Sales Offer Optimization through Repeat Behavior Prediction in E-Commerce using Machine Learning (ML) & Artificial Intelligence (AI)
This invention provides an artificially intelligent solution for offer optimization in E-Commerce. There are various factors that affect the customer’s loyalty to brands and merchants and consequently determine their repeating behavior. Some of these factors include a website’s layout for online shopping and product placement in shelves in retail, customer service and available payment methods etc.
Retailers often come across the problem of how to design promotional offers in a manner in which customers who avail these offers become loyal and buy these products beyond the life of the promotional offer. There are many subjective approaches for designing promotional offers that, are being utilized by retailers. Unfortunately, they come with relatively less data to back them up and very little automation to execute such plans. Therefore, a need exists in this field to design offers that, best suit each of the individual customers by intelligently identifying their repeat behavior. Our work here can be broadly classified into three parts:
I) Predicting repeat behavior of customers for sellers. II) Predicting repeat behavior of customers for products. III) Designing offers by exploiting the information obtained through above two points.
In first stage, we identify the repeat behavior of customers for sellers, given only the transaction history of customers. The repeat behavior shows whether a customer will buy a product in the near future or not. Three types of data is required for this purpose i.e. transaction history of customers, customer and target entity pair with repeat label and product information. We have broadly categorized the features into three different categories i.e. customer related features, target-entity related features and the features related to interaction of customer and target entity. The target entity in the first part is seller and product in the second part. Once these two behaviors are learned, depending on the campaign cost, offers are placed on the products that come from part II and are sent to the customers that come from part I.