ever-changing world of data laws and regulations, more than makes up for the current missing functionalities and transition headaches. 

I wouldn’t hesitate in launching the upgrade so that you can start familiarizing yourself with GA4, gathering precious data to feed into Google’s ML algorithms, activating BigQuery Integration, or just creating interesting visualizations inside the new Exploration Module.

But I would do this in parallel to your existing UA setup, so you still maintain the reliable and well-known components of an already mature solution while Google irons out some of its little brother’s imperfections. 

Cheatsheet: Differences between Universal Analytics and Google Analytics 4

Here’s a summary table listing the main differences between UA and GA4 properties at the time of publication.

UA

GA4

Measurement model

Session-Hit based. Hits have different types and schemas (pageview, event, eCommerce, social.)

In GA4 everything is an event with custom parameters and user properties. Easier cross-platform, cross-device tracking.

Quotas

Max 10 million monthly hits per property on the free tier. 200,000 hits per user per day, 500 hits per session.

No limit on total events for now. There are quotas for the number of custom event name definitions, custom parameters, and custom dimensions. See the official list here - Quotas

High-level structure

Property, views. Views can be created and filtered to meet reporting requirements. Property and view level access management.

Property, data streams.

No “views,” and more limited high-level data filtering options for now. Reporting is property-based.

Up to 50 data streams for different customer touchpoints (separate domains, app, web). Only property-level access management.

Events

Event hits have a rigid category, action, label, value, and custom dimensions (CD) model.

Events have event-scoped or user-scoped custom parameters. More flexible structure. By default, parameters show up only in BigQuery, but they can be marked as custom dimensions to be used in standard reporting UI. Types include Auto, enhanced, recommended, and custom.

Data retention

14, 26, 38, 50 months, or doesn’t expire.

2, 14 months (no non-expiring.)

Ecommerce

Many useful reports in enhanced ecommerce (EE). Easy ecommerce hit setup through GTM. Overall, still better ecommerce reporting in UA.

Extended data model, which is backward compatible with Enhanced ecommerce but not vice-versa. Some missing features compared to EE and difficult GTM setup. No product list attribution, product-scoped event parameters and these don’t appear in reporting even if marked as CD)

Cross-device, cross-platform tracking

The User ID can be set up in a separate view. Google signals can be activated, but available only in some of the reports.

Overall better cross-device, cross-platform tracking. The default reporting identity makes sure all identity spaces are used when recognizing visitors.

Attribution

UA attributes goal completions to the last non-direct click in standard reports. There’s a Multi-Channel Funnel Model comparison tool to see how attribution across channels, source, medium, and campaigns would change based on your attribution model.

GA4 offers a “Comparisons” feature where you can change the attribution type reflected in the standard report. No MCF and default channel grouping can’t be modified.

Custom dimensions and metrics

Setup based on CD slots requires dev or GTM setup. Different scoping options. (product, hit, session, user)

Custom event parameters can be marked as custom definitions at any time. The scope is therefore always event or user in case of user properties. (no session unification.)

Debugging

No built-in debug option

Built-in debug capability

Engagement metrics

Simple metrics are available like page view and bounce rate.

More refined metrics like engaged sessions, engaged time, engaged sessions per user. No notion of bounce rate.

Privacy

IP anonymization can be enabled. Ads personalization features can be disabled.

IP anonymization is automatically enabled.

Ads personalization exclusions can be set up for single conversions and audiences as well as for geo’s.

Integrations

Ads, Optimize, Search Console integration out of the box. BigQuery only in the Analytics 360 package.

Search Console, Adsense, and Optimize are not supported yet. BigQuery out of the box.

Sampling

Sampling already in standard reports.

No sampling in standard reports, only in the Exploration hub from 10 million events per query.

Analysis options, default reports.

Custom report option with explorer, flat table, map overlay. More options in Analytics 360.

Much more refined analysis options and types in the newly built analysis hub.

Default reporting identity (DRI)

It uses Client ID by default.

DRI is User ID, Google Signal, and Device ID by default. The setting is editable, takes place retroactively. All events go through this identification chain before being attributed.


Related Posts

Who's currently reading Briefly Experimental?