Recommendation Engine - Configuration

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

Scroll down to review available settings.

Configuring Similar Articles

While editing Recommendations you also have the possibility to specifically influence the determination of similar articles.

You can add fields for the determination of similar articles. If you go beyond the category and for example add the brand field, then only articles, which fit with the viewed product for all criteria, are returned – in this case only articles of the same brand. You must decide on your own whether it is sensible to add criteria additional to the categories.

If you only use a very narrow category level, you will get very similar, but possibly only very few, products. If you use a more broad level, you will get more, but possibly less fitting, products. Especially critical are linked categories: If there are sofas and armchairs in the narrowest category, then armchairs can appear as similar products to sofas.

Response Settings

Configure how many recommendations (1) are returned and how long the search can take at max (2).

The Recommendation Engine is fast enough to accomodate 20+ recommendations.

Personal Recommendations

If you have also licensed Personalization, you can also personalise recommendations. This means that a customer who likes Armani shirts will see offers for Armani ties, if shirts and ties are often sold together.

Just as with Personalization, you either choose a session-based method or one based on a shop-assigned user ID. With a session-based approach, FACT-FInder only uses clicks, add-to-cart ecvents and sals during the current session. A user ID enables you to tailor your recommendation to all events since creation of the user account.

First decide whether or not to personalise recommendations (1).

The Max per category value increased by factor (2) option multiplies categories. It governs, from how many categories recommendations are drawn.

This value is per default set to 10. Increasing it also increases the amount of necessary calculations and through that the time FACT-Finder needs to generate satisfactory recommendations. If the value is too low, it might not find any suitable products, so no personalisation occurs.

Matrix Settings

Relationships between categories are displayed as a matrix. Review and adjust them, on this page.

Matrix settings are essential, because the matrices, their weight and their self-reference have a strong influence on the Recommendation Engine's performance. The more strongly a matrix is weighted, the stronger its influence.

For example: a matrix accessing the brand field is weighted strongest. FACT-Finder now assigns the brand relationes managed by the matrix the highest priority when generating recommendations. Only then will it take categories into account.

You set the number of recommendations per matrix. This number should not be higher than the value set in Return Settings to facilitate quick recommendation calculations.

Self-references are recommendations aimed at the same brand or category. If you exclude self-reference, only other brand s or categories will be recommended.

Import Settings

Re-import the recommendation data to recalculate product/category relationships and generate the matrices under the menu point Recommendations.

Customise the settings for automatic and manual import. 

Page Contents

Chapter Contents