Recommendation Engine
A recommendation engine, used synonymously with the term Recommendations, tracks the user’s behavior on e-commerce sites and displays products that the shopper would be most likely to purchase. Behind the scenes, recommendations makes use of algorithms and pre-built datasets to present the most relevant items for the individual shopper. Recommendations can be placed on the site at various places: home page, product details page, landing page, checkout page, and even in email.
Delivering recommendations provide significant benefits to an eCommerceeCommerce, also known as electronic commerce, digital commerce, or internet commerce, refers to the buying and selling o... More site, including:
- Increased conversion rates – Due to high product relevance, an eCommerce site with a product recommendation engine will have higher conversion rates than a similar site with no recommendation engine.
- Increased average order value and number of products/cart – Recommendation engines have been proven to improve the average order value and the number of products per transaction.
- Increased time on site and number of pages/visit – Due to higher content relevancy, shoppers tend to stay longer and visit more pages per session.
- Increased content personalization – Shoppers view more products that they are interested in, thus making it more likely they continue shopping.
- Increased targeted traffic – Recommending relevant products in emails or on advertising banners increases the amount of highly targeted traffic.
A modern personalization engine (which includes a recommendation engine) such as Reflektion employs machine learning algorithms including Collaborative Filtering, Association Rules Mining, Convolutional Neural Networks (CNN), K-Means Clustering, and Latent Semantic Indexing (LSI) to generate datasets for Recommendation bars such as “You may also like”, “Similar Items”, “People who viewed X also viewed/bought Y”.
In addition, within Collaborative Filtering, the Association Rule Mining approach generates several datasets for determining personalized product recommendations on eCommerce sites. These datasets include:
- Co-viewed: People who viewed the product also viewed these products
- Co-bought: People who purchased this product also purchased these products
- Viewed & bought: People who viewed this product, purchased these products
- Co-list: General list for “You may also like” widget