RECOMMENDATION ENGINE

USE CASE

THE CHALLENGE

A retailer with over 500k customers and 60k products was looking to increase cross-selling and up-selling on their newly developed app.

OUR SOLUTION

We combined the available raw data about customer's past behaviour, similarity to other customers and similarity between products and processed it into a smart format in Keboola Connection to increase the speed of the calculations. We used a number of recommendation algorithms (UBCF, IBCF, Deep Learning, etc.) and combined them together with business logic rules.

RESULTS

Customers can now find their favourite products toghether with recommended ones, either from the categories they already purchased (up-sell) or from the suitable new categories (cross-sell). Overall revenue increase has been estimated at 5-15%.

Joint project with BizzTreat and Keboola