Convermax product recommendations became more relative and smart

When it comes to selling online, every merchant at least once heard about a recommendation system – a buzzworthy technology that many people would like to have on their websites. We were providing our customers with a simple product suggestions section for quite some time. Now our search engine is enhanced with a brand-new recommender system to make product suggestions even more relative and effective for your business.

How new recommendations work

In the past, our recommendations were based on the search. The search engine analyzed the attributes of the product the customer was viewing at a particular moment. Then it used the set of attributes to create a search results page with similar products to the currently searched item.

The new Convermax recommender system is based on machine learning. It utilizes user event history related to items and metadata about users and items. This data includes datasets as such as monitoring clicks, view times, purchases, etc. Product recommendations on the client’s store website are generated based on user-based, item-based, and ensemble methods. 

User-based filtering rests on the premise that similar users would like the same products. Here is how it works. There is an active user on the store website. We have his history of actions and the history of similar users. Based on this data engine can compare and suggest which products to recommend to the currently active website visitor.

Item-based filtering focuses on product similarity. If the engine knows that the users once bought or were browsing Nike Air Max 2, it can suggest to them an equivalent of Air Max models.

We use an ensemble of 5 different models. Ensemble modeling is a process of creating several models to make final predictions of what potential customers might like. So, it clusters different data that the engine knows about users’ activities and preferences. Then it creates multiple models and decides which model is the most correct one in a particular case to provide predictions. 

Here you can see how it works with accessories on our client’s Marjorie & Co website: 

Right now, we have a limited offer for those merchants who want to get brand-new product recommendations implemented into their stores. We won’t charge you any extra fee for the monthly plan you’re already on.

Seems like something your store needs? Schedule a call with Alex to discuss it.