Recommender traits are machine-learning traits that use the preferences of similar end users to identify the likelihood that a targeted user will be interested in a certain product from your FI. In turn, the system can then recommend products to individuals who are most likely to adopt a product based on historical data.
Every internal product has a coordinating recommender. These traits are tagged as Recommended products within the trait modal. You can create and build your own "Recommended for (product)" audiences by using our machine-learning traits and your identifying traits that determine which users may be good candidates for your product.
Similar to Nextflix, SMART uses a machines learning algorithm to recommend products to users based on the preferences of similar users. While Netflix recommends movies and television shows to users, SMART recommends banking products. For example, SMART may recommend you include a user in your campaign for a new auto loan because end users with similar product sets also have an auto loan. In the image below, SMART identified that 1,309 of your end users may be interested in a new auto loan. This number was calculated based on users who have similar product sets.
You can access these traits when you select the target audience for your campaign.
SMART uses collaborative filtering to identify a collection of end users with similar product types. This technique is shown in the diagram and explained below.
In the center of this circle is an end user who is surrounded by other end users who have product sets most similar to the center end user.
SMART computes the most popular products among this collection of users. The system counts every instance of each product and sorts the data into a list in descending order. The top items (specifically, the top 20%) in the list are the most common products among this collection of users. The system recommends these top products to the center user. This end user will be included in the recommended audience for campaigns featuring those products.
Recommended audiences may contain end users who already own the product featured in the campaign. When you define your target audience, you can exclude these users. Once a week, the algorithm checks for users who already own these products and removes them from the audience.
Note
This feature does not use or store any end user's demographic data.