References

E-commerce growth with smart recommendations

Written by Loihde Advance | 10.10.2016

Challenge

Talotarvike is one of the first e-commerce hardware stores in the Nordic countries. They sought help from Loihde for the automising product placement. With their e-commerce product selection of more than 25,000 items, effective and versatile linking is practically impossible to perform manually. Together with Loihde, smart recommendations were developed, providing a new and versatile method to find suitable products for customers, replacing manual linking.

Solution

The recommendation algorithm was developed by testing various scenarios and by analysing their results. Since the majority of customers do not login to the e-commerce store, recommendations cannot be made based on shopping history. Instead, the system has to use browsing history during the session. As the end result, a smart algorithm was created that performs product placement based on the browsing history of other visitors.

“The Loihde experts have rock-solid technical know-how, but they also understand business logic. Thanks to that, the concept was developed from a genuine business perspective. We wanted more shopping carts of higher value,” says Jani Molin from Talotarvike.com.

The solution itself was realised with open-source code. Our options were Recommenderlab library, in R, Mahout, which has been created in Java or a product such as PredictionIO. Since we wanted to have very influence on a very exact level, with our recommendation algorithms, we ended up realising the recommender Mahout, since Java is a more common coding language than R in the business world.

Results

In this case, the smart recommendations take advantage of browsing paths performed by other users, allowing the system to present more items of interest to the customer. The results tell us that the Loihde recommendation algorithm helps us to sell the same visitor amount more products, more often.

“Our e-commerce generates 5–10 % more revenue now, since implementing the smart recommendations lists. We get more shopping carts, more products and more valuable purchases. We are very happy with how Loihde realised the solution, and, above all, the end result,” says Jani.