Recommendation Engine - Configuration

Location

The Recommendation Engine Module offers recommendations based on product and category interrelations, as well as on field-based criteria.

Navigate More settings… → Recommendations to access the module:

The Recommendations heading tab should be active:

Scroll down to review available settings.

Settings types

Response settings

Provides an ability to configure the maximum number of product recommendations that the Recommendation Engine outputs and the maximum time it spends searching for recommendations.

The Recommendation Engine is so fast that 20 recommendations or more can be called.

Personal recommendations

If the license for the personalization module from FactFinder in addition to the Recommendation Engine is active, you can also have recommendations personalized for your customers, provided that corresponding products are available.

As with personalization, either a session-based method is selected here (FactFinder NG personalizes on the basis of the clicks, shopping cart additions, and purchases made in the session), or personalization of recommendations on the basis of a user ID assigned by the store. With a user ID solution, products purchased since the start of personalization tracking are also included in the personalized recommendations.

First, specify whether the recommendations should be personalized.

The increase max per category by factor is the factor by which the categories are multiplied. It controls from how many recommendations in a category FactFinder NG should pick out personalized ones.

By default, this value is set to 10. Do not set it too high, because this increases the computing effort - and thus the time required - for FactFinder NG to expand the recommendations to the point where personalization makes sense. If you set the value too low, it is possible that no products corresponding to the preference will be found. In this case, personalization will not take place.

Matrix settings

The individual category relations of a level are displayed in the form of a matrix. Here you can view the determined category relations and, if necessary, change them according to your ideas.

The matrix settings are particularly important because the matrices themselves, their weighting and the self-relation have a great influence on the behavior of the Recommendation Engine. The higher a matrix is weighted, the greater it's influence. For example, if a matrix that accesses the brand field is the highest weighted, FactFinder NG will give the brand relationships defined in this matrix the highest priority in recommendation determination. Only then come categories, for example.

ADD MATRIX : here one more matrix can be added (visible in the form of a matrix under Recommendations/Matrix). The example in the screenshot has 2 matrices defined: one for the field Brand and one for the field Category.

Field : in this column it is possible to set the related field of a matrix.

Weight : some of the pre-configured sensible values can be set here for the Weight value . Usually here the value "Normal" works best.

Weight value : these values will be set automatically when a value for was chosen. An arbitrary value (use carefully!) can be chosen if Weight has the value "Expert".

Max. products per value : there’s an ability to set an upper bound for the number of recommended products (in a recommendation result) that are recommended according to a field.

Self relation : if Self relation is enabled, for example for the field Brand, it can happen that the recommended (according to the field Brand) products have the same brand as the product which is the base for the recommendation. If Self relation is disabled only different brands are taken into account for the recommendation.

Copy matrix setting to a different channel : clicking this icon opens a dialog which makes it possible to copy a matrix setting to a different channel.

Delete : click this button to delete this matrix setting.

Import settings

By re-importing the Recommendation data, the product and category relations for the recommendations are recalculated and the matrices are created under the Recommendations menu item.