A good friend is celebrating her birthday in a few days. As a thoughtful person, as I was looking for a gift, it occurred to me that on our last visit to the lake, she didn’t have a beach bag along for her swimming accessories, towels, etc. She used two cloth bags to carry her things.
So I started looking online for a beach bag in a design and color that match my friend’s taste. I found a bag with a nautical design and the right size that appealed to me, but it didn’t have a zipper inside. I didn’t want a bag without this feature. In addition to the bag I liked, the online shop displayed other product recommendations. For example, under the heading “Matching items”, I found a selection of products beyond the current “Bag” category, such as sandals and towels. In addition, I was pleased to find more bags with the same look under the heading “Similar products”. Here I quickly found a beach bag with blue and white stripes in the right size and design, and it had a zipper inside. In just a short time, thanks to the product recommendations, I had the right beach bag in the shopping basket. With the ideal birthday present, I’m looking forward to celebrating with my friend.
Recommendations based on similar images – how is that useful?
Behind the product recommendations with the same look is the AI software prudsys Realtime Decisioning Engine (prudsys RDE for short), which calculates matching recommendations based on image similarities. Customers are very accepting of these types of recommendations. The result is increased customer satisfaction. At the same time, you achieve revenue uplift and reduce shopping basket cancellations.
For example: Your customer is looking for a special ring. Unfortunately, you currently don’t have any by this special brand in stock. What can you do? To prevent out-of-stock situations from leading your customers to leave the online shop, you can suggest similar rings to them. Based on image similarities, the AI software also calculates which rings are close in shape and design to the item in the search.
This new type of search could even be expanded in the direction of visual search. A scenario: During dinner with friends, a woman sees a stylish watch on her friend’s arm. She takes a picture of the watch with her smartphone so that she can use the retailer’s messenger service or app to get product recommendations that look similar to the watch she photographed. Or the woman is looking for bracelet that matches the look of the watch she photographed. This scenarios can also be played out in real time with the prudsys RDE AI solution.
Image similarities – how does it work?
An AI service for image similarities within the prudsys RDE provides recommendations for you based on similar images. The AI service regularly checks all product images for similarities. A multi-layer neural network trained in image analytics analyzes the product images and recognizes similarities. The detected similarity is mapped based on the strength of the relationship and then sent to the prudsys RDE as a model. The recommendation type for image similarities uses the relationship model to generate relevant recommendations.
You can use a number of non-visual settings to configure your recommendations in detail:
- You can specify that the AI service should only compare images of products that are currently available.
- You can determine a price corridor for the recommended products, e.g. depending on the initial product.
- You can set the AI service so that it only recommends products from the same category.
- You can exclude recommendations from display that fall below a certain threshold of relationship strength.
Would you like to learn more about how you can profit from using recommendations based on image similarities within your customer journey? Get in touch with your prudsys contact or let us know at