Web shops use recommender systems to help users find the products they find interesting in the large amount of available products online. An often used approach to do so is collaborative filtering. This method relies on historical user-item interactions and uses them to recommends products other users found interesting. Fashion is very reliant on quickly changing trends and personal preferences and requires a more personal and up-to-date approach. The focus of this research is to generate recommendations based on what products the user is currently searching for. It does this by detecting user behaviour based on the search scope of users and products user look at in the current session. Then new products are recommended by means of clustering new products to the most interesting products of the current session. This system was then compared with item-based collaborative filtering with an A/B test on the fashion platform Fashionchick.nl. It was found that traditional collaborative filtering was slightly more effective, but because of the small differences it is concluded that a behaviour driven recommender system are be promising and that more work is needed.