Real-time personalization of top online stores at the Prudsys personalisation summit 2017

Every year, Prudsys AG (addendum: in 2022 the company was renamed "GK Artificial Intelligence for Retail AG") presents the latest developments in the field of personalization through machine learning at a summit. A personalization can be, for example, a recommendation of another product through suggestions such as: "other customers also bought" or "the following might also interest you". In addition to the presentation of the new functions of the Recommendation Engine, the implementation at various customers will also be presented - in the form of short presentations.

While last year I only attended the event as a participant, this year we were asked by Prudsys to present the implementation at our customer Westfalia. Some important facts about the project have already been published in a case study.

I personally found the demonstrated concepts very interesting, as they showed that the recommendation system can be set up and used on the basis of a wide variety of data.

While many stores "only" use the purchasing and browsing behavior of users as a basis, some retailers already include parameters such as age and gender. Factors such as season or weather are also already being considered.
The whole thing can even be extended to the point where a camera at the point of sale can be used to determine a customer's mood and display appropriate recommendations based on this - this approach was presented by the company Sensape. However, such extensions are only possible if a sufficient number of customers or users are available in the system, otherwise the dataset is far too small.

Prudsys is currently working on a revised version of the Recommendation Engine, which is to be released at the end of this year. The presented screenshots and concepts show that especially the operation, as well as the evaluation in the backend will be in the focus of the RDE (Realtime Decision Engine). Here a significant improvement is definitely possible!

The reference solutions presented in the other presentations showed me that our current implementation and use at Westfalia is very good – the sales generated via the RDE show this as well. Nevertheless, I took away some suggestions that we want to implement in the next few weeks, such as the individualization of the order of the categories through the RDE.