AI Engine configuration

Before you can create your first recommendation, you must prepare a feed which will be the source of items displayed in the recommendations.

After that, you must select a recommendation type so the system can start the model training on the selected product feed and you can send relevant and precise recommendations.

Optionally, you can set default filters for each recommendation type. You can use these filters to define universal conditions an item must meet to be included in the recommendation, for example, that an item must be in stock in order to be displayed in recommendations.

The next steps concern selecting the attributes for the model training, the filterable attributes, and response attributes. All these steps are described in this article. After you perform them and initiate the model training, you can monitor the status. When the model is active, you can create a recommendation.

Preview of the model status


When you enable the model for the first time, it starts being trained. When the training ends, it receives the Active status which means you can use it. From time to time, the model is retrained, so it can work on the up-to-date data. This way, your recommendations are more relevant.

The status of the model is displayed in its settings and the table below contains all possible statuses a model can have.

Status Description
Active The model is enabled, trained, up to date, and ready for use.
Disabled The model is turned off.
Model queued for training The model is in queue for training. This status precedes the Training in progress status.
Training in progress Training or retraining (which occurs on a periodic basis). While retraining, the model still works and responds, but it might use outdated data.
Error There are two reasons for this status:
1. The model requires more data to be trained.
2. An unexpected error occurred while training the model - in such case contact the service desk to receive more information.
Disabled due to inactivity When the model is not used for 30 days in any way, for example, a recommendation based on this model is created and used in a dynamic content or you send API requests to the model, it is automatically turned off. You can turn the model back on by switching the Enable model toggle on.
Training paused due to inactivity When the model is not used for 10 days in any way, for example, you use a recommendation based on this model in a dynamic content or you send API requests to the model, the retraining is paused. If there is a request that uses the model (campaigns that use the model, or API requests to the model), the model retraining is re-enabled. The model still generates items in recommendations, but they are only up to date with the last model retraining.
Training paused Training or retraining of the model was paused manually (it’s possible only after a contact with the service desk). If it is trained for the first time, and the training is paused, the model will not respond. If the retraining of the model is paused, the model still generates items in recommendations but they are only up to date with the last model retraining.

Select product feed


The first step is selecting the feed from which recommendations will source the items. You can either select a catalog that contains a feed or use Google Merchant Feed.

WARNING: We recommend using Google Merchant XML instead of XML files due to the size limits (an XML file cannot exceed 10 MB).
  1. Go to Settings > AI engine configuration.
  2. Click Add feed.
    Result: A pop-up appears.
  3. Select the product feed you want to use.
    • A product feed uploaded to a catalog (available in Data Management > Catalogs)
    • Google Merchant
      Product feed uploaded to a catalog

      1. On the pop-up, select the type of catalog:
      2. From the dropdown list, select a catalog.
      3. Confirm by clicking Apply.
        Result: The selected catalog appears on the list in Settings > AI Engine configuration.

      Google Merchant

      1. Provide the following information:
        • the link to the Google Merchant feed,
        • the name of the field,
        • the type of the feed,
        • the frequency the feed is to be updated,
        • authorization type,
        • user name,
        • and password.
      2. Confirm by clicking Apply.
        Result: Feed appears on the list.

Configuring model settings for the selected feed


On the list of feeds, click the feed you added according to the Select product feed procedure.

Blank model configuration form
Blank model configuration form

Select catalog to be imported


The Item catalog section is filled out automatically after you performed the steps in the Select product feed procedure.

Select attributes for preview


You can define attributes, which values will appear in the Synerise platform preview.

Attribute for preview section
Attribute for preview section
  1. On the right side, from the dropdown list, choose the name of the attribute, which contains the value for the attribute on the left side.
    Example: If the title of your item is in an attribute that is called itemTitle, choose itemTitle from the drop down menu in the Response attribute column (right side), in the row that corresponds to Title in the Item attributes column (left side).
    Result: Values for the attributes that are chosen in the Response attribute column will be shown for products when previewing them in the Synerise platform.
Important: This setting is mutual for the Recommendations v2 preview as well as the AI Search Engine preview.

Select recommendation types and default filters


Select the recommendation models you want to apply for the selected product feed. Optionally, you can define the default filters for each model. You can use these filters to define universal conditions an item must meet to be included in the recommendation of a given type.

Note: Learn more about recommendation types.

In the Recommendation models section:

  1. On the recommendation model you want to select, click Show.
  2. Switch the Model enabled toggle on.
  3. Optionally, define default filters for the recommendation model.
  4. Repeat steps 1 to 4 for other recommendation models you want to select.
  5. Confirm by clicking Apply.
    Result: When the AI engine is configured, the default filters are available in the additional settings of recommendation campaign.
Important: If you selected Visual similarity or Similar items, you must define extra attributes (which are required for the model training) in the Training attributes section on the application interface.
What happens when default filters and filters in a recommendation campaign are mutually exclusive?

In such a situation, the filters defined in the recommendation campaign take precedence over default filters.

Select response attributes


A response attribute is any quality that can describe an item and it is visible to a customer (for example, the name of the item, price, description, size, and so on).

  1. In the Response attributes tab, click Show.
  2. Click Select attributes.
  3. Select the checkboxes next to the attributes which you want to display to the customers.
    Note:

    There are two types of attributes:

    • Textual attributes: These are all attributes that can have one value, for example, a color, an item name, a brand name, fabric, pattern, and so on.
    • Range attributes: These are all attributes that can have numerical values within a selected range, such as size, price, width, length, and so on.
  4. Confirm by clicking Apply.

Select filters


Optionally, you can define the attributes which you can use later to filter recommendations results. The attributes also become available in the Analytics module and filters.

  1. In the Filterable attributes tab, click Show.
  2. Click Select attributes.
  3. Tick the checkboxes next to the attributes which you want to use for filtering the recommendation results.
  4. Confirm by clicking Apply.
Important:

Attributes added to Filterable attributes will be automatically deleted if they fulfill all of the following conditions:

  • They are not used in recommendations for 30 days.
  • They were added to filterable attributes more than 10 days ago.
  • They are not used in a search or suggestion index configuration.
  • They are not set as filters or boosting rules in a recommendation campaign.
  • They are not set as a default filter in a recommendation configuration.
  • They are not used as additional filters in a recommendation API/SDK request.

The deletion of these attributes from Filterable attributes occurs daily.


An item link is an attribute of an item to which Synerise’s UTM parameters are added.

  1. In the Definition of item link tab, click Show.
  2. From the Attribute dropdown list, select an attribute that will be an item link.
  3. Confirm by clicking Apply.
    Result: Thesnrai, snr_content, and snr_id parameters are added to the URL of the item. For example: https://www.exemplary-shop.com./winter-shoes-camelbrown.html?snrai_campaign=QWERTY[…]e=&snrai_content=&snrai_id=123456789010305

Select training attributes


Important: Define the training attributes only if you selected Visual similarity or Similar items in the Recommendation types tab.
  1. In the Response attributes tab, click Show.
  2. Click Select attributes.
  3. Tick the checkboxes next to the attributes which you want to display to the customers.
    Note:

    There are two types of attributes:

    • Textual attributes: These are all attributes that can have one value, for example, a color, an item name, a brand name, fabric, pattern, and so on.
    • Range attributes: These are all attributes that can have numerical values within a selected range, such as size, price, width, length, and so on.
  4. Confirm by clicking Apply.

Select attributes to increase item variety


By using the distinct filters, you can define the number of items with the same value of the attribute (for example, a brand) that can be displayed in the recommendation frame.

  1. In the Attributes for distinct filters tab, click Show.
  2. Click Select attributes.
  3. On the pop up, from the list, choose up to 5 attributes.
  4. Click Apply.
  5. Confirm the settings of the tab by clicking Apply.
    Result: These attributes are available in the Distinct filter while creating a campaingn.

Switch AI Search for the selected feed


Optionally, you can enable the AI Search for the selected catalog or Google Merchant Feed.

  1. In the Applied search engines tab, click Show.
  2. Switch the Search engines toggle on.
  3. Confirm by clicking Apply.
😕

We are sorry to hear that

Thank you for helping improve out documentation. If you need help or have any questions, please consider contacting support.

😉

Awesome!

Thank you for helping improve out documentation. If you need help or have any questions, please consider contacting support.

Close modal icon Placeholder alt for modal to satisfy link checker