Rules are a kind of segments for products based on their parameters or on the behaviour of users who interact with the products. In the hierarchy of the recommendation service, rules are in the first place (Product Catalogue - Rules - Algorithm - Filters). Thus, if you add a rule, an algorithm and a filter to the recommendation block, the rule will be triggered first, and if there is space left in the recommendation block, it will be filled with products from the algorithm output, filtered by the selected filter.

Rules are self-sufficient, they do not require mandatory algorithm selection, as they work independently of them.

Rules come in two types:

• by human actions
• by product parameters

Human action rules are based on the user's purchasing activity: it is possible to recommend or not to recommend certain goods depending on the fact that each particular person looked at/not looked at/bought/not bought. For example, if you know that after buying a dishwasher, your customers usually buy products for it, you can set a condition of the form: “If bought from the “Dishwashers” category for all time, then recommend random 3 products from the “Dishwasher detergent” category”.

The rules for product parameters are based on the property values of the products in the product catalogue: you can define the relationship up to the point where you can specify that you should always recommend product M for product N. If you think that in your product catalogue, for example, if you have two categories of products that should not be related to each other in any way (for example, English textbooks and Chinese textbooks), you can set conditions like: “To the goods from the category “English textbooks” do not recommend from the category “Chinese textbooks”.

You can create rules in the Recommendations ➔ Rules.

Up to 10 conditions of both types can be set in one rule. Let's take a closer look at the customisation options.

By human actions

1. Operators by user activity (bought/not bought/looked at/not looked at)
2. Selection of goods (specific goods/goods from certain categories/goods with a certain value of a certain parameter)
3. Period for which the specified human activity should have occurred (all time/specific dates)
4. Recommendation operator (recommend/not recommend)
5. Selection of products (specific products/goods from certain categories/goods with a certain value of a certain parameter)
6. Number of products (if ”recommend from a category or recommend with a parameter“)

By product parameters

1. Product selection (specific products/goods from certain categories/goods with a certain value of a certain parameter)
2. Recommendation operator (recommend/not recommend)
3. Product selection (specific products/goods from certain categories/goods with a certain value of a certain parameter)
4. Number of products (if ”recommend from a category or recommend with a parameter“)

The settings of both types of rules overlap. We load the list of products, categories and parameters from your product catalogue . product catalogue.

In each recommendation block you can use up to 5 rules, each of which can consist of a maximum of 10 conditions. That is, a maximum of 50 conditions can be applied in one block. The same rule can be used in different recommendation blocks (if you change it, it will change in all blocks).

You can add rules to an active and working recommendation block. In this case, it will take some time to prepare them (during this time the block will give out products according to the algorithm as before), and after the rules are calculated, they will be automatically built into the output.