Recommendations for related products may be misleading

Recommendations two ways

Did you ever think about that when you visit a product in a store, that product may very well be related to other articles in the store. It is often the case that one product may be also be related to others depending on the context where it is used. If you provide the wrong recommendations, you may risk loosing a visitor instead of gaining a customer.

In my store, we sell LED-lighting products and often have products that are both applicable for kitchen and bathroom. This means that for those of our visitors that come visit us when browsing for LED-lighting for a bathroom are highly interested in getting primarily bathroom related recommendations on what is similar to what they are looking for.

Many solutions today, including the one we are currently using with recommendations won’t be able to tell the difference and will proudly present relations to the visitor based on the products that they visit and whenever they take a step into a product page that relates both to kitchen and bathrooms, we may lead the visitor on the wrong track.

This is not just something that happens in theory, we can see that from our customer service tickets, that it’s common to have questions based on the automatic recommendations where there a mix of products between several areas that they would like to use in their future redecoration project.

So there is still facetted navigation when the customer can set a filter on what to look for. This filter often applies for what is ON the product and not what is RELATED to the product. This means that the recommendations will become even more misleading.

The solution we are looking into is personalization and how to start using also this information that we know about the customer to make sure that the information that we present to them are more relevant based on the navigation history. We need to hit the spot right away, since we don’t have that many clicks to figure out what kind of person they are. In order to do this, we are in the middle of tagging our products in the backend PIM system for a set of areas of interest and connecting the same tags to the information pages so that we will get a quicker way of figuring out what our visitor is looking for. It must also be possible to re-evaluate the visitor really quickly if we are on the wrong track.


The following two tabs change content below.

Fredrik Gustavsson

CEO, IT-consultant at Jolix AB
Software and integration magician with an interest in how to improve commerce in a multi channel environment. Owns his own IT-consultancy business and runs an e-commerce store that has a physical store. Will write blog posts on how to take his own store into the future.