Google Analytics 4: Data Models & Functionalities Review

Google Analytics 4: Data Models & Functionalities Review

Understanding data models and the functionalities of analytics software is vital in the implementation of any successful SEO or CRO strategy. Over the past week, I have been focused on deepening my insights into GA4 as a conversion research tool. I took a CXL course to achieve this aim.

The course instructor was Charles Farina and the outline introduced me to the future of technical marketing, data models, and other components of an evolving measurement protocol.

Why a deep understanding of GA4 is necessary

In conversion optimization, changes are made on websites in order to upgrade the number and value of major and minor conversion events.

Nevertheless, these changes can only be made after developing behavioral models or by generating conversion hypotheses. While conversion hypotheses are the basis of test ideas (A/B, multivariate, or redirect tests) they can only be developed after some research has been conducted. A critical component of this research is the data obtained from web interaction analyses.

This is where a deep understanding of GA4 becomes imperative. Of all the existing web analytics tools, the most popular and accessible is Google analytics. This is why a thorough understanding of Google's vision for data capture and organization is important.

Brief History of Google Analytics

Google Analytics is about 16 years old. It started in 2005 when Google purchased urchin, and developed a free web analytics tool that rivaled the existing but costly alternatives at that time.

Google's vision was to make site-interaction data available to marketers and publishers, in order to improve the quality of internet experiences, and ultimately grow the global search volume.

Google's approach worked, global search volume grew and urchin was a success. This led to the creation of a more advanced version called Classic analytics. From classic analytics, universal analytics was developed, but due to certain cross-platform complications (rise of mobile apps, enhanced commercial capabilities on the internet, shifts in user behavior, etc) , Google moved to create GA4.

The Nature of GA4

Google maintains two measurement platforms with similar functions but vastly different features and capabilities. These two platforms are called Firebase and Universal analytics. Firebase is for mobile apps, while UA is for websites. Firebase has certain desirable capabilities (like funnels and BigQuery exports) that are not accessible to UA users on the free version. So what Google did was to unify both platforms (App+web) while rebuilding the UI, data model, and backend. This unification of firebase and UA is what makes for Google analytics 4.

GA4 vs Universal Analytics: Features and Capabilities

There are several points of divergence between UA and GA4 but I'll touch on9 which are:

1) Data model redesign

Universal Analytics uses a hit-driven data model where all tracked data units are aggregated in hit categories e.g page hits, event hits, social hits, etc.

GA4 on the other hand adopted the firebase data architecture, which is why it moved from a hit-based to an event-driven data model.

The event-driven data model is more flexible from a statistical point of view as it simplifies the task of a technical marketer or conversion researcher.

2) Automatic measurement capabilities:

In universal analytics, once the tracking code is installed, the default class of data being captured is the page view hits.

This is sort of insufficient in conversion research or to meet the needs of data-driven marketers. This is why Google modified GA4 into a lite version of google tag manager.

They did this by creating automatic measurements that expanded the scope of data capture to include; file downloads, video engagement, outbound link tracking, and scroll tracking.

This offered more data for less tracking code manipulation and is a major difference between GA4 and universal analytics.

3) New Report organization

In universal analytics, the default reports were organized using graphs and tables. In GA4, Google introduced the report card concept and embedded real-time interactivity into it.

In real-time reports, GA4 also offers more data, custom dimensions (or user scope dimensions) are visible in the dashboard, and there's also a separate reporting API.

4) Default reporting identity

In the existing version of Google Analytics, you can only track users by the default client ID or first-party cookie.

But if you wanted to track users across different devices or browsers, you would then have to implement your own user I (this is done in the user ID view).

So when someone logs in you'll send a user ID and then in a separate place in Google Analytics, it'll use those IDs to identify those users.

Issues with UA reporting identity

This setup is not ideal because you're forced to look at your anonymous users in one place and then your logged-in users at another place.

In addition, many marketers don't have a lot of user IDs or don't even have a concept of a login which makes everything on their site anonymous.

GA4's identity graph features and lots more

Where GA4 is helping marketers is that it captures client IDs and user IDs and actually uses both to create a better-connected view of your users in one dashboard.

Another huge improvement is the integration of Google Signals especially for marketers who don't have their own user IDs to work with. Google Signals is Google's identity graph or Google's logged-in users who have opted in and consented to share data with Google. This feature in GA4 allows marketers to get better visibility into what's happening on their platforms and experiences.

With this data, even without your own user IDs, you can use Google's identity graph to add user-based contexts to all of your reports.

5) Debugging capabilities

In GA4, tracking code issues and other reporting discrepancies can be investigated from within the UI. In universal analytics, this isn't possible.

The only way to debug data errors in UA is to use the tag assistant or Google analytics debugger extension. Alternatively, you may have to open your browser's developer console to investigate these issues like a web developer.

With GA4 you can clean up your tracking code, data quality and improve your reporting data right within the UI. All you need to do to activate the debug view by:

a) Using the preview mode in GTM

b) Using the GA debugger extension

c) Sending a debug mode flag through on-page events

This debugging window within the UI is a major improvement and this is important for conversion researchers who are concerned about data quality.

READ: GA4 Is Here: Here’s What You Need to Know

6) Segmenting and Audience Reports and Conversion tracking

Google analytics 4 now has a hit scoped segments feature (something Adobe users used to brag about) which allows us to identify the exact set of hits that met certain advanced conditions.

GA4 also has an audience builder which adds more flexibility to the segmentation process. In GA4 the concept of time has become more malleable unlike the relatively static time component in UA reports. This allows marketers to build audiences based on how quickly forms are completed, how much time is spent on a blog post etc.

GA4 also ups UA due to the temporary and permanent exclusion feature in its segments. This is particularly useful in remarketing allowing us to create flexible remarketing lists on the fly.

In Conversion tracking, UA allowed for the conversion actions to be based on the user triggering certain events or viewing certain pages. In GA4, every tracked event can be tagged as a conversion with just a simple toggling of the settings. The goals can also be archived which allows more flexibility than was possible in UA. This is because UA has a limit of 20 goals per view in its free version which is restrictive for older accounts.

7) Funnel Visualization

In GA4, the funnel visualization features allow for the creation of open and closed funnels, funnel customization, trended funnel analytics, and elapsed time functionality. All of these are not available in UA which is another point of divergence between both versions

8) Pathing

In Universal Analytics, there were these user flow reports that were very difficult to work in any meaningful way. To address this, GA4 now incorporates a Pathing report which allows for the tracking of user paths based on URLs, events, page titles, etc. They also introduced backward Pathing which allows for us to start the path trace from an end goal (checkout, form submit) and reverse the average user journey from that end goal

9) Potential for Big Query/CRM integration

Use case example 1

If you want to build advanced visualizations or you want to use Tableau with Google Analytics, it's basically impossible to do that at its full power with the free version. This is because you have to use the API and those APIs are sampled, only allowing you to access subsets of your data.

Use case example 2

Maybe you want to integrate Google Analytics with your CRM. To do this, you need to have access to all the raw data & user IDs, but since UA doesn't have a great way to get your data out, the only solution was to upgrade to Google Analytics 360 which would unlock the BigQuery or the raw data feeds.

With GA4, the improvement is that everyone has access to a free BigQuery linking allowing a complete data export to either BigQuery's data warehouse, or to Azure or AWS. In these data warehouses, machine learning algorithms and predictive analytics can be performed on the data to uncover deeper insights.

All these are some of the new features that differentiate GA4 from the standard version of Google analytics. So as a CRO, SEO or digital marketer, my take from the course is that you should use a dual implementation of both versions in order to upgrade your practice.

Course Summary

This particular course was rich and very insightful. This is why I made it the sole focus of this week's review. If you'd like to learn more about GA, you can consult the resources below.

Charles Farina’s Blog

Data blog — Krista Seiden