[UA] MCF Data-Driven Attribution model application and limits [Legacy]

You are viewing a legacy article about Universal Analytics. Learn more about Google Analytics 4 replacing Universal Analytics.
This feature is only available in Google Analytics 360, part of Google Marketing Platform.
Learn more about Google Marketing Platform.

Note: The GDN Impression Beta feature was deprecated for all the traffic inside and outside the EEA (European Economic Area) on 29th January 2024. Learn more about Deprecated Universal Analytics features and corresponding Google Analytics 4 features.

In this article:

Minimum threshold

In order for MCF Data-Driven Attribution to generate a model for your selected conversion type, you must meet the minimum conversion threshold for the past 28 days. The current threshold is:

  • 400 conversions per conversion type with a path length of 2+ interactions (i.e., 400 conversions for a specific goal or transaction, not a sum of 400 over all conversion types)
    AND
  • 10,000 paths in the selected reporting view (roughly equivalent to 10,000 users, although a single user may generate multiple paths)

If you have enabled MCF Data-Driven Attribution but don't meet this threshold, you will see a notification in the Model Explorer, ROI Analysis, and Model Comparison Tool reports alerting you that a Data-Driven model can't be generated due to insufficient data.

Model refresh

Your MCF Data-Driven model is applied on an ongoing, weekly refresh basis. The model weights (visible in the Model Explorer report) are refreshed on a weekly basis and include a trailing time window of conversion history (currently 28 days).

Note: When you enable MCF Data-Driven Attribution, Google Analytics uses historical conversion data from the past 28 days to build the initial model. You can view the model for the past 28 days of historical data immediately after enabling MCF Data-Driven Attribution.

Maximum number of interactions and lookback window

Up to 4 interactions within the 90 days prior to each conversion are incorporated into the model. If you have questions, contact your 360 support team.

Types of interactions included

All MCF Data-Driven models include clicks and direct traffic. If (1) you have asked your account manager to link your Campaign Manager 360 account to Analytics or (2) you're participating in the Google Display Network Impression Reporting beta, your MCF Data-Driven model(s) will also include impressions and rich media interactions.

How Direct traffic is treated

For each conversion, MCF Data-Driven Attribution removes the last 24 hours of inferred direct sessions, i.e., the model doesn't give credit to the Direct channel if there was a marketing activity that occurred within 24 hours prior to the direct session.

Gaps in data

If there are any gaps in your data, and no model for the time period could be generated, those gaps are filled with the most recent Data-Driven model. For example, if you drop below the conversion threshold for a short time period, MCF Data-Driven Attribution won't be able to create a model for that time period and will instead use the previous time period's Data-Driven model.

Insufficient data to generate a model

If the model doesn’t find sufficient data for a conversion type, the MCF Data-Driven Attribution algorithm will not create a model for that conversion type.

This can occur when:

  • A large number of conversion paths considered for the model have a single touchpoint.
    To verify that this is the cause of insufficient data, check the Path Length report to see if a large percentage of paths have an interaction count of 1.
  • The specific path is so infrequent that the MCF Data-Driven Attribution algorithm cannot be statistically confident of the validity of the model for the selected conversion type (although the path is still valid).
    • To verify that this is the cause of insufficient data, check the Top Conversion Paths report to see whether a high percentage of paths have a low interaction count.
    • Configure your default Analytics Channel Grouping to collapse low-volume channel definitions into a single definition. This will result in fewer paths with low interaction counts.

Notifications in your reports

As you are using the MCF Data-Driven Attribution reports, you may receive in-product notifications about the availability of your Data-Driven attribution model.

A Data-Driven model has not yet been created
You will receive this message if you've very recently enabled MCF Data-Driven Attribution. It can take up to 7 days to create your first Data-Driven model, so check back after more time has passed.
Unable to create Data-Driven model due to insufficient data
You may get this notification if at least one of the following is true:
  1. More time is needed to collect data for your selected conversion. For example, if you've recently set up Goals or E-Commerce Tracking, you may need more time to meet the minimum conversion threshold.
  2. There's a temporary lull in conversions for your selected conversion type, and you don't yet meet the minimum threshold.

If you continue to see this notification, it may indicate that your selected conversion type will not meet the minimum conversion threshold. Consider investigating a different conversion type.

Unable to create Data-Driven model due to internal error
You will get this message if MCF Data-Driven Attribution encounters an internal error. Check back later and notify your 360 support team if it persists.
No Data-Driven model is available for the selected week, and the most recent model is displayed.
You will see this message if a Data-Driven model can't be generated for the current week, and there is a Data-Driven model for at least one previous week. The most recently generated model will be displayed instead.
For one or more of your selected conversions, a Data-Driven model couldn't be generated due to insufficient data.
You will see this error if a Data-Driven model couldn't be generated for at least one of your conversion types. Read the insufficient data notification (above) to learn why there is insufficient data.

Was this helpful?

How can we improve it?
Search
Clear search
Close search
Google apps
Main menu