Creating a Data Source for Recommendations

The data source is a recommendation filtering algorithm that can be applied to the website, media channels, and mobile application.

Creating a data source is mandatory when setting up any recommendation type in eSputnik.

Automatic Data Source Creation for Triggers

When the Professional plan is activated for an organization with web tracking enabled, eSputnik automatically creates the predefined data sources required for recommendation triggers.

The data sources are created only once. If some of them already exist, eSputnik creates only the missing ones. A data source that was created automatically and then deleted is not recreated.

If a product feed has not been uploaded yet, the data sources are created with predefined required fields, such as product ID, name, and price. You can still create data sources manually for custom or other recommendation algorithms.

Creating a Data Source in Account Settings

1. Go to account settings:
2. In the menu on the left, select the section Data sources and click New data source → Recommendations for messages or website:

3. Select a recommendation algorithm.

• Recommendations based on visitor data. The recommended products will be unique for each visitor. If the visitor is not authorized, bestsellers of all products will be displayed.

• Recommendations based on product data. for pages that show a product or a product category.

• General recommendation algorithms.

In addition, you can add custom filtering rules for the block.

4. Specify a unique source name.

5. Configure the available data source settings, and click Save.

The data source settings page contains the following settings for all recommendation types:

• Rules for product algorithms: custom rules that adjust the algorithm output.
• Products count: the maximum number of products the data source can return. In the Max field, enter the upper limit for products in the recommendation result.
• Product preview shows only products that the system will send to recommendations. You set the appearance of product cards when you create recommendations.

Recommendation preview algorithms correspond to the website display algorithms: some data sources require a contact, product, or product category to show recommendations, while others show recommendations without input parameters.

Advanced Settings

For Recommendations based on visitor data, you can configure advanced settings for using data sources in messages. To open them, click Advanced settings on the data source settings page.

Advanced settings include:

• Filters: conditions for selecting products.
• Product exclusion conditions: conditions for excluding products from recommendations.
• Field transformers: rules for modifying product field values, for example, converting prices to the accepted format.
• Regular expressions: expressions for converting product field values to the accepted format.

Filters work differently depending on how the data source selects products:

• If the data source can generate recommendations, you can select rules for it or use rules that you have already selected. In this case, the system applies filters to the products it uses to generate recommendations and uses only matching products to build the recommendation set.
• If the data source returns products from the user's previous actions, such as cart or viewed products, you cannot select rules for it. In this case, the system applies filters to the products the user has interacted with, and the result includes only matching products.

The created data source will appear in the general list.

You can use the generated source when setting up recommendations for your site, app, and campaigns.