Thresholding, Sampling, and Cardinality in Google Analytics 4

For all its advanced capabilities, GA4 introduces complexities in how data is processed, analyzed, and interpreted.

Central to these complexities are three critical concepts: data thresholding, sampling, and cardinality.

Understanding these concepts is of great importance to marketers and data analysts who seek to leverage GA4’s capabilities effectively.

In Google Analytics 4, the following icon shows you if a form of sampling of thresholding is applied to the data in the overview:

Now that we know where to see if any of the aforementioned data concepts are applied to our data, let’s update our understanding on what these concepts entail.

And also, let’s look into what their implications for data analysis are.

Understanding the Core Concepts

Data Sampling

Sampling in GA4 is the process of analyzing a subset of data rather than the entire dataset.

This approach is adopted to manage and process vast amounts of data efficiently, ensuring that reports are generated swiftly.

However, sampling introduces a trade-off between speed and accuracy, as the selected subset may not perfectly represent the whole dataset.

Underlying Mechanics of Sampling

GA4 employs advanced algorithms to determine which data points are included in the sample.

This decision is based on a variety of factors, including the size of the dataset and the computational resources required to process the request.

The aim is to provide a snapshot that is as representative of the entire dataset as possible, but discrepancies can arise, especially in complex or highly segmented analyses.

Sampling Occurences

n GA4, data sampling becomes a nuanced process activated when event volumes in a report, exploration, or data request exceed the property’s predefined quota limits.

This method helps manage large datasets effectively.

The quota limit for event level queries is 10 million events for standard Google Analytics properties and up to 1 billion events for Google Analytics 360 properties.^{(Google source)}

The sampling mechanism involves a strategic selection of a subset of the data.

This dataset’s subset is then extrapolated to deliver results that, while not exhaustive, aim to accurately reflect the broader trends and patterns inherent in the entirety of your data.

Strategies to Counteract Sampling Effects

1. Leverage Standard Reports: Whenever possible, utilize GA4’s standard (pre-aggregated) reports for your analysis. These reports are less likely to be subject to sampling.
2. Analyze in Chunks: For large datasets, consider breaking down the analysis into smaller, incremental time frames (date ranges) or segments. Analyze these chunks individually and then aggregate the insights manually.
3. GA4 to BigQuery Integration: For users needing precision, exporting data from GA4 to BigQuery allows for analysis without sampling, offering a granular view of the data.
4. Sampling Settings for 360 Users: If you’re using GA4 360, take advantage of the ability to adjust sampling settings. GA4 360 offers the option to choose between faster analysis with more significant sampling or more detailed analysis with less sampling, depending on your needs.

In GA4 360, if you find that your data is sampled, you can choose between a higher level of detail in reports or faster creation by clicking the the data quality icon:

• More detailed results: Uses the maximum sample size possible to give you results that are the most precise representation of your full data set
• Faster results: Uses a smaller sampling size to give you faster results

Data Thresholding

Thresholding is GA4’s response to privacy concerns, ensuring that data cannot be used to identify individual users.

This mechanism is particularly relevant in reports with small user bases or those that delve into detailed demographic information.

Thresholding’s Activation Triggers

Thresholding is activated when the granularity of the data poses a risk of identifying individual users.

GA4 evaluates the potential for user identification in each report, applying thresholding as needed to anonymize the data.

Navigating Through Thresholding

• Broadening Query Scope: Expanding the dataset can dilute the granularity to levels where thresholding is not required. E.g:
  • Extending date ranges
  • reducing the level of filters applied
  • using broader categories
• Anonymization Techniques: Employing methods that further anonymize data, such as using User IDs. Ensure they are anonymized or hashed. This can help protect user privacy and reduce the likelihood of thresholding by making the data less personally identifiable.
• Exporting to BigQuery: By exporting GA4 data to BigQuery, you can analyze detailed data without GA4’s thresholding constraints. This approach allows for the use of advanced queries and analyses that might not be possible directly within GA4 due to thresholding. Further reasoning behind setting up the GA4 data export to BigQuery can be found here.

The different type of icons representing sampling and thresholding states in Google Analytics 4 are as follows:

State	Details	Sample message
	You are seeing all available data for the selected dimensions and metrics.	Unsampled report This report is based on 100% of available data.
	You are seeing some available data for the selected dimensions and metrics.	Thresholding applied Google Analytics has applied thresholding to one or more cards in this report and will only display the data in the cards when the data meets the minimum aggregation thresholds. Learn more
	You are seeing a small percentage of the available data for the selected dimensions and metrics.	Heavily sampled exploration This report is based on 8.88% of available data. A smaller sample size means that the data in this report is less accurate. Learn more

Cardinality

Cardinality refers to the number of distinct values that a dimension can take.

It plays a crucial role in how data is structured and analyzed within GA4, impacting both the granularity and usability of the data.

Cardinality example

Consider the dimension of Device Type, which typically encompasses a limited array of unique values: desktop, tablet, and mobile. In this case, the cardinality of the “Device Type” dimension is 3, indicating a low cardinality since the number of potential values is small and finite.

Now, when we expand on the example of the Device Type dimension, let’s consider another dimension that could be tracked within the same dataset, such as Operating System.

This dimension might include values like Windows, macOS, Android, and iOS.

If we combine these two dimensions, the potential combinations increase significantly. For instance:

• Device Type: Desktop, Tablet, Mobile (3 unique values)
• Operating System: Windows, macOS, Android, iOS (4 unique values)

When these dimensions are combined, the total number of unique combinations (cardinality) becomes 12 (3 device types * 4 operating systems).

This example shows that cardinality levels increase from low to medium when transitioning from analyzing individual dimensions to combining two dimensions.

The dataset encompasses a broader array of unique values with the addition of an extra dimension.

Cardinality challenge in GA4

However, cardinality becomes a challenge within GA4 when it’s high.

High cardinality occurs in dimensions that have a broad range of potential unique values.

For instance, the Page URL dimension can exhibit high cardinality due to the vast number of unique URLs a website might have.

Similarly, Event Names might display high cardinality if a wide variety of custom events are tracked, each with unique identifiers.

In these examples, the variability (the range of different values a dimension can have) is substantial, leading to high cardinality.

Implications of High Cardinality on Data Analysis

GA4’s ability to process and present data can be hampered by high cardinality.

This is because dimensions with a large number of unique values require more computational resources to analyze and can lead to two main issues:

1. Increased Sampling: To manage the computational load, GA4 may resort to sampling data, analyzing only a portion of the dataset and extrapolating results from this sample. This can affect the accuracy of data analysis, especially for detailed reports.
2. Aggregation into “(other)” Category: When the number of unique dimension values exceeds GA4’s processing capabilities, excess values are often aggregated under the “(other)” label. This aggregation masks specific details, hindering the ability to derive precise insights from the data.

The (other) row is a row that appears in a report, exploration, or Data API response when the number of rows in a table exceeds the table’s row limit.

Strategies for Effective Management of High Cardinality

1. Dimension Simplification: One approach to managing high cardinality is to simplify dimensions. This can involve consolidating similar values or reducing the variety of tracked values.
2. Leveraging BigQuery for Detailed Analysis: Exporting GA4 data to BigQuery offers a solution to cardinality challenges. BigQuery does not have the same cardinality limits as GA4, enabling detailed analysis of the entire dataset without the need for aggregation or sampling. This allows for a deeper dive into the data, uncovering insights that might be lost due to the limitations within GA4 itself.
3. Custom Dimension Creation: Tailoring dimensions to your specific analytical needs can also help manage cardinality. By creating custom dimensions that aggregate high-cardinality data into more manageable categories, you can maintain the depth of analysis while avoiding the pitfalls of high cardinality.

By understanding and addressing the challenges posed by high cardinality, analysts can ensure that their GA4 data remains both accessible and insightful, facilitating more informed decision-making processes.

Now that we understand sampling, thresholding and cardinality..

These insights hopefully provide you with new confidence in navigating the complexities of data analysis in GA4.

This blog aims to enhance your analytical skills and improve your data collection methods. It also focuses on refining report accuracy to help you make more informed strategic decisions.

Leveraging Your Insights

Now that you’ve expanded upon your foundation of knowledge with new insights on data thresholding, sampling and cardinality, you can:

• Streamline your GA4 setup, balancing detail with privacy.
• Adjust your reporting strategies to counteract the challenges that data analysis can provide and ensure more reliable outcomes and better understanding of outcomes in general.
• Use your insights to influence strategic business decisions, driving growth and innovation.
• Promote a culture of data-driven decision-making within your team, enhancing collective understanding of the data and foster increased team successes.

Thanks!

If you found this guide helpful, subscribe to our LinkedIn to receive updates whenever we post a new blog!

Share your experiences or questions about the Facebook Pixel in the comments below!