ogle Analytics 4 features to recover lost data

We’re now in the age of Google Analytics 4 (GA4), with the legacy version of Google Analytics soon to be retired . Apart from a major facelift, data model changes were made. One of the most important upgrades to the platform was the enhancement and refinement machine-learning capabilities.

Google Analytics now allows you to combine unobserved and observed data. This is a great benefit and a necessity, as browser cookies and user identifiers are constantly changing.

We must adapt to the fact that our analytics and tracking tools are losing data. GA has some simple features that will compensate for this loss and allow you to remain data-informed.

Explore deeper: 3 secret’ tools for marketing in Google Analytics 4

Unobserved Data: Why it Matters and How it Works

No matter what analytics tool you are using, leverage unobserved data to stay in tune with digital marketing trends. The difference between observed and unobserved data is the collected and model data.

It used to be easier to track users using cookies, since most browsers accept cookies. It works by stamping users with cookies when they visit a website. GA uses this cookie to allow them to identify users based on their device information, location, demographics, and most importantly, a randomly generated ID that is “sticky”.

GA recognizes the user’s ID when they return to the site. This combines the user’s past activity with their past information. The behavior for mobile apps is the same. Devices are given an advertising ID instead of cookies. (Android and iOS use different versions.

Things have changed slowly over the years, and they will continue to improve. This old behavior was problematic because it allowed users little or no control over the sharing of their personal data. Privacy was not a consideration. Organizations had 100% control of the information of their audience.

Google Analytics does not track personally identifiable information (PII). However, this is against the terms. The definition of PII can change depending on how policies and interpretations are made by different law and security teams.

Users can now opt out or block analytics tools from gathering data. Automatic opt-out is the default in GDPR, and other countries’ laws will likely adopt it. It is the ” cookieless.”

We won’t be able to access the same amount or detail of user data as we used to. So it’s time for us to fill that gap. There are many features in Google Analytics 4 that can make up lost data. These features require very little or no lifting once tracking is set up. You can therefore test them and take advantage of their benefits today. These are three examples:

• Data-driven attribution
• Predictive metrics
• Behavior modeling

Get deeper: What are marketing attribution tools and predictive analytics tools?

1. Data-driven attribution

If you are not familiar with GA4, data driven attribution may be difficult to locate. It is located in the Advertising screen, not the Reports area. Because they offer a different view of your data, the Advertising reports are very interesting and can be split out.

Universal Analytics (sometimes called GA3) has a close equivalent: the Multi-Channel Funnel reporting. This is a useful descriptor as these reports allow for more detailed analysis of conversions, and offer a richer user experience. Data-driven attribution used to be only available to 360 paid accounts, but it is now available to everyone.

DDA attribution models use a statistical model that shows how important a channel was in helping a conversion. Although there may be 5,000 organic search channel purchases in GA4 acquisition reporting, the previous touchpoints from Paid Search may have been significant to the final purchase.

The statistical model will use data from users and their paths to conversions to determine the credit that each touchpoint should receive. Credit would not be 100% going to organic as in the previous example. Instead, credit would instead be divided by percentages across all channels that users used before making a transaction.

The Advertising > Conversion paths report contains the visualization of DDA (pictured above).

2. Predictive metrics

Although we have data on what users saw and what they engaged with, what next? Because it is “future behavior”, this is the best example of unobserved information. This feature is currently only applicable to ecommerce and churning.

Before predictive metrics and predictive audiences can use predictive models, ecommerce tracking must be established. The top areas to use predictive modeling if you have ecommerce tracking are the Explore reports, and the Audience tool.

Predictive metrics can be best used in the User Lifetime method. This report type allows you to choose which metrics to import based upon purchase probability, predicted revenue, and churn probability. These metrics are covered in the section of the selection screen.

GA4’s predictive data (both here and in Audience) is based upon past user activity. The model can learn trends from the data points about users who have purchased compared to those who didn’t. The model uses data points to identify users who have become inactive and active users to predict who will not return to your app or site in the coming week.

These insights can also be used in Google Analytics. Segments and audiences can be used to identify likely purchasers. You can create a predictive audience with just a few clicks by going to Administration > Audiences> New Audience > Predictive. These will provide pre-made templates for you to customize as you wish (pictured below).

3. Behavior modeling

Because it directly affects the source of user tracking, the identifier, behavior modeling is the most powerful machine-learning feature. This involves integrating GA4 into your cookie consent management tool, so Google Analytics can collect data about users who have not consented to being tracked.

Although it may sound counterintuitive, the data is anonymized so that they are not linked to any cookie or user identifier. Instead, anonymous event-only data will be used to determine user level activity. Because it is based on data from your app or site, it is powerful. A machine-learning model is built from the behavior of tracked users (users who have opted in to tracking). It can then be used to predict the behavior of those who have opted out.

Google’s documentation consentmode provides information and guidance to help you start conversations about behavior modeling. Reporting Identity > Blended gives you the option to enable behavior modelling in your GA4 account.

GA4’s machine learning features: Making the most of them

These tools can help you to transform questions about users and data from “How many page views did page X get?” to “Which users most likely to make large purchases within the next seven days?”. This sophistication is much easier to use.

GA4’s machine learning methods can be combined with audience-sharing and remarketing to launch your analytics beyond analysis. You can even use GA4’s audience-sharing tools for immediate use cases, audience engagement, and RoAS impact.

Explore GA4 more with these stories

The post 3 Google Analytics4 features to make up data was first published on MarTech.