The Product Recommendations recipe enables you to provide personalized product recommendations that are tailored to your customer’s needs and interests. With an accurate prediction model, a customer’s purchase history can provide you with insight on what products they may be interested in.
In modern day, a retailer can offer a multitude of products, giving their customers a lot of choices which can also hinder their customers search. Due to time and effort constraints, customers may not find the product they desire, resulting in purchases with a high level of cognitive dissonance or no purchase at all.
The Product Recommendations recipe uses machine learning to analyze a customer’s interactions with products in the past, and generate a personalized list of product recommendations quickly and effortlessly. This optimizes the product discovery process and eliminates long, unproductive, irrelevant searches for your customers. As a result, the Product Recommendations recipe can improve a customer’s overall purchasing experience, leading to higher engagement and stronger brand loyalty.
You can get started by following the Adobe Experience Platform Lab tutorial (see Lab link below). This tutorial will show you how to create the Product Recommendations recipe in a Jupyter Notebook by following the notebook to recipe workflow, and implementing the recipe in Experience Platform Data Science Workspace.
This recipe uses custom XDM schemas to model the input and output data:
The Product Recommendations recipe utilizes collaborative filtering to generate a personalized list of product recommendations for your customers. Collaborative filtering, unlike a content based approach, does not require information about a specific product but rather utilizes a customer’s historical preferences on a set of products. This powerful recommendation technique uses two simple assumptions:
This process is broken down into two main steps. First, define a subset of similar customers. Then, within that set, identify similar features among those customers in order to return a recommendation for the target customer.