Skip to main content

Audience Recommender

Under Review

This section reflects the previous release and will be updated soon. If you need to use this feature before documentation is updated, please reach out to your Client Success Representative or Technical Solutions Architect.

note

This feature may not be available in your version of Console depending on configuration.

Creating audiences — collections of members with similar or the same attributes or affinities — helps marketers more accurately target similar members with suitable offers. An audience can be built manually, but another option is to have an audience built by a generative AI engine that responds to an audience description entered by the user. In ES Loyalty™, this is referred to as the Audience Recommender.

The Audience Recommender in the ES Loyalty platform allows for the creation of member audiences using natural language prompts. It uses a generative AI engine to generate audiences based on descriptions provided by the user. This tool is particularly useful for marketers, as it helps target members with similar attributes or behaviors more effectively. By describing the features of a desired member audience, the Audience Recommender builds that audience using member data and returns a description of the logic used to capture that audience for the user to review and verify. Once the audience meets the specified requirements, it can be named, saved, and used in conjunction with offers.

The Audience Recommender can work with a diverse set of member data to create the desired audience, including:

  • Standard profile information: Such as city, province or state, country, gender, postal or zip code, and so on.

  • Past purchase information: Spending on particular products as defined by department, group, vendor, category, sub-category, or brand (either name or code).

    • Note: The feature can answer product ID (SKU/GTN) queries with some success. However, this isn't an officially supported use case as it hasn't been coached or vetted by QA.

  • Extended member data: Additional information provided by and relevant to the client.

  • Partner linking information: Information about which partners the member is linked to and details of those links.

  • Member point balance: The point balance on the member account.

  • Offer engagement: Member engagement with specific offers and types of offers, including accepting and completing offers, accepting offers at a particular time, completing a specified number of offers, accepting or completing any offer, completing offers with a particular reporting identifier (such as "Flyer"), or multiple conditions (for example, a member who didn't accept an offer in 2024 but has in 2025).

tip
  • As with all generative AI systems, the Audience Recommender may display inaccurate information, so review responses carefully.
  • If the initial response isn't satisfactory, a useful strategy is to submit the same prompt again. Because the generative AI regenerates the response each time, the result may differ from the initial response.

To use the Audience Recommender, describe the features of the member audience you want to build. The Audience Recommender builds that audience based on member data using a generative AI engine and returns a description of the logic used to capture that audience for you to review and verify. When the audience meets your requirements, you can name and save it, then use it with offers.

The feature only returns active, registered members, in alignment with the audience builder. Unregistered accounts don't receive emails and don't have an account to log in to view offers, so they don't belong in these marketing-oriented lists. Historical data includes data generated up to the previous day, but not the current day's data.

Set permissions for the Audience Recommender​

note

This procedure applies only to client administrators in the loyalty program or to Exchange Solutions personnel with administrative rights such as a Technology Solutions Architect (TSA).

  1. In the Console's header, navigate to Users > Roles.
  2. Select a role for Audience Recommender users. To create a new role, see Roles.
  3. Ensure that the user role assigned allows the following permissions: all Member Management > Audience permissions (including all Audience Recommender permissions).
  1. Click Save Changes.

To open the Audiences section of the Console:

  1. Select Membership > Audiences. If required, select a business unit in the sub-menu. The Audiences page opens.
  2. Click the Audience AI button.

View, Edit, Enable, or Disable an Audience​

To view, edit, or enable/disable an audience, see the instructions in the Manage an Audience topic. Note that an audience generated using the Audience Recommender can't currently be duplicated.

The sections below outline how to generate audiences in general and based on specific factors using the Audience Recommender:

Specific Prompt Scenarios​

The Audience Recommender coaching script has been optimized to support specific prompt scenarios. Optimizations include the following:

  • Consistency coaching: Better support for prompts such as "likely to buy based on past purchases."
  • AND/OR coaching: Reinforces interpretation of distinctions between AND and OR conditions. For example, "Members who have purchased from Beauty and Snacks" returns only members who have purchased from both categories.
  • Offer Code: Defined as the value of the same term as used in the Exchange Solutions data warehouse. Distinct from "Offer ID."
  • Point Balance: Can look for members who: (1) do or don't have balances that are greater than, less than, or between specified point balance values; or (2) are in the top 10% of balance holders, or who do or don't have a zero balance.
  • LTC Meaning/Synonyms: Allows prompts to relate a Load to Card (LTC) offer to the offer activation.
  • Redemptions: Provides appropriate error messages for redemption offer prompts that can't return a legitimate audience.

Generate an Audience with Audience Recommender​

The Audience Recommender accesses a generative AI engine to take an audience description entered by the user and determine which members in the database meet the criteria in the description. The Audience Recommender generates a completed audience based on the prompt.

To create an audience using Audience Recommender:

  1. Select Membership > Audiences. If required, select a business unit in the sub-menu. The Audiences page opens.
  2. Click the Audience AI button.
  1. The Audience Recommender opens. Generating the recommended audience can take up to a minute per prompt.
  2. In the Input text box, enter a prompt or edit a generic prompt to describe the features of the audience in plain language. For example: "All males who spent over $100 on vitamins last year." The query must be between 8 and 2,048 characters long. Alternatively, click one of the prompts in the Example Prompts section, this appears in the description text box and can be edited to suit your needs.
  3. Once you've completed the prompt, click Generate Audience. It may take 20 to 40 seconds to generate the response. Once complete, the plain language interpretation of the input description is shown in the Description field.
  1. To view the SQL generated from the query, select Enable query preview. A field containing the SQL opens. Feature availability depends on permissions settings.
  2. To see an estimate of how many members are currently included in this audience, view the Targeting Results field. This number may change over time as members meet or no longer meet the criteria.
  1. To provide feedback on the Audience Recommender results, click the thumbs up button (results were satisfactory) or thumbs down button (results didn't meet expectations). Input from these buttons is monitored. These buttons aren't available for some clients.
  1. Once the description has been refined to your specification and audience generation is complete, click Save Audience. In the Save Audience window, enter the audience ID (3 to 128 alphanumeric characters and underscores only) and the audience name (3 to 128 alphanumeric characters, allowed symbols .@+-$&', French letters, and spaces). Click Save Audience.
  1. If you use an audience ID that has already been used, an error message appears and you'll need to change the audience ID. You can reuse an existing audience name as long as the audience ID is unique. The Audience Recommender page closes and you are returned to the Audiences page. The published audience is shown in the list, along with an icon next to the audience name indicating that it was generated using AI.

Generate an Audience Based on Offers​

One of the query types for which Audience Recommender has been optimized involves offers — for example, which members have accepted a particular offer or accepted a vendor-funded offer.

This feature allows offer engagement data to be used in offer targeting. For example, members can be targeted if they meet one of the following conditions (each followed by an example):

  1. Accepted Basic: Members who accepted a specific offer ID. For example: "Members who accepted offer ID_FREQ_123."
  2. Completed Basic: Members who completed a specific offer ID. For example: "Members who completed offer ID_TIRES_456."
  3. Accepted or Completed Basic: Members who accepted or completed a specific offer ID.
  4. Time-based: Members who accepted or completed a specific offer on a particular date. For example: "Members who completed offer ID_FREQ_123 on October 1, 2024."
  5. Offer Counts: Members who have completed at least a specified number of offers in a given period. For example: "Members who have completed at least 10 offers in the past month."
  6. Any Offer: Members who have accepted or completed any offer ever, or any offer in the past month; or members who have not accepted any offer in the past month. For example: "Members who have accepted or completed any offer in the past month."
  7. Multiple Conditions: Members who meet more than one condition, such as accepting or completing offers within specific timeframes. For example: "Members who have not accepted an offer in the first three months of this year but did accept an offer during the past five months."
  8. Reporting Identifier: Members who have completed an offer based on an associated reporting identifier. For example: "Members who completed an 'ES Loyalty Boost' offer in the past 5 weeks" or "Members who completed a 'Flyer' offer."

By including member engagement data in targeting, marketers can create a virtuous loop to reinforce desirable member engagement behavior and support repeat sales.

Generate an Audience Based on Products​

The Audience Recommender has knowledge of all product categories, sub-categories, and brands that have been loaded into ES Loyalty, with the exception of those that haven't been purchased in the last 18 months. The Audience Recommender can also use product departments, family groups, and vendors in queries. It has also been enhanced to use codes rather than human-readable names for any of these product identifiers (for example, "Department 0123") and, for one client, to allow units, litres, and/or NACS codes in queries.

This feature allows product data to be used in queries:

  1. Categories: Used to determine members who did or didn't purchase within a specified product category; can include category name or number. For example: "All members who purchased batteries in the past year" or "All members who bought from category code 01234 in the past year."
  2. Sub-categories: Used to determine members who did or didn't purchase within a specified product sub-category; can include sub-category name or number. For example: "All members who purchased rechargeable batteries in the past year" or "All members who purchased from sub-category codes 00123, 00456, or 000789 in the past year."
  3. Brands: Used to build an audience from members who did or didn't purchase a specific brand; can include brand name or brand code. For example: "All members who purchased Duracell batteries in the past year" or "All members who did not buy brand 245070 in the past year."
  4. Department: Members who accepted or completed offers for products categorized in a particular department; can include department name or number. For example: "Members who bought cosmetics in the past six months" or "Members who bought from department 035354 in the past six months."
  5. Family/Group: Products that can be grouped by a family group name or number. Clients have family or group identifiers for these groupings, which can be used in Audience Recommender queries. For example: "I want all active members who purchased items in the family group called AQUAF" or "Active members who purchased items from family group 452405 in the past year."
  6. Vendors: Generates the audience based on specified product vendors; can include primary vendor name or ID number. For example: "Members who purchased Jamieson products during the past year" or "Members who bought products by vendor 525255 in the past year."
  7. Multiple Conditions: The query includes more than one of the product data types listed above. For example: "All members who purchased batteries and cosmetics in the last three months" or "All members who purchased from sub-category 52435 and from department 5245 in the past year."

By including product data in targeting, marketers can focus on the success of particular products and categories within the loyalty program.

Generate an Audience Based on SMS Subscription Status​

The Audience Recommender can access information about SMS subscription status for members. This feature allows SMS subscription data to be used in queries, as long as using subscription status for targeting is explicitly allowed in the configuration file. The query can also include whether or not a phone number is on file for the member. For example:

  • "I want all active members who have their SMS subscription status for short code 'X' as either opted in or pending and who have provided their phone number."
  • "I want all active members who have their SMS subscription status for short code 'X' as opted out and who have a phone number added to their account."
  • "I want all active members who have provided a phone number in their profile."

By including SMS subscription status in targeting, marketers can direct messages to members with a particular status to provide them with tailored rewards or encourage an action such as providing their phone number and opting in to an SMS subscription.

Using a Generated Audience in an Offer​

  1. When creating a new promotion, go to the Targeting section.
  2. In the Audience Targeting section, click the Specific Audiences button.
  3. When asked how you would like to select your audience, choose the Audiences option.
  4. Select an audience from the dropdown and choose the audience you created from the list of pre-configured audiences.
  1. Complete the remaining fields for your offer and publish your promotion by clicking Publish.

See also​

Manage an Audience

Last updated on