An up-and-coming eCommerce brand has been working on improving their website user experience by showing relevant products immediately when visitors land on their homepage. The problem is that the company’s recommendation models are based exclusively on their internal data, limiting their effectiveness when dealing with new users, and offering limited results to existing ones.
Our customer built two parallel models to improve their recommendation engine — one for existing users, and one for cold-start recommendations. The models were built on smarter user profiles based on enriched datasets.
The existing user recommendation model uses historic purchase data, as well as users’ clicks, pages visited, and other website data, and connects it with data from Explorium, including:
The cold-start user recommendation builds profiles for new users with no prior history on the site. It connects with Explorium’s Enrichment Catalog to take advantage of the following data:
The result has been a 17% increase in per-session value for each visitor to the site and improving overall revenues for the company’s website by 4% on a monthly basis. More importantly, the company has been able to make its website more dynamic depending on the visitor, whether new or returning.
