Use the Supervisory Controls pages to increase or decrease the likelihood of recommending a category, brand, product, or promotion. You can apply changes to multiple items in the current view or reset all items to their default recommendations.
This screenshot shows the supervisory controls Products page with two categories, one that's set to never offer, and the other with the default setting.
Guidelines for Adjusting Recommendations
You can adjust the likelihood of specific types of recommendations being shown by boosting or constraining on the slider. Boosting doubles the likelihood of a recommendation whereas constraining reduces the likelihood by half. You can also choose to never offer a specific product, brand, category, or promotion, which overrides any boost or constraint setting.
Choosing to never offer also overrides any policy settings. Each time you make an adjustment, you must apply your changes.
This screenshot shows the slider to boost or constrain the likelihood of recommendations and the Never Offer option.
When to Make Adjustments
Before making any adjustments, it’s important to understand the impact that they could make on the machine learning outcome. So it's best to make adjustments only in special circumstances when you know a crucial piece of information that the machine learning doesn't know.
All adjustments you make will override the decisions that the adaptive intelligence models carefully calculated. And your adjustments can affect recommendations in ways you don’t expect. For example, if you boost a product based on incorrect market research, shoppers may ignore the recommendations. A low accept rate leads to less frequent recommendations. The end result could be fewer recommendations that before you boosted the product.
The Reset All Option
Clicking Reset All reverts every boost and constrain adjustment in the application back to the default values, even the Never Offer option. You might want to use Reset All after making adjustments that are no longer relevant to restore the default values provided by the adaptive intelligence models.
Adjustments at Different Levels
You can adjust product recommendations at a category, brand, or individual product level. If you adjust at a category level, the adjustment applies to all products in that category. The same is true for the brand or promotion level. Adjustments apply to products in that brand or promotion. Adjustments for an individual product only affect that product.
How the Never Offer Option Affects Promotions
A promotion containing even just one category, brand, or product that's set to Never Offer won't be recommended. For example, promotions aren't recommended in these scenarios:
A promotion includes multiple products, and one or more is set to Never Offer.
A promotion includes products in multiple categories, but one category is set to Never Offer.
A promotion includes products of multiple brands, but one brand is set to Never Offer.
Set policies for price threshold, inventory threshold, and brand exclusivity to control product and brand recommendations. The policies you set will filter the recommendations derived by the adaptive intelligence models and machine learning.
How Policies Affect Recommendations
Policies can control recommendations at the site level. You can add policies based on:
Minimum product price
Minimum product inventory
Brand exclusivity rules
These policies and rules have higher priority than boosts and constraints, with one exception. If a product or brand is set to never offer, it's never recommended. It won't even be recommended if it fits within any policies or rules that you defined.
Create Brand Exclusivity Policies
Use brand exclusivity policies to better control which types of products are recommended together in a product carousel widget.
You can create as many brand exclusivity policies that you need, and even create policies with seemingly redundant values. For example, setting the same brand in both lists can prevent products from a boosted brand from being shown more than once in the carousel.
To create brand exclusivity policies:
On the Policies page, click Add Policy.
Select a brand in the first list, and then select the brand to exclude when the first brand is recommended.
Repeat these steps to add additional policies.
To remove a policy, click Delete in the row that you want to remove, and then save your changes.
To edit an existing policy, click Edit in the row that you want to change, and then save your changes.
Examples of Supervisory Controls
It's best to only apply supervisory control in special circumstances when you know a crucial piece of information that the machine learning doesn't know. These examples here are some of these special circumstances.
New Product or Product Line
You've added a new product or product line that market research has suggested it could likely be very popular with consumers. To bring more initial visibility to this product, you boost it. Boosting recommendations gives the product a better chance of visibility against existing products with historically high purchase rates.
New Marketing Campaign
You have a new promotion for one free item of a product when purchasing another related item. Your experience and research suggests that the savings offered by the new promotion will make both products more attractive to consumers. And you think the cost savings will ultimately raise their purchase rates. You boost the promotion, and optionally, the products associated with the promotion.
Too Much Inventory
You have an excessive inventory of a particular product. You set the competing products to never offer, which increases recommendations for the product with high inventory.
You have several items in your inventory that might run out soon. To prevent purchases of those items when they could be out of stock, you create an inventory threshold policy. The policy excludes products with inventory fewer than a specified number from being offered.
Change in Marketing Strategy
Your marketing approach has recently changed and now you want to emphasize and place more attention on higher-priced items. To prevent lower-priced items from being recommended, you create a price policy to exclude offers for items under the specified price.
Because the school season is about to resume, you want to increase sales of individual back-to-school items. You boost selected products for the four weeks before school starts and then you reset them to their default values.
You have holiday items that you don't want recommended because the holiday has passed, but you don't want to hide them completely from potential consumers. To recommend them less often and save inventory for the holiday season, you constrain the holiday category.
There was a defect with a product and you want to ensure it won't be offered to any consumers. Until the defect can be repaired and the defective product restocked, you set the Never Offer option for the product, which stops all recommendations for it.
Dominating Brand in Carousel Widgets
You have good reason for boosting a specific brand, but you don’t want it to dominate the carousel widget as a result. To prevent the boosted brand from being shown in the carousel more than once, you create a brand exclusivity policy with the brand in both lists.
You can filter promotions on the Promotions page by the promotion type. The promotions you see are imported from the commerce application, which is where they're created and managed.
|Promotion Type||Description||Discount Examples|
|Amount Discounts||For price discounts off the original price.||$5 off
Buy boots, get $5 off socks
|Percent Discounts||For percentage discounts off the original price.||10% off
Buy boots, get a second pair for 50% off
|Fixed Price||For a new price that's discounted from the original price.||Now only $19.99
Buy boots, get luxury socks for $5 a pair
|Free||For free, normally tied with another offer.||Free shipping over $75
Free sample with purchase
Buy boots, get a free pair of luxury socks