Recommendation | dtlr






In the new big data era, the data being produced in all areas of the retail industry is growing exponentially, creating a competitive advantage for those analyzing this data. As digitalization accelerated, the physical stores had to cope with a new channel, the e-commerce retailers. Major e-commerce sites set novel purchasing strategies: faster, occasionally cheap, and more targeted. However today the online retailers are in pursuit of new approaches; employing personalized recommendations to improve customer satisfaction by matching customers with relevant products at specific times and conditions.


DTLR Recommendation is developed based on latent factor model with Matrix Factorization (MF) method to incorporate personalized purchase behaviors with product/item attributes. Two algorithms form the basis for matrix factorization: Mix and Discover. Discover algorithm makes recommendations for products 'not purchased' by the customer, whereas Mix algorithm makes recommendations for products both 'purchased' and  'not purchased' by the customer.


Benchmarked with competitor algorithms, DTLR Recommendation stands out with the figures; 'Click to purchase rate' is higher by at least 14 percentage points and 'purchased amounts' are %52 higher than the alternative algorithms.