Return Prediction
Published:
An online sales company had difficulties with their returns, as free-return policies were cutting heavily into profits.
All available data was analyzed and solutions implemented to predict how likely a user was to return the product before they even purchased it.
An in-depth exploration of the data alongside shipping costs and return costs revealed key relationships. A major finding was that particular categories such as socks and shoes were causing massive returns, as users ordered multiple sizes. A system was proposed to direct users towards a sizing chart for those categories.
A prediction tool was provided that estimates the probability a user will return an item before purchase, achieving an accuracy of 84%.