Pages

Learn how to create recommendations in the Frosmo Platform:

Before you start

Recommendations typically rely on data collected from the site, but they can also use data that is directly fed or imported to the Frosmo Platform. Before you can create a recommendation, you must have the necessary data tracking implemented, or you must otherwise fetch and store the data to the Frosmo back end.

For example, a retail product recommendation relies on product data and transaction data. The product data gives you:

  • The product information, such as name and price, that you display in the recommendation
  • The product view statistics that tell the recommendation engine which products get the most views and which products visitors typically view together

The transaction data, in turn, tells the recommendation engine how well different products sell and which products visitors typically purchase together.

You can use either the product view statistics or the transaction data, or both, to drive how product recommendations are generated. But you need to first get the data:

  • For the product data, you must have product tracking set up on the site, or you must use a product data feed or file import to store the data to the Frosmo back end.
  • For the transaction data, you must have transaction tracking set up on the site.

Once the data is accumulating in the Frosmo back end, you can start generating product recommendations from it. To check whether the data is coming in from the site, check the site status.

Getting the data required to generate product recommendations

Figure: Getting the data required to generate product recommendations (click to enlarge)

For more information about tracking data on a site, see:

If you want to use a data feed or file import, contact Frosmo support, as only a Frosmo team can currently implement these solutions.

Creating a recommendation

A recommendation is the end product of multiple Frosmo components working together. Some of the components are mandatory, while some are optional.

Components of a recommendation coming together on the page

Figure: Components of a recommendation coming together on the page

You can use a workspace to develop and test the modification, placement, and optionally template of the recommendation.

To create a recommendation:

  1. Plan the recommendation. This involves answering questions such as:
    • What business goal do you want to achieve with the recommendation?
    • What type of recommendation do you want to create?
    • What items do you want to recommend?
    • Who do you want to target with the recommendation?
    • When, where, how do you want to display the recommendation on the site?
  2. Create a recommendation configuration. The recommendation configuration defines the logic and settings for generating the recommendation data.
  3. Optionally, create a template for the recommendation. If you want to use the same recommendation in multiple modifications, create the recommendation as a template.

    For example, if you want to create a customizable recommendation carousel that can be used in different parts of the site, create the carousel as a template with the relevant content options, and then use and configure the template in individual modifications.

  4. Create a modification for the recommendation. Create the necessary placement and optionally segments for the modification.

    The modification both defines the recommendation content and displays that content on the site. You can define the content either directly in the modification as custom content for a variation, or in a template that you then use in a variation (see step 3 above).

    Irrespective of where you define the content, the content must:

    • Create the web page element for the recommendation. This is the static part of the recommendation: the HTML, CSS, and JavaScript that together define how the recommendation looks and behaves on the page.
    • Fetch the recommendation data, and populate the web page element with that data. This is the dynamically generated content of the recommendation: the details of the recommended items. You fetch the recommendation data using the Recommendations API.
  5. Test the recommendation. You can use variation preview or test mode.
  6. Publish the recommendation. You do this by activating the modification (or, if the modification is already active but in test mode, by disabling test mode for the modification).

The recommendation is now live on the site. You can freely edit the different components to improve or otherwise change how the recommendation looks and works.

For an end-to-end example of creating a product recommendation, see Example: Recommending products purchased together.