Overview of various algorithms used in the Acoustic Personalization Product Recommendations.
Acoustic Personalization - Product Recommendations provides the following algorithms that you can choose from: Most popular, View to view, Buy to buy, View to buy and Recently viewed.
Apart from the primary product recommendation strategy, you can optionally create a Fallback strategy. For more information, see Creating Fallback strategy
For each product recommendation model, you can specify the number of products to be returned by the model, and also the number of products to be displayed on the zone. By default, the algorithm returns 10 items. The channel developer should determine how to render the products returned by the model, on the zone.
- View to View and Buy to Buy
View to View and Buy to Buy are collaborative filtering algorithms based on the wisdom of the masses. These algorithms use the Product View and Cart Purchase events to generate product recommendations on the zone.
View to View algorithm uses the event data to provide a list of other products viewed by those who had viewed the current product. For example, if you are viewing a model A of running shoes, the View to View algorithm returns a list of other products that were viewed by other visitors who had also viewed the shoe model A.
Buy to Buy algorithm uses the event data to provide a list of other products purchased by those who had bought the current product. For example, if you are buying the model A of running shoes, the View to View algorithm returns a list of other products that were purchased by other visitors who had also purchased the shoe model A.
- View to Buy
View to Buy is based on the Frequent Pattern Mining algorithm. It finds the pattern between the Product View and Cart Purchase events to generate product recommendations.
View to Buy algorithm uses the event data to provide a list of other products bought by those who had viewed the current product.
For example, suppose a visitor viewed products A, B, C, D and finally purchased the product D. The View to Buy algorithm captures this behavior. If another visitor views the products A, B, C, or D, then the product D is shown in the recommendations to the visitor.
- Most Popular
Most Popular algorithm determines the most popular products based on various parameters, such as the product view count or product purchase count or revenue or margin. When creating the product recommendation strategy, you must specify the parameter to be used.
To use the Most Popular algorithm based on revenue or margin, ensure that you have uploaded this data to Acoustic Personalization using the Product Data Catalog. Further, the recommendations are generated based on the selected measure of popularity. If the measure of popularity is revenue or margin, then the products purchased by the customers and having highest cumulative revenue or margin are recommended.
If measure of popularity is View count, then products recommended will be based on the product view events flowed through Acoustic Personalization.
- Recently Viewed
Recently Viewed algorithm tracks the products that are recently viewed by the visitors on the channel and saves it in the Local Storage of the web browser.
When a visitor arrives at the channel (which has "Recently Viewed" zone is configured on the homepage), there will be no products to be displayed for his first visit. If the visitor clicks on "product A" and returns to the homepage, then "Product A" will be displayed on the "Recently Viewed" zone. The product details, such as the image, price, etc. are fetched from the product catalog.
When configuring the Recently viewed model on your channel, you need to specify only the Lookback time frame as the parameter. The model tracks the recently viewed products by a visitor within the specific Lookback time frame, irrespective of the user sessions on the channel.
You cannot create a Fallback strategy for this algorithm.
If a channel visitor has not yet viewed any products on the channel, the algorithm will not have any input data to show product recommendations on the Recently Viewed zone. Here are some of the ways you can handle this scenario:
- Hide the Recently Viewed zone until Acoustic Personalization sends recently viewed products based on the visitor browsing behavior
- Display an information message, for example: "There are no recently viewed products."