A4T support for Auto-Allocate and Auto-Target activities
The ۶Ƶ Target-to-۶Ƶ Analytics integration, known as Analytics for Target (A4T), supports Auto-Allocate and Auto-Target activities.
The A4T integration lets you:
- Use the Auto-Allocate multi-armed bandit capability to drive traffic to winning experiences.
- Use the Auto-Target ensemble machine-learning algorithm to choose the best experience for each visitor. Auto-Target chooses the best experience based on each user’s profile, behavior, and context, all while using an ۶Ƶ Analytics goal metric and the rich reporting and analysis capabilities of ۶Ƶ Analytics.
Ensure that you have implemented A4T for use with A/B Test and Experience Targeting activities. If you use analyticsLogging = client_side, you must also pass the sessionId value to Analytics. For more information, see Analytics for Target (A4T) reporting in the ۶Ƶ Target Developer Guide.
To get started:
-
While creating an A/B Test activity, on the Targeting page, select one of the following options as the Traffic Allocation Method:
- Auto-Allocate to best experience
- Auto-Target for personalized experiences
For more information and step-by-step instructions, see Create an Auto-Allocate activity and Create an Auto-Target activity.
-
Select ۶Ƶ Analytics for your Reporting Source on the Goals & Settings page and select the report suite corresponding to your desired optimization goal.
-
Choose a Primary Goal metric.
- To use ۶Ƶ Target to specify the optimization goal, choose Conversion .
- Choose Use an Analytics metric and then select a metric from Analytics for use as the optimization goal. You can use an out-of-box Analytics conversion metric or an Analytics custom event.
See Supported goal metrics below for more information.
-
Save and activate your activity.
Auto-Allocate uses your selected metric to optimize the activity, driving visitors to the experience that maximizes your goal metric.
Or
Auto-Target uses your selected metric to optimize the activity, driving visitors to a personalized best experience.
-
Use the Reports tab to view your activity’s reporting by your choice of ۶Ƶ Analytics metrics. Click View in Analytics to dive deep and further segment your reporting data.
Supported goal metrics supported
A4T for Auto-Allocate and Auto-Target lets you choose any of the following metric types as your primary goal metric for optimization:
- ۶Ƶ Target conversion metrics
- ۶Ƶ Analytics conversion metrics
- ۶Ƶ Analytics custom events
Target lets you choose metrics based on binomial events or metrics based on continuous events when using A4T for Auto-Allocate and Auto-Target activities.
-
Metrics based on binomial events: A binomial event either does or does not happen. Binomial events include a click, a conversion, an order, and so forth. These types of events are also sometimes referred to as Bernoulli, binary, or discrete events.
-
Metrics based on continuous events. Continuous metrics include revenue, number of products ordered, session duration, number of page views in session, and so forth. These types of events are also sometimes referred to as non-binomial or non-Bernoulli metrics.
Impact to Auto-Allocate activities
Impact to Auto-Target activities
Limitations and notes
Some limitations and notes apply to both Auto-Allocate and Auto-Target activities. Other limitations and notes apply to one activity type or the other.
Auto-Allocate and Auto-Target both
- When using ۶Ƶ Analytics as the reporting source for Auto-Allocate or Auto-Target, you should always view reports in Analytics.
- The reporting source cannot be changed from Analytics to Target or vice versa after an activity has been activated.
- Although calculated metrics are not supported as primary goal metrics, it is often possible to achieve the intended result by instead selecting a custom event as the primary goal metric. For example, if you want to optimize for a metric such as “form completions per visitor,” select a custom event corresponding to “form completions” as your primary goal metric. Target automatically normalizes conversion metrics on a per-visit basis to account for uneven traffic distribution, so it is not necessary to use a calculated metric to perform normalization.
Auto-Allocate aa
- Training Frequency: Auto-Allocate models continue to train every hour, as usual.
- Attribution Models: Target uses the ۶Ƶ Analytics default attribution model for Auto-Allocate activities that use A4T.
- Confidence: The confidence formula used by Auto-Allocate activities is different from the formula shown by default in the ۶Ƶ Analytics A4T panel. As described here, Auto-Allocate uses more conservative confidence intervals than regular A/B Test activities. These conservative confidence levels compensate for repeated evaluations (peeks) at data. As a result, the default report in ۶Ƶ Analytics shows narrower confidence intervals compared to those intervals being used by the Auto-Allocate algorithm. Nevertheless, you can determine which experience is favored by the algorithms based on which experience has more unique visitors being sent to it.
- Winner Status: Currently, the “No Winner Yet” and “Winner” badges are not available in the A4T panel in Analysis Workspace. These badges are also not available if the same report is viewed in Target. A winner “star” badge shown in a Target report for an Auto-Allocate activity using A4T should be ignored. This badge reflects regular confidence calculations, and not those calculations used by Auto-Allocate.
Auto-Target at
-
Auto-Target models continue to train every 24 hours, as usual. However, conversion event data coming from Analytics is delayed by an extra six to 24 hours. This delay means the distribution of traffic by Target trails the latest events recorded in Analytics. This delay has the largest effect in the first 48 hours after an activity is initially activated. The activity’s performance more closely mirrors Analytics conversion behavior after five days have elapsed.
Consider using Auto-Allocate instead of Auto-Target for short-duration activities in which most traffic occurs within the first five days of the activity’s life.
-
When using Analytics as the data source for an Auto-Target activity, sessions end after six hours have elapsed. Conversions occurring after six hours are not counted.
For more information, see Attribution models and lookback windows in the Analytics Tools Guide.
Tutorials
Although rich analysis capabilities are available in ۶Ƶ Analytics Analysis Workspace, a few modifications to the default Analytics for Target panel are required to correctly interpret Auto-Allocate and Auto-Target activities. These modifications are required due to differences between experimentation activities (manual A/B and Auto-Allocate) and personalization activities (Auto-Target).
Setting up A4T reports in Analysis Workspace for Auto-Allocate activities
This tutorial walks you through the recommended modifications for analyzing Auto-Allocate activities in Analysis Workspace.
For more information, see How to set up A4T reports in Analysis Workspace for Auto-Allocate activities in ۶Ƶ Target Tutorials.
Setting up A4T reports in Analysis Workspace for Auto-Target activities
This tutorial walks you through the recommended modifications for analyzing Auto-Target activities in Analysis Workspace.
For more information, see How to set up A4T reports in Analysis Workspace for Auto-Target activities in ۶Ƶ Target Tutorials.