Unica Interact - Better Handling of Learning Attributes with Auto-Binning

Segmentation in marketing allows brands to tailor their efforts in the right direction to the right audience. It lets you divide your customer base into segments based on their characteristics, demographics, age, gender, salary, and many such factors. Those efforts can relate to both communications and product development. Unica Interact allows marketers to perform real-time segmentation based on their behavior and their actions while exploring your product and website. One such great feature of Unica Interact is Auto Binning that lets marketers analyze the data collected after segmentation or categorization; the following article talks about the same.

Auto Binning – What is it?​

Interact’s built-in learning algorithm works partly by saving and analyzing the values of learning attributes when the offers were contacted and responded to. Some attributes may have a virtually unlimited number of unique values. However, due to limited resources and/or practical categorization requirements in an Interact system, we can save only a small number of them. 

How will it help?

It’s often observed that it’s more reasonable to do the analysis based on the ranges of the values rather than its actual n number of values. Auto binning allows customers to create bins(Segments/Ranges) in Interact, and the learning sub-system will automatically do the mapping.

How to define Bins in Interact DT?

The bin definitions can be created from Interact -> Global Learning -> All Bin Definitions page, using the mapped learning attributes. The bin could be either Range Type or List type.

For example, A Salary attribute can have n number of distinct values, and when this attribute is used in learning, analysis on these distinct values could be quite difficult. Hence we can divide different value ranges of Salary attributes into separate bins.

We have created 3 bins on the Salary attribute 

Low Income >=35000

Medium Income between 36000 -50000

high income > 50000 as displayed in the below image

How does Auto Binning work?

A bin definition is global data across all interactive channels and across all learning models. All bin definitions will be deployed as part of Global Deployment Data. You can deploy them in any interactive channel, deploying once and deployed for ALL.

When global data is deployed to the run time, all the bin definitions will be parsed and compared to existing ones. If there is any change in the bin definition, all the existing data for that attribute will be cleared.

When a contact or response event is posted, the value of a learning attribute is mapped to a bin if such a bin exists. If bins are defined for an attribute, The “bin” values are used while logging to the learning tables. If bins are defined for the attribute and the attribute value is not part of any bin definitions, then attribute value will be logged as OTHER in learning tables.

For example, as displayed in the image, the value of SALARY =51000, 51000 is the original value stored in UACI_LearningAttributeHist, and “High income” is the Bin value stored in the UACI_OfferStats table.

Upgrade Impact

A New Configuration Parameter SaveOriginalValues is added under Affinium|interact|offerserving|Built-in Learning Config, with values All Values, Binned Values, None. A New Table UACI_LearningAttributeHist is added to the learning schema, which stores the original values and datatypes of the attribute on which learning is performed.

The auto binning feature is helpful in analyzing the data, which can have a large number of Values. It helps to divide the data between Bins(Segments) based on the Ranges of Values or a list of Values. To understand more about it you can have a read to our Product Guide and you can even reach out to us for any more queries, and we will be happy to help.