Personalized recommendations have long been associated with the need to track user behavior and build user profiles. Product browsing history, previous purchases, clicks, or time spent on a website – all this data helps tailor an offer, but using it is not always possible or necessary. That is why we are developing our recommendation engine so that it also works effectively when a user does not consent to profiling.
This is one of the most important improvements in recent months and, at the same time, a solution that stands out on the market. The engine does not make the quality of recommendations dependent on creating an individual user profile. Instead, it can automatically switch to a mode based exclusively on the context of the products currently being viewed
Learn more about Quarticon’s AI recommendation engine:
Two modes of operation for a single engine
Depending on the consents available and the needs of a given implementation, the engine can operate in two complementary modes.
In personalized mode, recommendations may take into account information related to a user’s preferences and behavior – of course, when the user has given the appropriate consent. This makes it possible to create a more individualized experience, for example by presenting products similar to those the user has viewed previously or tailoring suggestions to their interests.
The second mode works without profiling. In this case, the engine automatically uses previously trained models and primarily analyzes the relationship between the product being viewed and other products in the offer. Recommendations may be based, among other things, on similarity, complementarity, popularity, or frequently occurring relationships between products – without creating an individual profile of the person visiting the site.
Recommendations tailored to the product, not the person
The most important difference is that the system does not need to know who the user is or which products they viewed previously. It only needs information about which product is currently displayed on the screen.
If a customer is browsing a specific shoe model, the recommendation engine can suggest products suited to the current product, based on the past sessions of other users. This approach helps maintain the usefulness and higher effectiveness of recommendations even during a first visit, when the user is not logged in, or when they have not consented to profiling. The user receives help discovering the offer, while the store can continue supporting the sale of similar and complementary products.
Automatic switching without complicating the experience
The change in operating mode takes place automatically. It does not require any additional configuration on the user’s side or a manual choice of how recommendations are presented. The system recognizes which data it can use and selects the appropriate operating model.
This is also important from a business perspective. A store does not have to give up recommendations simply because some visitors have not consented to profiling. The engine can still support pseudo-personalized navigation through the offer, increase product visibility, and make it easier for customers to find what they are looking for.
One solution, different needs
The new feature addresses the growing need to combine business effectiveness with greater user control over the use of their data. Personalization remains available wherever the user consents to it, but it is not a prerequisite for activating recommendations.
For owners of online stores and e-commerce platforms, this means greater flexibility. They can use AI recommendations in both personalized and pseudo-personalized models, without having to switch to contextual recommendations, which are significantly less effective.
It is also worth remembering that the recommendation engine personalizes or pseudo-personalizes the offer in real time for up to 100% of website traffic, which sets it apart from marketing automation or CDP systems that require a user profile (in a traditional e-commerce store, this corresponds to approximately 5% of website traffic). Recommendation engine: 100% > marketing automation 5%. We have written more about this, among other places, here: recommendation engine vs. marketing automation.
An innovation that sets a practical direction for the development of recommendations
The ability to operate without profiling is not merely a limited version of traditional personalization. It is a separate, deliberately designed way of recommending products, based on an analysis of the offer and the context of the visit. As a result, the Quarticon recommendation engine combines two approaches: personalization available with the user’s consent and effective pseudo-personalized product recommendations that do not require user consent.
