Configuring Recommendations

The product recommendation configuration possibilities include the following:

Algorithm Availability

Define availability of algorithm in Manage Recommendations and Recommendation Models.

Data Source Pre-Filters

Define filters to limit the data retrieved from a data source during model generation. Import custom fields to be used as pre-filters.

Recommendation Algorithm Defaults

Define default values for algorithm and data source pre-filter parameters.

Runtime Impressions

Aggregate or Transform Runtime Recommendation Impressions.

Runtime Parameters

Enhanced Recommendation Runtime Parameters

You can define the availability of algorithms in Manage Recommendations and Recommendation Models. Manage Recommendations contains a subset of features and algorithms that are available in Recommendation Models. If you make an algorithm available in Manage Recommendations, it is automatically made available in Recommendation Models also.

Step One

Navigate to Manage Your Solution to view the available apps.

1. User Story: The admin can control which algorithms are available to the recommendation applications tailored for the Marketing Expert and Business Analyst.
2. New Capabilities: Algorithms can be made available in any of the recommendation applications.

Select the Recommendation Algorithms application.

Step Two

Now that you're in the Recommendation Algorithms app, let's explore some of the available settings.

1. Usage of algorithms: These settings allow you to define the availability of algorithms in Recommendation Models and Manage Recommendations. Once an algorithm is set to be available in an application, the algorithm will be shown in the list of available algorithms once the user selects to add an algorithm in the model.
2. Benefits for customers: Determine which pre-filters best fit the users of the two applications and accordingly make them available. This will enable the users to see only what’s relevant to their role.

Step Three

You can define data source pre-filters that can be applied to an algorithm within Manage Recommendations and Recommendation Models. The pre-filters are data source type specific and limit the data retrieved from a data source during model generation.

1. Usage of pre-filters: Once a data source pre-filter is set to be available in an application, the pre-filters will be shown in the list of available data source pre-filters once the user selects to add a pre-filter in an algorithm.
2. User story: The admin can control which data source pre-filters are available to the recommendation applications tailored for the Marketing Expert and Business Analyst.
3. Benefits for customers: Determine which pre-filters best fit the users of the two applications and accordingly make them available. This will enable the users to see only what’s relevant to their role.
4. New capabilities: Pre-filters can be made available in any of the two recommendation applications.

Step Four

This area is where you can define the default values algorithm and data source pre-filter parameters.

1. Recommendation Algorithm Defaults: This feature helps the user by avoiding to add the same parameters each time an algorithm is used within a model by saving the parameters as default values.
2. In this example, when the Top Viewed algorithm is used, Interaction Type and Use Interaction Data will be used and values will be defaulted (i.e. SHOP_ITEM_VIEW and look back in the last 15 days).

Here is a list of parameters. You can define the following algorithm defaults:

Maximum rule size

User specified maximum item set size.​​

Maximum result set

User specified maximum result set to return.​​

Minimum confidence

User specified minimum probability that a combination of products are grouped together in a transaction. The value must be strictly greater than 0.0 and strictly less than 1.0.​​

Minimum lift

User specified minimum figure that is the ratio between the confidence and the support of the dependent items (expected confidence).​​

Minimum support

User specified minimum figure associated to a product association rule that describes the frequency with which two or more products are grouped together in a transaction. ​The value must be strictly greater than 0.0 and strictly less than 1.0.​​​

Score type

User specified method of ranking recommendations that is based on one of the following:

• Confidence
• Lift​​​
• Support

Generation refresh rate (in hours)

User specified number of hours after which an algorithm is regenerated. Once an algorithm has been added to a model, the algorithm parameters can be changed and saved. The updated parameters are then only applied within the context of that individual model.​

Data source pre-filters are applied during model generation to limit the data retrieved from a data source. The data source pre-filters available to an algorithm are contingent on the data source type assigned to the algorithm.

You can define the following data source pre-filter parameters:

• SAP Marketing Interactions
• Communication Medium
• Product Category
• Interaction Type
• Interaction Time Stamp
• Use Interaction Data
• SAP Marketing Interaction Contacts
• Target Group
• SAP Marketing Product
• Product Category

If you want to learn about custom recommendation algorithms, check out the resource below:

Blog post

Use Custom Algorithms in Product Recommendation.