Criteria in 蜜豆视频 Target Recommendations control the content of your Recommendations activities. Create criteria to show the recommendations that are most appropriate for your activity. These criteria use the visitor鈥檚 actions to determine which content or products to display.
The following sections explain how to create a new criteria.
Access the Create New Criteria screen
There are multiple ways to reach the Create New Criteria screen. Some screen options vary depending on how you reach the screen.
On the Recommendations > Criteria library screen, click Create Criteria > Create Criteria. Criteria you create here are automatically made available for all Recommendations activities.
When you are creating a Recommendations activity using the Visual Experience Composer (VEC), you are immediately taken to the Select Criteria screen after you select an element on your page and click Replace w/ Recommendations, Insert Recommendations Before, or Insert Recommendations After. You can then select an available criteria or you can click Create Criteria. If you create a new criteria, you have the option to save your criteria for use with other Recommendations activities. For more information, see Create a Recommendations activity.
When you are editing a Recommendations activity, click in a Recommendations Location box on your page, and select Change Criteria. On the Select Criteria screen, click Create Criteria. You will have the option to save your new criteria for use with other Recommendations activities.
The following steps assume you access the Create New Criteria screen by using the first method: the Recommendations > Criteria library screen.
Click Recommendations > Criteria.
Click Create Criteria > Create Criteria.
Configure the information in the following sections.
Basic Information info
Type a Criteria Name.
This is the 鈥渋nternal鈥 name used to describe the criteria. For example, you might want to call your criteria 鈥淗ighest margin products,鈥 but you don鈥檛 want that title to display publicly. See the next step to set the public-facing title.
Type a public-facing Display Title to appear on the page for any recommendations that use this criteria.
For example, you might want to display 鈥淧eople who viewed this viewed that鈥 or 鈥淪imilar products鈥 when you use this criteria to show recommendations.
Type a short Description of the criteria.
The description should help you identify the criteria and might include information about the purpose of the criteria.
Select an industry vertical based on the goals of your recommendations activity.
table 0-row-2 1-row-2 2-row-2 3-row-2
Industry Vertical
Goal
Retail/Ecommerce
Conversion resulting in purchase
Lead Generation/B2B/Financial Services
Conversion with no purchase
Media/Publishing
Engagement
Other criteria options will change based on the industry vertical you select.
Select a Page Type.
You can select multiple page types.
Together, the industry vertical and page types are used to categorize your saved criteria, making it easier to reuse criteria for other Recommendations activities.
Make recommendations based on the user鈥檚 cart contents.
People Who Viewed These, Viewed Those
People Who Viewed These, Bought Those
People Who Bought These, Bought Those
Popularity-Based
Make recommendations based on the overall popularity of an item across your site or based on the popularity of items within a user鈥檚 favorite or most-viewed category, brand, genre, and so forth.
Most Viewed Across the Site
Most Viewed by Category
Most Viewed by Item Attribute
Top Sellers Across the Site
Top Sellers by Category
Top Sellers by Item Attribute
Top by Analytics Metric
Item-Based
Make recommendations based on finding similar items to an item that the user is currently viewing or has recently viewed.
People Who Viewed This, Viewed That
People Who Viewed This, Bought That
People Who Bought This, Bought That
Items with Similar Attributes
User-Based
Make recommendations based on the user鈥檚 behavior.
Recently Viewed Items
Recommended for You
Custom Criteria
Make recommendations based on a custom file you upload.
Custom Algorithm
note note
NOTE
If you select Items/ Media with Similar Attributes, you will have the option to set content similarity rules.
As required, select an Item Attribute and Profile Attribute to Match, a Recommendation Key, Filtering Key, and/or Analytics Metric to configure the algorithm.
The remaining algorithm configuration options vary depending on the selected algorithm. To finish configuring the algorithm, select a Recommendation Key, Filtering Key, Co-Occurrence Basis, Analytics Metric, and/or Item Attribute and Profile Attribute to Match.
Select the desired Behavioral Data Source: 蜜豆视频 Target or Analytics.
note note
NOTE
The Behavioral Data Source section displays only if your implementation uses Analytics for Target (A4T).
If you chose Analytics, select the desired report suite.
If the criteria uses 蜜豆视频 Analytics as the behavioral data source, once created, the time for criteria availability depends on whether the selected report suite and lookback window has been used for any other criteria, as explained below:
One-time report suite setup: The first time a report suite is used with a given data range lookback window, Target Recommendations can take from two to seven days to fully download the behavioral data for the selected report suite from Analytics. This time frame is dependent on the Analytics system load.
New or edited criteria using an already available report suite: When creating a new criteria or editing an existing criteria, if the selected report suite has already been used with Target Recommendations, with a data range equal to or lesser than the selected data range, then the data is immediately available and no one-time setup is required. In this case, or if an algorithm鈥檚 settings are edited while not modifying the selected report suite or data range, the algorithm runs or re-runs within 12 hours.
Ongoing algorithm runs: Data flows from Analytics to Target Recommendations on a daily basis. For example, for the Viewed Affinity recommendation, when a user views a product, a product-view tracking call is passed into Analytics close to real-time. The Analytics data is pushed to Target early the next day and Target runs the algorithm in less than 12 hours.
Set the Lookback Window to determine the time range of available historical user behavior data to use when determining which recommendations to show. This option is available for all algorithms with the exception of Items with Similar Attributes and Custom Algorithms.
If your site has a lot of traffic and behaviors change frequently, choose a shorter data window. A shorter window enables Recommendations to be more responsive to changes in the market and in your business. For example, a shorter window means that Recommendations will detect changes in visitor behavior as your visitors begin seasonal shopping, such as back-to-school shopping or Christmas, and will recommend items appropriate to those shopping seasons.
If you don鈥檛 have a lot of data, or visitor behavior does not change frequently, you might select a longer window. However, for many sites, a shorter window results in higher-quality recommendations.
Popularity-Based algorithms when the selected Behavioral Data Source is 蜜豆视频 Target
One day
Algorithm runs every 12-24 hours
Popularity-Based algorithms
Two days
Algorithm runs every 12-24 hours
Popularity-Based algorithms
Item-Based algorithms
User-Based algorithms
Cart-Based algorithms
One week
Algorithm runs every 24-48 hours
Popularity-Based algorithms
Item-Based algorithms
User-Based algorithms
Cart-Based algorithms
Two weeks
Algorithm runs every 24-48 hours
Popularity-Based algorithms
Item-Based algorithms
All User-Based algorithms
Cart-Based algorithms
One month (30 days)
Algorithm runs every 24-48 hours
Popularity-Based algorithms
Item-Based algorithms
User-Based algorithms
Cart-Based algorithms
Two months (61 days)
Algorithm runs every 24-48 hours
Popularity-Based algorithms
Item-Based algorithms
User-Based algorithms
Cart-Based algorithms
Backup Content content
Backup Content rules determine what happens if the number of recommended items does not fill your recommendations design. It is possible for Recommendations criteria to return fewer recommendations than your design calls for. As an example, if your design has slots for four items, but your criteria causes only two items to be recommended, you can leave the remaining slots empty, you can use backup recommendations to fill the extra slots, or you can choose to display no recommendations.
(Optional) Slide the Partial Design Rendering toggle to the 鈥渙n鈥 position.
As many slots as possible will be filled but the design template might include blank space for remaining slots. If this option is disabled and there is not enough content to fill all available slots, recommendations are not served and default content is displayed instead.
Enable this option if you want recommendations served with blank slots. Use backup recommendations if you want recommendation slots to be filled with content based on your criteria with empty slots filled with similar or popular content from your site, as explained in the next step.
(Optional) Slide the Show Backup Content toggle to the 鈥渙n鈥 position.
Fill any remaining empty slots in the design with a random selection of most-viewed products from across your site.
Using backup recommendations ensures that your recommendation design fill all available slots. Suppose that you have a 4 x 1 design, as illustrated below:
Suppose your criteria causes only two items to be recommended. If you enable the Partial Design Rendering option, the fist two slots are filled, but the remaining two slots remain empty. However, if you enable the Show Backup Recommendations option, the first two slots are filled based on your specified criteria and the remaining two slots are filled based on your backup recommendations.
The following matrix shows the result you鈥檒l observe when using the Partial Design Rendering and Backup Content options:
table 0-row-3 1-row-3 2-row-3 3-row-3 4-row-3
Partial Design Rendering
Backup Content
Result
Disabled
Disabled
If fewer recommendations are returned than the design calls for, the recommendations design is replaced by default content and no recommendations are displayed.
Enabled
Disabled
The design is rendered, but may include blank space if fewer recommendations are returned than the design calls for.
Enabled
Enabled
Backup recommendations will fill available design 鈥渟lots,鈥 fully rendering the design. If applying inclusion rules to backup recommendations restricts the number of qualifying backup recommendations to the point that the design cannot be filled, the design is partially rendered. If the criteria does not return any recommendations, and inclusion rules restrict backup recommendations to zero, the design is replaced with default content.
Disabled
Enabled
Backup recommendations will fill available design 鈥渟lots,鈥 fully rendering the design. If applying inclusion rules to backup recommendations restricts the number of qualifying backup recommendations to the point that the design cannot be filled, the design is replaced by default content and no recommendations are displayed.
Use Content Similarity rules to make recommendations based on item or media attributes.
NOTE
If you selected Item-Based/ Media with Similar Attributes as your Algorithm Type and Algorithm, you have the option to set content similarity rules.
Content similarity compares item attribute keywords and makes recommendations based on how many keywords different items have in common. Recommendations based on content similarity do not require past data to deliver strong results.
Using content similarity to generate recommendations is especially effective for new items, which are not likely to show up in recommendations using People Who Viewed This, Viewed That and other logic based on past behavior. You can also use content similarity to generate useful recommendations for new visitors, who have no past purchases or other historical data.
When you select Item-Based/ Media with Similar Attributes, you have the option to create rules to increase or decrease the importance of specific item attributes in determining recommendations. For items such as books, you might want to boost the importance of attributes like genre, author, series, and so on, to recommend similar books.
Because content similarity uses keywords to compare items, some attributes, such as message or description, can introduce 鈥渘oise鈥 into the comparison. You can create rules to ignore these attributes.
By default, all attributes are set to Baseline. You do not need to create a rule unless you want to change this setting.
NOTE
The content similarity algorithm might use random sampling in computing similarity between items. As a result, similarity ratings between items might vary between algorithm runs.
Inclusion Rules inclusion
Several options help you narrow the items that display in your recommendations. You can use inclusion rules while creating criteria or promotions.
Inclusion rules are optional; however, setting these details gives you more control over the items that appear in your recommendations. Each detail you configure further narrows the display criteria.
For example, you can choose to display only women鈥檚 shoes that have an inventory of more than 50 and a price between $25 and $45. You can also weight each attribute so those items that are more important to your business are most likely to appear.
As another example, you can choose to display job openings to visitors who visit your site only from certain cities and who have the required college degrees.
Inclusion rule options vary by industry vertical. By default, inclusion rules are applied to backup recommendations.
IMPORTANT
You should use inclusion rules cautiously. They are useful if, for example, your organization has rules that demand that one brand is not recommended while another brand is being shown. However, there is an opportunity cost to this feature. You could possibly lose a percentage of lift by restricting some items from not showing when they would normally be shown by the activity criteria.
The inclusion rules are joined with an AND. All rules must be met to include an item in a recommendation.
To create a simple inclusion rule, as mentioned previously, to display only women鈥檚 shoes that have an inventory of more than 50 and a price between $25 and $45, perform the following steps:
(Conditional) Slide the Allow recently purchased items to be recommended? toggle to the 鈥渙n鈥 position.
This setting is based on the productPurchasedId. The default behavior is to not recommend previously purchased items. In most cases you do not want to promote items a customer has recently purchased. It is useful if you sell items that people typically purchase only once, such as kayaks. If you sell items that people come back to purchase again on a repeated basis, such as shampoo or other personal items, you should enable this option.
Set a price range for the products you want to recommend.
Set the minimum inventory amount for the products you want to recommend.
Configure the recommendation to display items only when they meet certain criteria.
You can specify that items are included only when one of the attributes in the list meets or does not match one or more specified conditions.
The available evaluators depend on the value you choose in the first drop-down. You can list multiple items. These items are evaluated with OR.
Multiple rules are combined with an AND.
note note
NOTE
This option limits the items that are displayed in the recommendation. It does not affect which pages the recommendation is displayed on. To limit where the recommendation displays, select the pages in the experience composer.
You can add multiple rules to 鈥渘udge鈥 the algorithm based on important information or metadata about the content catalog so that certain items are more likely to be shown.
For example, you can apply a higher weighting to on-sale items so they appear more often in the recommendation. Non-sale items are not completely excluded, but they appear less often. Multiple weighted attributes can be applied to the same algorithm, and the weighted attributes can be tested on split traffic in the recommendation.
Choose a value.
The value determines the type of item that is more likely to display, based on one of several available criteria.
Choose an evaluator.
Type the keyword to complete the rule attributes.
For example, the complete rule might be 鈥淐ategory contains substring shoes.鈥
Select the weight to assign to the rule.
Options range from 0 to 100 in increments of 25.
Add additional rules if desired.
When finished, click Save.
If you are creating a new Recommendations activity or editing an existing one, the Save criteria for later check box is selected by default. If you do not want to use the criteria in other activities, clear the check box before saving.
Training video: Create criteria in Recommendations (12:33)
