Return Prediction

An online sales company had difficulties with their returns. As their returns were cutting heavily into their profits.

As their return policy was free returns and they bore the cost of shipping.

We were given all the data and asked to find solutions.

We implemented solutions to predict how likely the user was to return the product before they purchased it.

We also explored the data in depth along with shipping costs, and return costs to find relationships in the data that caused returns.

We discovered and proposed to the clients that particular categories are causing massive returns. Including socks and shoes, where the user orders different size shows. We then proposed a system to direct users towards a sizing chart for those categories. We also provided them a tool to generate a prediction of a probability that the user will return an item before they purchase the item with achieved an accuracy of 84%.