Use the Personalization Insights Reports

Automated Personalization and Auto-Target activities use advanced machine learning to serve the most tailored experience to each visitor based on his or her individual customer profile and the behavior of previous visitors with similar profiles. The Personalization Insights reports provide information on how these models make their decisions.

Let’s look at the personalization insights reports, that come with automated personalization and auto target activities. These types of activities use Adobe Sensei personalization models, to deliver the right experience to the right customer, at the right time. The personalization insights reports are designed to provide you more information on how these machine learning models are making content decisions. While setting up your AP or AT activity, the only thing special you need to do, is make sure you use a conversion based goal metric. If you use a revenue based goal, you won’t get the reports. So be sure, that the goal is conversion based before you push the activity live. After launching your activity, you’ll see menu options for the two insights reports. At first, they’ll be disabled. After the activity has sufficient data and after a minimum of 15 days since going live, the reports become available. Once they’re available, you’ll be able to click on the icons. Let’s start with the important attributes report. One thing to point out, right away is the date range selection, is a little bit different from other target reports. You begin by selecting the end date and then use the duration dropdown to choose how far back you want to look. The important attributes report, shows you which individual attributes are most important to the model’s decision making. These can be built in the attributes, like time of day, and geo location, or custom attributes, like your Experience Cloud Audiences or the profile parameters you pass in your target calls. The interface displays the top 10 attributes in their weighting. The bar graph at the top is to help you visualize the relative importance of each attribute. You can download a CSV to see all 100 attributes. The top 100 attributes will total 100 percent. Now, let’s pivot to the automated segments report. Automated segments goes a step further to identify segments of visitors, made up of specific attribute values, sometimes combining multiple attributes. On the left you’ll see the 20 largest automated segments identified by the model. These aren’t the segments that you created or shared from another experience cloud solution. These are segments discovered by the personalization models. This segment definitions should mostly be self-evident although sometimes you might need to look up some latitude or longitude coordinates to understand what this segment represents. Also keep in mind, that the segments aren’t mutually exclusive. A single visit might end up in multiple segments in the report. When you click on a segment, you’ll get more details showing you how the segment responded to the content options in your activity, on the right hand side. The blue bar represents the number of visits in each experience for the selected segment. The pink line represents the conversion rate of those visits. The grey dash line represents the conversion rate for all personalized visits in the activity. For example, in this segment of evening visits, twenty one thousand one hundred seventy six visits were personalized with the credit card experience and had a conversion rate of 7.28% The top ribbon shows us how many visits fell into this segment. In my segment one almost twenty three thousand of the roughly thirty three thousand personalized visits fell into this segment, or almost 70 percent and had an overall conversion rate of 7.65%. One thing you might notice, is that the number of personalized visits shown on the top, is not the same as the number of visits in the activity as a whole. This is because the insights reports only use the personalized visits traffic. This excludes the control traffic selected, when you set up the activity as well as traffic that gets served by the overall winner model. The overall winner model is something AP and AT use when a single experience is outperforming the personalized approach. So, I hope that personalization insights reports shed some light on what Target’s personalization algorithms are doing behind the scenes. They might trigger ideas for new hypotheses to test, new audiences to build, and new ideas for additional data to send to targets profiling system. If you want to go deeper into the math, you can read the Adobe’s research team’s white paper which is available online and can be found through the documentation. Thanks a lot. Take care. -

Additional Information